The Economic Cost of Methamphetamine Use in the United States,

http://www.rand.org/content/dam/rand/pubs/monographs/2009/RAND_MG829.pdf#statistics%27
Preface
This monograph presents the first national estimate of the economic cost of methamphetamine
(meth) use in the United States. Our analysis suggests that the economic cost of meth use in
the United States reached $23.4 billion in 2005. Given the uncertainty in estimating the costs
of meth use, this study provides both a lower-bound estimate of $16.2 billion and an upper-bound estimate of $48.3 billion.
The analysis undertaken to generate these estimates considers a wide range of conse-quences due to meth use, including the burden of addiction, premature death, drug treat-ment, and aspects of lost productivity, crime and criminal justice, health care, production
and environmental hazards, and even child endangerment. There are other potential harms
due to meth, however, that could not be included either due to a lack of scientific evidence
or due to data issues. Although meth causes some unique harms, the study finds that many
of the primary cost drivers are similar to those identified in economic assessments of other
illicit drugs. Among the most costly elements are the intangible burden of addiction and
premature death, which account for nearly two-thirds of the economic costs. The intangible
burden of addiction measures the lower quality of life (QoL) experienced by those addicted
to the drug. Crime and criminal justice costs also account for a significant share of economic
costs. These costs include the burden of processing and incarcerating drug offenders as well
as the costs of additional nondrug crimes generated by meth use. Other costs that signifi-cantly contribute include lost productivity, the costs of removing a child from his or her par-ents’ home due to meth, and the cost of drug treatment. One unusual cost captured in the
analysis is the cost associated with the production of meth. Producing meth requires toxic
chemicals that can result in fire, explosions, and other events.
The study was sponsored by the Meth Project Foundation, a nonprofit group dedicated to
reducing first-time meth use. Additional research support was provided by the National Insti-tute on Drug Abuse.
The RAND Drug Policy Research Center
This study was carried out under the auspices of the Drug Policy Research Center, a joint
endeavor of RAND Infrastructure, Safety, and Environment and RAND Health. The goal of
the Drug Policy Research Center is to provide a firm, empirical foundation on which sound
drug policies can be built at the local and national levels. The center’s work has been sup-ported by the Ford Foundation, other foundations, government agencies, corporations, and
individuals.
iv The Economic Cost of Methamphetamine Use in the United States, 2005
Questions or comments about this monograph should be sent to the project leader, Rosa-lie Liccardo Pacula (Rosalie_Pacula@rand.org). Information about the Drug Policy Research
Center is available online (http://www.rand.org/multi/dprc/). Inquiries about research projects
should be made to the center’s co-directors, Rosalie Liccardo Pacula (Rosalie_Pacula@rand.
org) and Beau Kilmer (Beau_Kilmer@rand.org).
v
Contents
Preface........................................................................................................... iii
Tables............................................................................................................ ix
Summary........................................................................................................ xi
Abbreviations................................................................................................. xvii
CHAPTER ONE
Introduction.....................................................................................................1
General Approach to the Study.................................................................................3
Included and Excluded Costs...................................................................................5
Organization of This Monograph..............................................................................6
CHAPTER TWO
The Cost of Methamphetamine Treatment................................................................9
Care Received in the Specialty Sector.........................................................................9
Costs of Hospital-Based Drug Treatment....................................................................14
Other Federal Treatment.......................................................................................16
CHAPTER THREE
The Cost of Methamphetamine-Related Health Care Among Methamphetamine Users........19
A Literature Review.............................................................................................21
Health Service Costs Associated with Amphetamine Use................................................ 24
Amphetamine-Induced Conditions........................................................................ 24
Amphetamine-Involved Conditions....................................................................... 26
Suicide Attempts............................................................................................. 28
Emergency-Department Care.............................................................................. 30
Health Administration and Support.......................................................................31
Limitations.......................................................................................................32
CHAPTER FOUR
Premature Death and the Intangible Health Burden of Addiction..................................33
Premature Death................................................................................................35
Number of Methamphetamine-Related Deaths..........................................................35
Placing a Value on Premature Mortality.................................................................. 36
The Cost of the Health Burden Associated with Being Addicted........................................ 38
Limitations...................................................................................................... 42
vi The Economic Cost of Methamphetamine Use in the United States, 2005
CHAPTER FIVE
Productivity Losses Due to Methamphetamine Use................................................... 43
Literature Review.............................................................................................. 44
Lower Probability of Being Employed..................................................................... 46
Absenteeism................................................................................................... 48
Lost Work Due to Incarceration............................................................................52
Employer Costs of Hiring Methamphetamine Users...................................................... 54
Limitations.......................................................................................................55
CHAPTER SIX
The Cost of Methamphetamine-Related Crime..........................................................57
Cost of Methamphetamine-Specific Arrests.................................................................58
Methamphetamine Offenses at the State and Local Levels..............................................58
Methamphetamine Offenses at the Federal Level........................................................62
Cost of Community Corrections Revocations............................................................. 64
Cost of Methamphetamine-Induced Crime.................................................................67
Background: Methamphetamine’s Link to Property and Violent Crime..............................67
Approach and Results........................................................................................69
CHAPTER SEVEN
The Methamphetamine-Related Cost of Child Maltreatment and Foster Care....................75
Social Costs of Child Maltreatment..........................................................................75
Attributing Child Maltreatment and Neglect to Methamphetamine................................... 77
A Crude Calculation of the Cost of Methamphetamine to the Foster-Care System....................78
Number of Children Entering the Foster-Care System in 2005........................................78
Cost of Sending a Child to Foster Care....................................................................79
Medical, Mental Health, and Quality-of-Life Costs for Victims of Abuse and Neglect...............81
Limitations.......................................................................................................82
CHAPTER EIGHT
The Societal Cost of Methamphetamine Production...................................................85
Physical Injury from Lab Mishaps............................................................................85
Number of Events and Events with Victims.............................................................. 87
Calculating the Costs of Injuries and Deaths............................................................ 88
Lab-Cleanup Cost ..............................................................................................89
Other Costs Associated with Production: Decontamination, Evacuation, Sheltering, and
Hazardous Waste...........................................................................................91
CHAPTER NINE
Consideration of Costs Not Included......................................................................93
Treatment....................................................................................................... 94
Health........................................................................................................... 94
Cost of Health Care Received Outside the Inpatient Setting...........................................95
Other Health Care Costs for Conditions Caused by Methamphetamine Use........................95
Dental Costs.................................................................................................. 96
Productivity..................................................................................................... 97
Contents vii
Health Care and Workers’ Compensation Costs......................................................... 97
Other Productivity-Related Losses......................................................................... 98
Crime and Criminal Justice.................................................................................. 98
Incarceration for Misdemeanor Possession............................................................... 98
Violent Crime................................................................................................ 99
Identity Theft................................................................................................. 99
Other Crime Issues.........................................................................................100
Child Maltreatment and Foster Care.......................................................................100
Methamphetamine Production.............................................................................. 101
Decontamination, Evacuation, and Sheltering.......................................................... 101
Hazardous Waste........................................................................................... 101
Summary....................................................................................................... 102
CHAPTER TEN
Conclusion.................................................................................................... 103
External Versus Internal Costs.............................................................................. 105
To Conclude...................................................................................................106
APPENDIXES
A. Supporting Information for Estimating the Cost of Methamphetamine-Related
Health Care: Inpatient Days.......................................................................... 109
B. Additional Calculations to Support Productivity-Loss Estimates............................. 111
C. Additional Information to Support the Cost of Methamphetamine-Related Crime...... 119
D. Deriving Methamphetamine Attribution Factors from the Inmate Surveys................ 125
Bibliography...........................................................................................
Summary
After marijuana, amphetamines are the most widely used illicit drug worldwide (UNODC,
2008). The United Nations estimates that amphetamine users nearly equal the number of
cocaine and heroin users combined (25 million versus 28 million). In the United States, the
recent increase in the prevalence of amphetamines, particularly methamphetamine (meth), is
cause for concern. The meth situation in the United States is a complicated story of conflicting
indicators, however. On the one hand, national reporting systems monitoring drug use among
household- and school-based populations suggest that meth is a relatively minor drug of con-cern (NSDUH, 2006; Johnston and O’Malley, 2007). According to the National Survey on
Drug Use and Health (NSDUH), only 0.5 percent of the household population reported use
in 2005, far below prevalence rates for marijuana and cocaine and slightly lower than those for
heroin (NSDUH, 2006). Similarly, reports from high-school seniors suggest that meth abuse
is relatively minor, with annual prevalence rates in 2005 of just 2.5 percent as compared with
rates of 5.1 percent for cocaine, 9.7 percent for hydrocodone (e.g., Vicodin®), and 33.6 percent
for marijuana (Johnston and O’Malley, 2007).
On the other hand, regional data systems, law-enforcement agencies, and county hospitals
indicate that meth is the most significant problem facing the populations they serve (NDIC,
2007b; NACO, 2005, 2006; NIDA CEWG, 2005). According to information reported by
the National Institute on Drug Abuse Community Epidemiology Work Group, in 2004,
meth was the primary drug of abuse in 59 percent of treatment admissions in Hawaii, 51
percent of treatment admissions in San Diego, and 38 percent of treatment admissions for the
entire state of Arizona. (These percentages all exclude treatment for alcohol abuse.) In 2005,
39.2 percent of reporting state and local law-enforcement agencies cited meth as their greatest
drug threat, exceeding the percentage reporting cocaine and crack to be the greatest threat
(35.3 percent) (NDIC, 2007a).
While it is clear that the prevalence of meth problems is greater in western and rural
states, there is evidence of a national problem. Data from the Treatment Episode Data Set
(TEDS) show that treatment admissions for meth more than doubled nationally between
2000 and 2005 (NDIC, 2007b). Growth in amphetamine-related treatment admissions,
which are dominated by meth, increased in every region between 1992 and 2005. Further-more, information from the 2005 Drug Abuse Warning Network (DAWN), which monitors
drug-related emergency-department (ED) episodes, reveals that stimulant-related admissions,
including those due to meth, are just as likely as heroin admissions nationally once the margin
of statistical error is taken into account. Yet, meth remains far less part of national discussions
than are cocaine, heroin, marijuana, and prescription-drug abuse. Meth has not been the focus
of media campaigns like the Marijuana Prevention Initiative. And only with the 2006 reautho-
xii The Economic Cost of Methamphetamine Use in the United States, 2005
rization did the National Youth Anti-Drug Media Campaign require that at least 10 percent of
the fiscal year 2007 appropriation focus on reducing meth use.
Meth is a highly addictive substance that can be taken orally, injected, snorted, or smoked.
When smoked or injected, the user immediately experiences an intense sensation followed by a
high that may last 12 hours or more. Concerns about meth use arise from its highly addictive
nature and its association with a number of adverse physical effects, including hypertension
and other cardiovascular effects, seizures and convulsions, pulmonary impacts, and dental
damage. Users also suffer psychological effects, such as anxiety, irritability, and loss of inhibi-tion, which can lead to risky sexual and other types of behavior. Meth users who inject the
drug expose themselves to additional risks related to injection-drug use, including contract-ing human immunodeficiency virus (HIV), hepatitis B and C, and other bloodborne viruses.
Meth production imposes additional health risks on users and nonusers alike. The process
for producing meth, a synthetic substance, is hazardous and susceptible to fire and explosion.
Moreover, the production of each pound of meth results in 5–6 pounds of toxic by-product.
As a result, health and environmental effects may also follow from the production process.
Despite the documented harms associated with use, research to date has not attempted to
quantify the social cost or burden this drug places on society today.
In this monograph, we attempt to fill the void by building the first national estimate
of the economic burden of meth use, based on information available for 2005. We chose to
focus on estimating the burden in 2005 because it is the most recent year for which we were
able to obtain the data necessary to construct our estimate. Unlike data for other substances
of abuse, the data necessary to build such an estimate nationally are far from complete or
comprehensive. Furthermore, the scientific literature has yet to develop consistent evidence
of causal associations for many of the harms that meth is believed to cause. Thus, calcula-tion of the cost of consequences must be necessarily imprecise, as it can only reflect the state
of the knowledge available today, which will change as more research is done in particular
areas. To capture the relative imprecision of what is actually known today, we produce lower
and upper bounds of all our point estimates. The variation represented in these lower- and
upper-bound estimates reflects, in some instances, sampling variability in data sets that are
available and, in other instances, the lack of data from which to obtain a precise point esti-mate. Despite this variability, we attempt to generate a best estimate of these costs based
on what we understand of the science today. But we also point out that there are many cost
areas that contribute to the economic burden that we are unable to measure and hence have
to exclude from our calculations.
Our results are surprising for a substance that has received limited national attention.
Even before monetizing the consequences associated with meth use, we see in Table S.1 that
the toll in terms of premature death and lost well-being, as measured in quality-adjusted life-years (QALYs), is substantial. We estimate that meth use was responsible for nearly 900 deaths
in 2005 and resulted in a total loss of more than 44,000 QALYs. According to data from
the DAWN mortality publications, there were far more meth-related deaths than marijuana-related deaths in any given year (SAMHSA, 2007b). Yet, marijuana has remained the focus of
the national antidrug campaign (ONDCP, 2008).
Following previous cost-of-illness (COI) studies estimating the burden of other sub-stances of abuse, we consider in this monograph the cost of numerous consequences associ-ated with use, including the cost of meth treatment, the excess health care service utilization
Summary xiii
Table S.1
The Burden of Methamphetamine in 2005 in Terms of Premature Death and Lost Quality of Life
Burden Lower Bound Best Estimate Upper Bound
Premature mortality 723 895 1,669
Lost QALYs 32,574 44,313 74,004
associated with meth use and dependence, productivity losses due to the drug, and the cost
of meth-associated crime. In contrast to previous COI studies, however, this monograph also
considers the cost to society associated with the production of meth, the intangible burden
borne by those addicted, and child endangerment.
Using the national data available on each of these cost consequences, our best estimate
of the economic burden of meth use in the United States in 2005 is roughly $23.4 billion
(Table S.2). This figure includes the estimable costs associated with drug treatment, other
health costs, the intangible burden of addiction and premature death, lost productivity, crime
and criminal justice costs, child endangerment, and harms resulting from production.
Many of our estimates are subject to substantial uncertainty, however, so for all of our
estimates, we provide lower- and upper-bound estimates that help us better understand where
paucity of data or scientific research might influence the credibility of our single point esti-mate. The degree of uncertainty, as indicated by these lower and upper bounds, varies con-siderably across cost components, with some categories showing much greater uncertainty
than others (see Table S.2). When considered together, the uncertainty about each component
contributes to the uncertainty about the total. Taking this aggregation of uncertainties into
account, we estimate that the true economic burden is likely to be in the range of $16.2 billion
to $48.3 billion.
In reviewing the key contributors to the total estimated costs, we found that most of
meth’s costs are due to the intangible burden that addiction places on dependent users and
to premature mortality. We estimate the cost associated with these losses at $16.6 billion,
representing nearly two-thirds of our total cost estimate. The majority of these costs are due
to the intangible cost of addiction ($12.6 billion). That number is the product of the number
Table S.2
Social Costs of Methamphetamine in the United States in 2005 ($ millions)
Cost Lower Bound Best Estimate Upper Bound
Drug treatment 299.4 545.5 1,070.9
Health care 116.3 351.3 611.2
Intangibles/premature death 12,513.7 16,624.9 28,548.6
Productivity 379.4 687.0 1,054.9
Crime and criminal justice 2,578.0 4,209.8 15,740.9
Child endangerment 311.9 904.6 1,165.7
Production/environment 38.6 61.4 88.7
Total 16,237.3 23,384.4 48,280.9
NOTE: Due to rounding, numbers may not sum precisely.
xiv The Economic Cost of Methamphetamine Use in the United States, 2005
of people dependent on the drug and the monetary value of the lost quality of life (QoL), mea-sured by a reduction in QALYs. The estimate is subject to great uncertainty, because of assump-tions underlying our upper- and lower-bound estimates of the number of people dependent
and the valuation of their lost QALYs. The remaining cost ($4.0 billion) is due to premature
mortality among users. The uncertainty in this estimate is also significant, reflecting the varia-tion in assumptions as to which deaths are attributable to meth. Moreover, we caution that
both of these estimates depend on the value placed on a lost life, which, based on the litera-ture, we take to be $4.5 million. We use $4.5 million as the value of a life rather than ascribe
a range of estimates, but we provide our estimates of the number of events so that the reader
may recalculate the associated costs with alternative valuations.
Crime and criminal justice costs represent the second-largest category of costs at $2.5 bil-lion to $15.8 billion and a best estimate of $4.2 billion. Meth-specific offenses represent more
than half of these costs, totaling $2.4 billion. These are the costs associated with processing
offenders for the possession and sale of meth. Meth-induced violent and property crimes that
are generally attributable to actions of people under the influence of meth or in need of meth
represent an additional $1.8 billion in costs. Finally, an additional $70 million is due to parole
and probation violations for meth offenses. It is possible that these costs are significantly under-estimated, however, as the scientific literature regarding the causal association between meth
and property and violent crime is inconclusive. We conducted our own analyses to explore the
causality and association and, in our best estimate, find sufficient support to include an esti-mate of meth-induced property crime, but not violent crime. The very large bounds on this
element are due to alternative assumptions of causality that warrant additional research.
Costs associated with productivity losses represent another substantial category of costs.
The best estimate for total productivity losses is $687 million. Most of the productivity losses
are due to absenteeism ($275 million) and incarceration ($305 million). Smaller contribu-tors are the costs due to a lower probability of working among meth users ($63 million) and
the cost of employer drug testing ($44 million). We do not attempt to estimate any losses
embodied in the potential changes in wages paid to meth users vis-à-vis nonusers. Nor can
we include any estimates of the higher health care and workers’ compensation costs paid by
employers because of employees’ meth use.
We calculate the costs associated with drug treatment at approximately $545 million,
almost all of it in the community-based specialty treatment sector ($491 million). The total
also includes $39 million of federally provided specialty treatment, almost entirely through the
Indian Health Service (IHS) and the U.S. Department of Veterans Affairs (VA), and $15 mil-lion for treatment received in general short-stay hospitals. We did not have access to data on
the cost of treatment received in the general, non–hospital-based medical sector, so these costs
are omitted.
We estimate approximately $351 million for additional health care costs among meth
users. These include $27 million for hospital admissions induced by the use of meth, $14 mil-lion for the incremental costs of caring for patients admitted for another cause but whose con-ditions are exacerbated by meth use, $46 million for ED care of meth patients not admitted to
the hospital, and $14 million for hospital inpatient care of suicide attempts to which meth use
is a likely contributor. The largest contribution is an additional $250 million for health admin-istration and support. The health care total is likely an underestimate because it includes only
the incremental costs for other conditions even though a share of those conditions may have
been caused by meth and meth-induced behaviors.
Summary xv
Child-endangerment costs are estimated at $905 million. Our estimates are limited to
children who are removed from their homes by the foster-care system, so these costs are likely
an underestimate of the full burden of meth abuse. Substance abuse is a key contributing factor
in two-thirds of those removals, though we must make some assumptions about the specific
role of meth. The largest contributor to these costs is the medical, mental, and QoL losses suf-fered by children ($502 million), though the burden on the foster-care system is similar in size
($403 million). The substantial uncertainty derives from the uncertainty regarding how many
of those substance-associated removals are related to meth and our inability to measure accu-rately the cost of these episodes.
Potentially unique costs of meth are the harms associated with production. We estimate
the social costs associated with the meth-production process at $61 million. About half of
those costs are due to injuries and deaths from hazardous-substance events, such as explosions
and fires ($32 million). About half the casualties are suffered by responding emergency person-nel, but the more serious and costly events are not suffered by first responders. The other half of
the production costs are due to cleanup of hazardous wastes at discovered laboratory facilities
($29 million). The substantial range—from $39 million to $89 million—results largely from
uncertainty in estimating the number of deaths attributable to meth production.
While our methods are not completely comparable to those of other prevalence-based
COI studies for the abuse of other drugs or drugs in general (e.g., Mark, Woody, et al., 2001;
ONDCP, 2004b), our results are similar in a few key ways. The major cost drivers for meth,
if we ignore the intangible burden of addiction (which is omitted from other estimates of the
cost of drug use), are similar to those for other illicit drugs, with losses associated with prema-ture death and crime being major components. Importantly, if we take out the intangible cost
of addiction from our estimate (representing $12.6 billion), our revised estimate of the total
economic burden of meth use ($10.8 billion) represents 5.5 percent of the total cost of illicit
drug use in the United States reported by the ONDCP (2004b). While this might seem like
a small fraction, it represents a greater share of the economic burden than simple consump-tion rates would suggest. According to annual prevalence data from the NSDUH, meth users
represent only 3.7 percent of all illicit-drug users (NSDUH, 2006). If we use our upper-bound
estimate of the cost of meth use (still excluding intangibles), we find that meth users represent
7.2 percent of the total cost of illicit drug use, approximately twice the burden suggested by
consumption alone.
If we consider the cost per meth user, our best estimate translates into $26,872 for each
person who used meth in the past year or $74,408 for each dependent user. The per capita cost,
like our overall estimate, is sensitive to the inclusion of the intangible burden of addiction. If
we ignore this intangible cost, the costs in the past year for each user and meth-dependent user
are appreciably smaller at $12,395 and $34,322, respectively. This suggests that the average
cost per heavy meth user (ignoring intangible cost of addiction) is at least 75 percent of the
estimated average cost per heroin addict in 2005 dollars (Mark, Woody, et al., 2001).
While this monograph highlights key components of the costs of meth that we were able
to quantify, further research is needed in a number of areas before a true accounting of the
full economic burden can take place. Throughout this monograph, we identify and (when-ever possible) provide evidence on the potential magnitude of various cost components that
we are unable to include explicitly in our estimate because of data issues and inconclusive sci-ence regarding causality. These are obvious areas in which more research would be fruitful.
Specific areas that are likely to translate into substantial costs in terms of the overall burden
xvi The Economic Cost of Methamphetamine Use in the United States, 2005
include meth-associated crime, child endangerment in non–foster-care settings, employer costs
of hiring meth workers, and health care costs associated with treating meth-induced health
problems. Estimates of the total cost and its components do not provide adequate information
to inform policymakers regarding the most effective policies to reduce the harms associated
with meth use. More research should be conducted to identify cost-effective strategies for deal-ing with the problem and reducing particular harms associated with use.
In developing policies to address meth, it is also important to recognize the key differ-ences between meth and other substances, such as the harms to persons and the environment
that result from meth’s unique production process, of which only a fraction are estimated here.
Of course, this monograph can only highlight the key components of the costs of meth, in the
hope that these will be more salient to policymakers and that their attention will be directed
to the more important aspects of the problem. This research cannot directly support one policy
or another. Further research is needed to inform policymakers on the most effective strategies
to reduce these harms.
A final insight from this study is that, in the case of meth abuse, we should be cautious
interpreting evidence from national household surveys and school-based studies as indicators
of the problem. Clearly, the burden of meth abuse is substantial, far exceeding what would be
implied by simple prevalence measures from either of these populations. Moreover, as is the
case with other substances of abuse, it is probably not the recreational meth user who imposes
the greatest burden on our society, but rather those who become addicted, engage in crime,
need treatment or emergency assistance, cannot show up for work, lose their jobs, or die pre-maturely. These are the individuals who impose the greatest cost on society, yet they are also
generally populations that are not adequately captured in household- or school-based surveys.
Abbreviations
$QALY dollar value of one quality-adjusted life-year
2SLS standard two-stage least squares
ACASI audio computer-assisted self-interview
ACF Administration for Children and Families
ADSS Alcohol and Drug Services Study
AHRQ Agency for Healthcare Research and Quality
AIDS acquired immune deficiency syndrome
ASI Addiction Severity Index
ATSDR Agency for Toxic Substances and Disease Registry
BJS Bureau of Justice Statistics
BLS Bureau of Labor Statistics
CCS Clinical Classifications Software
CDC Centers for Disease Control and Prevention
COI cost of illness
CPI consumer price index
CSAT Center for Substance Abuse Treatment
DALY disability-adjusted life-year
DASIS Drug and Alcohol Services Information System
DATCAP Drug Abuse Treatment Cost Analysis Program
DAWN Drug Abuse Warning Network
DEA Drug Enforcement Administration
DHHS U.S. Department of Health and Human Services
DoD U.S. Department of Defense
xviii The Economic Cost of Methamphetamine Use in the United States, 2005
DSM-IV Diagnostic and Statistical Manual for Mental Disorders
E-code external-cause code
EAP employee-assistance program
ED emergency department
EMT emergency medical technician
FBI Federal Bureau of Investigation
FY fiscal year
HCUP Healthcare Cost and Utilization Project
HIV human immunodeficiency virus
HSEES Hazardous Substances Emergency Events Surveillance
ICD-9 International Statistical Classification of Diseases, ninth edition
ICU intensive-care unit
IHS Indian Health Service
IV instrumental variable
LOS length of stay
LSD lysergic acid diethylamide
meth methamphetamine
MIC methamphetamine-induced caries
MILQ Multidimensional Index of Life Quality
MRSA methicillin-resistant Staphylococcus aureus
NCRP National Corrections Reporting Program
NCS National Compensation Survey
NDACAN National Data Archive on Child Abuse and Neglect
NESARC National Epidemiologic Survey on Alcohol and Related Conditions
NI not included
NIS Nationwide Inpatient Sample
NLSY National Longitudinal Survey of Youth
NSDUH National Survey on Drug Use and Health
ONDCP Office of National Drug Control Policy
PCP phencyclidine
Abbreviations xix
QALD quality-adjusted life-day
QALI quality-adjusted life index
QALY quality-adjusted life year
QoL quality of life
QWB Quality of Well-Being Scale
QWB-SA Quality of Well-Being Scale, Self-Administered
RVU relative value unit
SAMHSA Substance Abuse and Mental Health Services Administration
SATCAAT Substance Abuse Treatment Cost Analysis Allocation Template
SF-12® SG SF-12® standard gamble weighted
SIFCF Survey of Inmates in Federal Correctional Facilities
SILJ Survey of Inmates in Local Jails
SISCF Survey of Inmates in State Correctional Facilities
STD sexually transmitted disease
TC therapeutic community
TEDS Treatment Episode Data Set
UCR Uniform Crime Report
USSC U.S. Sentencing Commission
VA U.S. Department of Veterans Affairs
WHO World Health Organization
WTP willingness to pay
CHAPTER ONE
Introduction
Methamphetamine (meth), together with cocaine, heroin, and marijuana, is one of the four
illegal drugs of most consequence in the United States today. Although lifetime and annual
prevalence rates for meth use in the household population are 5 percent and 0.5 percent,
respectively (NSDUH, 2006), meth use and abuse have risen substantially in the United States
during the past 15 years. From 1992 to 2005, meth-related treatment admissions increased
more than tenfold. This increase is due to both significant geographic expansion from west to
east and demographic expansion of meth users.
Meth use remains a more significant problem in western states, though states in the Mid-west and South contribute a large share of treatment admissions for meth. Western states com-prise 65 percent of primary meth treatment admissions nationally, while the Midwest and South
contribute 19 percent and 15 percent, respectively (SAMHSA, 2008b). Amphetamine-related
treatment admissions also increased by 920 percent in the Midwest, 560 percent in the South,
455 percent in the West, and 45 percent in the Northeast between 1992 and 2002 (Dobkin
and Nicosia, forthcoming). Although meth use was originally highly concentrated among
white men, users are now increasingly female and Hispanic. The emergence of meth is also a
significant concern for the criminal justice system. The majority of county law-enforcement
agencies now report meth as their primary drug problem (NACO, 2005). Moreover, the share
of meth-related treatment admissions referred by the criminal justice system is approximately
50 percent higher than for other substances (SAMHSA, 2008b).
Concerns about meth use arise from its association with a number of adverse physical and
psychological effects. Meth users suffer from of a wide variety of physical symptoms, including
headaches and chest pain (Rawson, Huber, et al., 2002; Rawson, Anglin, and Ling, 2002). Use
has also been associated with mental health events, such as hallucinations, paranoia, and violent
behavior (Rawson, Huber, et al., 2002; Rawson, Anglin, and Ling, 2002; NIDA, 2002). There
is also concern that meth-related reductions in inhibition can result in an increase in injuries
and sexually transmitted diseases (Sheridan et al., 2006; Schepens et al., 1998; Winslow, Voor-hees, and Pehl, 2007; Shoptaw et al., 2003; Lyons, Chandra, and Goldstein, 2006).
Furthermore, meth use is supplied by a production process unique among major drugs.
Meth is a synthetic substance produced by numerous labs in a hazardous process susceptible
to fire and explosion. Although many of the superlabs
1
have migrated to Mexico in response
to tougher precursor-chemical laws, meth production continues in the United States. In addi-tion, the production of each pound of meth results in 5–6 pounds of toxic by-product (DEA,
1
Superlabsare defined as those capable of producing 10 pounds or more of meth per production cycle. See NDIC
(2005).
2 The Economic Cost of Methamphetamine Use in the United States, 2005
undated). As a result, health and environmental effects may follow from meth production,
which may affect users, nonusers, and the environment.
While estimates have been made of the overall social costs of drug abuse, with several
studies examining the specific costs of heroin and cocaine, no estimate has yet been forth-coming for meth alone. Such a cost estimate would be useful for several reasons. First, the esti-mated cost burden of meth would provide policymakers with valuable information regarding
the magnitude of the social burden that this particular drug, as compared with other drugs
of abuse, imposes on society. This information will be useful for guiding resource-allocation
decisions regarding where to spend limited drug-prevention and drug-enforcement dollars.
Second, it could provide insight regarding the need for new policy approaches aimed at reduc-ing the unique harms imposed by meth production and abuse that may not be common with
other illicit drugs. Third, it could provide policy analysts with valuable information relevant
for the construction of cost-effectiveness analyses evaluating alternative strategies to reduce the
problem of meth abuse.
With these benefits in mind, RAND researchers embarked on an effort to calculate the
annual economic burden of meth in the United States. Considering the potentially broad
scope of this task and the availability of reliable data and literature, we focused on estimating
the costs associated with the following:
meth treatment that is delivered in general, short-stay hospitals and the specialty treat- t
ment sector
health services used in the treatment of medical conditions attributed to or exacerbated t
by meth use, such as overdose, acute respiratory and cardiovascular problems, accidents
and injuries, meth-exposed infants, and mental health conditions
lost productivity due to absenteeism, unemployment, or premature death t
crimes attributable to meth use as well as criminal justice costs associated with enforcing t
meth laws
environmental and personal harms resulting from meth-related production t
meth-related child endangerment, including the burden on the foster-care system, due to t
parental involvement with meth
intangible cost of meth addiction. t
While we recognize that these domains do not capture all the costs associated with meth
use, they represent those components for which we believe that reasonably good data are avail-able and for which a preliminary national estimate could be built. Although other domains are
not included in the cost estimate, we make an effort to provide some evidence on their scope.
The calculations provided in this monograph represent an assessment of these costs based
on measures of the problem and consequences in 2005. To the extent that use has changed
since 2005, the scope of the corresponding costs will follow. But the costs associated with
this use will not necessarily increase or decrease proportionally with the level reported in this
monograph, as specific costs are tied to particular types of use (e.g., dependence versus regu-lar use) as well as assumptions regarding how that use translates into harms (e.g., crime). For
example, the cost of lab cleanup may be declining with a reduction in lab busts, but crime
among meth users may be on the rise. Similarly, the fraction of the meth-using population that
is represented by dependent or heavy users may change in proportion to light users, thereby
affecting costs differentially.
Introduction 3
General Approach to the Study
We take a prevalence-based, cost-of-illness (COI) approach to identifying and measuring the
costs associated with meth use. This approach generates a monetary value of the economic
burden to society in a given calendar year and considers the direct, indirect, and intangible
costs associated with meth use and treatment of the problem in that calendar year (Rehm et
al., 2007; Harwood, Fountain, and Livermore, 1998; Rice et al., 1990). The prevalence-based
COI approach has been used to estimate the cost burden of various physical and mental health
conditions, such as asthma, diabetes, and depression, as well as drug use.
COI studies have come under increasing criticism by academics for a variety of reasons.
First, they tend to provide a single point estimate of the measure of the problem in a manner
that suggests more precision than what is actually inherent in these calculations (Reuter, 1999;
Moore and Caulkins, 2006). Rarely are estimates of the degree of a causal relationship, the
cost of treating a medical condition, or the fraction of the population affected known with
certainty. Second, because COI estimation is embedded in the medical literature, which has
traditionally resisted placing a dollar value on pain and suffering, the calculations omit values
for these intangible costs even though the research shows they exist. The resistance to assign-ing monetary values to these intangible costs places health problems at a distinct disadvantage
when compared with programs that have other social or economic impacts, as traditions in
these literatures are much more likely to include estimates of the intangible costs. For example,
in the crime literature, it has been shown that the intangible costs of crime are frequently
more than three times the estimated tangible costs (T. Miller, Cohen, and Wiersema, 1996).
Third, COI studies applied to substance abuse in particular are inconsistent in their treatment
of future costs associated with use of the drug today. While it is common practice to include
the full net present value of future lost productivity associated with a premature death caused
by drug use in the period in which it occurs, the future costs of incarcerating a person caught
in possession or selling a drug today, for example, is not fully considered. Finally, while COI
estimates may provide information on the total magnitude of a problem for society, they say
little about who bears the brunt of those costs (government, families, or the user). Without
knowledge of who bears the burden of these costs, it is difficult to make policy recommenda-tions regarding appropriate places for government to step in.
We attempt to address some of these limitations of prior COI studies. Specifically, when
possible, we attempt to capture the uncertainty related to the causal attribution by providing
high and low estimates of the causal association between meth and these specific outcomes
when the literature or our own analyses support such an approach. In doing this, we produce
a range of estimates that enable us to capture the uncertainty inherent in the construction of
our estimate as well as in the data that underlie these estimates. Further, in our estimates, we
include an estimate of the intangible health burden associated with living with addiction and
the welfare loss associated with sending a child to foster care. In addition, we treat criminal jus-tice costs in a manner consistent with the losses associated with premature death, by incorpo-rating the full net present value of future expenditures associated with a current arrest or con-viction today.
2
Finally, in the last chapter, we provide some context for our findings and draw
a distinction between those costs that are borne by the individual and those that are borne by
2
It is not possible, however, to include the full present costs for components that are less understood, such as long-term
health problems of those convicted.
4 The Economic Cost of Methamphetamine Use in the United States, 2005
society, so as to facilitate policy recommendations and to guide policymakers regarding areas
in which different policies might reduce the cost to society in general.
A key element of this study is identifying when or how much of each cost should be
attributed to meth. Unlike research on other substances of abuse, established attribution fac-tors for meth-specific diseases or other harms are not available.
3
Therefore, we must infer attri-bution fractions from the scientific literature or from our own analysis of existing data.
When constructing estimates from our own examination of existing data, we discovered
a number of methodological issues that are important to keep in mind. First, the data sources
used in the analyses identify meth with varying levels of precision. Some identify only stimu-lants, others only general amphetamines, and still others identify meth use specifically. When
meth is not identified separately from other stimulants and amphetamines, we make adjust-ments to the totals so that our estimates reflect only what we believe to be meth-specific cases.
A second problem is the inconsistent measurement of meth use across the data sets. Some data
sources capture any use of meth in the past month or year, while other data sources capture
more involved measures of use, such as dependence or abuse within a one-year time frame. This
creates problems in trying to identify the costs associated with a particular type of use consis-tently across data sets. Because this work focuses on the costs associated with any meth use,
we do our best to use the most appropriate measure that is tied to a given outcome (e.g., meth
dependence when looking at drug treatment, meth use when looking at productivity effects). A
third and more difficult data problem is that often we are given only the total budget allocated
to an issue (e.g., substance abuse treatment, law enforcement) and very little about how those
funds are used, specifically those targeting meth users or meth harms. In these cases, we again
attempt to use data from outside sources to help us ascertain a reasonable allocation of these
total budgets, but our estimates are fundamentally based on assumptions that allocations are
made consistently with the data we use.
Clearly, our attempts to deal with these challenges add some imprecision to the esti-mates we generate. We attempt to highlight this imprecision through the presentation of lower
and upper bounds of our best point estimates. But the bounds are influenced by more than
just the imprecision related to the issues mentioned here. There are two additional sources
of uncertainty underlying the estimates presented in this monograph. The first is statistical
uncertainty. Statistical uncertainty comes from sampling and model parameter variation that
is natural whenever a probabilistic sample is used to infer information about an entire popu-lation. Statistical uncertainty is traditionally demonstrated in statistical analyses through the
use of 95-percent confidence intervals, which specify—under the assumption of normality—
that the actual population value should be within the range provided by the interval 95 per-cent of the time.
A second form of uncertainty influencing our estimates, to which we will refer as struc-tural uncertainty, arises from factors influencing the reliability of estimates, regardless of the
sample from which they are drawn. For example, when pulling estimates from the literature
regarding the value of a crime or a lost life or the average cost for a given length of stay (LOS)
for treatment, there are substantial ranges reported. Without having access to the original data
from which these are constructed, we are left to infer ranges from those reported in the data.
3
For example, Popova et al. (2007) and Collins and Lapsley (2008) provide morbidity-attribution factors for other
substances.
Introduction 5
While we recognize that these adjustments do not address all the criticisms raised regard-ing the COI approach, they represent a significant improvement over prior analyses of the cost
of illicit drugs in the United States.
Included and Excluded Costs
In addition to explicitly addressing some of the criticisms raised about the COI methodology,
this monograph differs from previous estimates of the cost of illicit drug use in that it includes
costs that are specific to meth, such as production-related incidents that are never directly con-sidered in general estimates of the cost of drug abuse. Similarly, we explicitly consider the issue
of child endangerment, which is so clearly a potential cost of both meth production and meth
use when the persons involved are parents. As is indicated in Chapter Two, these meth-specific
costs are not trivial. Nevertheless, there are some categories of costs that should be captured in
a full assessment of the cost of meth use that are not reflected in our estimate because data are
not available on the prevalence or cost of these events. Table 1.1 provides an overview of the
cost components that we are able to include and the specific components not included. There
may well be additional cost components that are not included. The reasons for exclusion of
specific elements are discussed in the chapters that follow. When suitable data sources were not
available to generate cost estimates, we make an effort to provide evidence that speaks to the
potential scope of the component whenever possible but do not provide direct estimates.
Finally, in some cases, we can include only the current costs associated with use. As is the
case for many other COI studies, the longer-term implications are generally less understood
and therefore difficult to quantify.
Table 1.1
Costs Included in and Excluded from the Study
Cost Included Not Included (NI)
Drug treatment
Care received in specialty sector x
Care received in general hospitals x
Other care received in general medical setting x
Other care received through the U.S. Department of Veterans Affairs
(VA), Indian Health Service (IHS), Federal Bureau of Prisons, U.S.
Department of Defense (DoD)
x
Excess health service utilization
Care received in hospital settings x
Care received in other medical settings x
Dental x
Health infrastructure x
Intangible costs associated with dependence x
6 The Economic Cost of Methamphetamine Use in the United States, 2005
Cost Included Not Included (NI)
Productivity losses
Associated with premature death x
Reduced income associated with meth use
Due to increased unemployment x
Due to fewer hours worked x
Due to treatment-related absenteeism x
Due to other meth-involved absenteeism x
Due to lower wages x
Lost productivity due to incarceration x
Employer costs
Drug testing x
Work-related injury x
Higher health care and benefit costs x
Crime
Arresting and adjudicating users and sellers x
Property and violent crime by meth users (includes intangible costs) x
Nonindex and nondrug crime (e.g., identity theft) x
Incarceration for misdemeanor possession x
Crime and violence related to meth market x
Harms related to meth production
Environmental cleanup x
Physical injury and death x
Additional waste x
Personal decontamination, shelter, and evacuation x
Child endangerment
Foster-care placement x
Child malnutrition and victimization x
Other costs (e.g., adoption) x
Organization of This Monograph
Each chapter of this monograph attempts to quantify the identified meth-related costs in a
particular area (e.g., health care, productivity, crime, child endangerment, production). The
chapters begin with a summary of findings followed by a review of the peer-reviewed scientific
Table 1.1—Continued
Introduction 7
literature supporting an association between meth and particular outcomes captured within
that chapter. In many places, although the science suggests a relationship between meth and
particular outcomes, no cost estimates can be constructed, because the data are insufficient
for doing so. It is important that the omission of these costs is not misinterpreted as an infer-ence that the costs are small or zero. Indeed, we dedicate an entire chapter (Chapter Nine) to
a discussion of some potentially plausible magnitudes of these costs under differing assump-tions and whether their omission is likely to be a large or small factor in terms of the overall
estimate. Within the specific chapters, however, we simply note their omission so that the esti-mates we produce are based on the strongest science available. The specific methods, data sets,
and assumptions used to generate these estimates are all contained in the individual chapters,
although technical regression results supporting parameters used in estimate construction
within the chapter are placed in the appendixes. We conclude the monograph with a discus-sion of the magnitude of these costs vis-à-vis other substances of abuse and what we can infer
regarding who bears the burden of these costs.
We add just a final word of caution regarding the information provided in this mono-graph. The information is based on the best available national data and the state of the science
at the time we wrote this. Given the relatively early stages of the meth epidemic in some states,
we anticipate that the knowledge base will change substantially in the next few years, warrant-ing a reconsideration of these costs and updating of these estimates. These are costs based on
incidences of meth-related problems in 2005, which will not be reflective of the problem in
2008 or 2010, given that this is an evolving social issue
CHAPTER TWO
The Cost of Methamphetamine Treatment
In Table 2.1, we provide a summary of our findings with respect to the cost of meth treatment.
Numerous calculations, assumptions, and data sources were involved in the construction of
these estimates, which are described in greater detail in this chapter.
Care Received in the Specialty Sector
The Substance Abuse and Mental Health Services Administration (SAMHSA) maintains an
administrative data system designed to provide annual data on the number and characteris-tics of persons admitted to public and private nonprofit substance abuse treatment programs.
The Treatment Episode Data Set (TEDS), part of the Drug and Alcohol Services Information
System (DASIS), collects information from any treatment provider receiving public funding
Table 2.1
Estimated Cost of Methamphetamine Treatment in the United States in 2005 ($ millions)
Cost Lower Bound Best Estimate Upper Bound
Community-based specialty treatment
Hospital-based treatment 5.6 14.9 14.9
Specialty treatment sector 274.1 491.2 977.2
Care received in other settings NI NI NI
Total for community-based treatment 279.7 506.1 992.1
Federally provided specialty treatment
DoD 0.2 0.2 0.9
IHS 4.7 24.4 24.4
Federal Bureau of Prisons 5.7 5.7 10.0
VA 9.1 9.1 43.5
Total for federally provided specialty treatment 19.7 39.4 78.8
Total cost of drug treatment 299.4 545.5 1,070.9
NOTE: Data are for treatment in which meth is the primary drug of abuse.
10 The Economic Cost of Methamphetamine Use in the United States, 2005
(through block grants, state agencies, or government insurers). The reporting is on all patients,
regardless of the funding for any particular patient.
1
TEDS identifies meth cases using primary, secondary, and tertiary diagnosis. Cases in
which meth is the primary drug of abuse are identified as those cases in which the primary
diagnosis is meth.2
As shown in Table 2.2, there were 160,101 cases in which meth was the pri-mary drug of abuse and treatment did not take place in a hospital setting in 2005. Only 40,831
of these were cases involving meth alone.
3
All other cases included alcohol or some other sub-stance of abuse.
4
The number of cases exclusive to meth omits the additional 67,585 treatment
episodes that list meth as a secondary or tertiary drug of abuse. As is standard in previous COI
studies (ONDCP, 2004b; Harwood, Fountain, and Livermore, 1998; Mark, Woody, et al.,
2001), we do not include in our cost calculation those cases in which meth is a nonprimary
drug of abuse. Likewise, when meth is the primary drug of abuse but there are other drugs of
abuse as well, all costs are ascribed to meth.
Table 2.2
Methamphetamine Treatment Admissions
Admission Type
Number of Primary Meth Admissions
Meth as Primary Drug,
Other Substances Included
in Subsequent Drug Codes
Meth as Primary Drug, No
Other Substances Included
in Subsequent Drug Codes
Meth as Primary Drug,
Combined
Outpatient, methadone
maintenance
218 144 362
Standard outpatient 62,425 22,061 84,486
Intensive outpatient 16,413 4,444 20,857
Short-term residential 11,884 3,406 15,290
Long-term residential 18,551 5,688 24,239
Methadone detox 317 163 480
Detox, free standing
and ambulatory,
nonmethadone
a
9,462 4,925 14,387
Total 119,270 40,831 160,101
SOURCE: SAMHSA (2007c).
a
Estimate includes free-standing residential detox and ambulatory, nonmethadone detox. It does not include
hospital detox.
1
SAMHSA estimates that more than 95 percent of substance abuse patients receive treatment from a facility receiving
public funds for at least some patients.
2
In TEDS, the primary, secondary, and tertiary diagnosis variables are labeled SUB1, SUB2, and SUB3, respectively.
Meth is categorized by any of these having a value of 10. In Oregon, cases involving meth in 2005 were categorized as
“amphetamine.” Personal correspondence with staff responsible for the Oregon database confirmed that these amphetamine
cases could be considered meth cases.
3
Based on this Stata 8.1 programming code: SUB1==10 & SUB2==1 & SUB3==1 & NUMSUBS==1.
4
The category also includes 37 episodes in which meth is likely the only drug but could not be verified as such (i.e., when
SUB1<2 & (SUB2==10|SUB3==10) & NUMSUBS==1).
The Cost of Methamphetamine Treatment 11
We also report in Table 2.2 the number of cases in which meth is the only reported sub-stance of abuse, so the interested reader could calculate the total cost of meth-only episodes if
desired. Similarly, if one is comfortable making an assumption regarding the fraction of cases
in which meth were a secondary or tertiary code that had additional costs due to meth, one
could use the weighted average cost of a treatment episode across modalities ($3,487, accord-ing to the Substance Abuse Treatment Cost Analysis Allocation Template, or SATCAAT) to
estimate the additional cost of these episodes. For example, if one assumes that the presence
of meth as a secondary or tertiary drug either caused the dependence on the primary drug or
raises the cost of treating the cases in which it is secondary or tertiary by, say, 20 percent, one
would raise our estimate of the cost of drug abuse by 0.20 × 67,585 × $3,487 = $47.13 mil-lion. For consistency with previous estimates of the cost of drug abuse in the United States,
however, we include in our costs all cases for which meth is reported as the primary substance
of abuse.
Although TEDS provides a census of people receiving treatment from all publicly funded
treatment facilities and programs, it does not contain any information on the cost of care
received. Thus, information on the cost of each treatment episode must be inferred based on
information on the number of treatment episodes (as indicated by admissions into each service
setting) and typical per episode costs from other sources. Unfortunately, no current national
data system provides information on the cost of treatment by drug and service setting. How-ever, several recent studies provide estimates of the cost of drug treatment by setting for large
geographically dispersed areas, including the Alcohol and Drug Services Study (ADSS), the
Drug Abuse Treatment Cost Analysis Program (DATCAP), and SATCAAT. These sources
do not differentiate treatment costs based on the drug of abuse, but rather examine costs by
treatment modality (e.g., inpatient, intensive outpatient, outpatient, methadone maintenance).
They differ in the types of costs considered or included and tend to focus on the accounting or
economic costs of delivering the services.
DATCAP is a data-collection instrument and interview guide developed in the early
1990s to estimate the economic cost of substance abuse programs (Roebuck, French, and
McLellan, 2003; French et al., 1997). It is based on standard accounting and economic prin-ciples and, hence, measures both economic and accounting costs. The instrument is designed
to capture and organize detailed information on resources used in the delivery of treatment.
The DATCAP methodology has been applied in more than 100 different treatment programs
since its development. However, most of the programs that have used the instrument have
been involved in clinical studies of new treatments or efficacy studies. Thus, the programs that
have used the instrument do not represent a random sample of treatment programs or facili-ties. Nonetheless, the programs that have adopted the costing instrument are geographically
dispersed across the United States and therefore do capture some of the geographic variability
in cost. Roebuck, French, and McLellan (2003) provide a summary of findings from 85 stud-ies that employed DATCAP over a 10-year period and present average episode cost estimates
in 2001 dollars for nine treatment modalities from these studies.
SATCAAT is a unit cost instrument that was piloted in a study of 43 providers represent-ing 406 treatment programs sponsored by Center for Substance Abuse Treatment (CSAT).
The study collected cost information from 43 providers in six states and Puerto Rico.5
The
5
The six states were Michigan, Massachusetts, Pennsylvania, New York, Florida, and Texas.
12 The Economic Cost of Methamphetamine Use in the United States, 2005
SATCAAT instrument considers the accounting cost of delivering treatment services plus the
imputed value of donated items; thus, the instrument does capture some of the economic costs
associated with the delivery of substance abuse treatment. Average treatment costs are reported
by population (women, men, children, and combined for all adults) and service-delivery unit
(e.g., residential, detox, intensive outpatient, standard outpatient). For the purposes of this
work, we use estimates reported by service-delivery unit for adults in general.
SAMHSA conducted the ADSS study between 1996 and 1999 using information from
280 facilities, consisting of 44 nonhospital residential treatment facilities, 222 outpatient treat-ment facilities, and 44 methadone maintenance treatment facilities.6
The cost data were col-lected through in-person interviews with administrators that took place during site visits and
validated using an automated program that tested the relationship of costs to client counts,
staffing, and resource and utilization measures. Treatment costs were estimated by each type
of treatment facility and modality and reported in constant 1997 dollars.
A critical strength of the DATCAP and SATCAAT instruments is that they attempt
to measure more than accounting costs by including in their estimate of unit costs the value
of all resources that are donated or shared across programs or facilities. However, estimated
service-unit costs using these instruments are publicly available only from a summary of
clinical-study results that involved selective populations and facilities that meet specific crite-ria for the purposes of the clinical study. Hence, because neither the sample of the clinics nor
that of the patients is representative, the costs obtained from these studies may not be reflec-tive of the typical costs.
The ADSS is the only study that provides information on the average (and median) cost
per episode of drug treatment based on a nationally representative sample of treatment facili-ties serving regular patients across the country. Thus, the estimates of unit cost from this study
are likely to provide a more realistic reflection of the typical accounting costs of serving regular
patients. But, to the extent that the treatment facilities and programs rely on donated or sub-sidized resources, accounting costs (as captured in the ADSS) will clearly underestimate the
actual value of resources used in the delivery of treatment services.
In the first column of Table 2.3, we report the inflation-adjusted average unit cost by
modality generated from each of these three studies. It is important to note that we have infor-mation on the average cost of detox in the specialty sector only from the SATCAAT instru-ment. For all other treatment modalities, we have information on the average cost per episode
from each of these unit-costing methods. In all cases, we inflate the average cost from pub-lished studies using 2005 dollars and the medical-care consumer price index (CPI). As can be
seen in this table, there are indeed large differences in the average cost per modality when we
use different instruments to estimate costs. The ADSS estimate is consistently lower than the
other two regardless of the modality, which reflects the fact that the ADSS omits the economic
value of donated and volunteer resources used in the delivery of these treatments. What is per-haps even more surprising is how much larger the DATCAP estimates are than the SATCAAT
estimates, which reflects some important differences in the method of calculating costs, partic-ularly in short-term residential facilities, as well as possible differences in the types of facilities,
programs, and patients observed in the clinical studies adopting these cost instruments.
6
Some facilities provided treatment through more than one modality.
Table 2.3
Estimated Costs of Outpatient Treatment Episodes, by Service Unit
Treatment
Category Data Source Service unit
Per Episode
Unit Cost
Estimate ($)
Episodes
with Meth
as Primary
Diagnosis
Total Service-Unit Cost ($)
Average Cost
Per Meth
Treatment
Episode by
Instrument ($)
Detox
a
SATCAAT All 1,215 14,867 18,064,892 1,215
Inpatient and
outpatient
services
ADSS Nonhospital
residential
b
3,398 38,594 131,128,132
Nonhospital
outpatient
c
1,268 96,744 122,684,936
Methadone
maintenance
6,561 341 2,237,281
Total 256,050,349 1,887
DATCAP
d
Intensive
outpatient
e
5,178 20,656 106,962,345
Standard
outpatient
f
2,339 76,088 177,979,723
Long-term
residential
g
21,904 22,655 496,227,870
Short-term
residential
h
10,981 15,939 175,025,681
Methadone
maintenance
8,572 341 2,922,991
Total 959,118,610 7,069
SATCAAT
i
Intensive
outpatient
3,981 20,656 82,241,038
Standard
outpatient
2,331 76,088 177,361,889
Long-term
residential
7,885 22,655 178,645,776
Short-term
residential
2,187 15,939 34,858,912
Methadone
maintenance
n.a. 341
Total 473,107,614 3,487
NOTE: Estimates for meth as primary drug; includes use of other substances. n.a. = not available. Due to
rounding, totals may not sum precisely.
a
Detox includes free-standing residential, ambulatory detox services, and methadone detox in TEDS.
b
Cost is applied to short- and long-term residential visits in TEDS.
c
Cost is applied to outpatient and intensive outpatient visits in TEDS.
d
Rates are weekly, not daily. Thus, unit costs are not directly comparable, but totals are.
e
Costs are based on estimates for the adult intensive-care unit (ICU).
f
Estimates are a weighted average of the reported standard and adolescent outpatient services, based on the
ratio of adults to children in these services in TEDS. LOS, in days from TEDS, is converted to weeks.
g
Estimates are those specified as therapeutic community (TC) programs.
h
Estimates are those specified as adult residential programs.
i
Estimates are based on programs servicing adults.
14 The Economic Cost of Methamphetamine Use in the United States, 2005
In the “Episodes with Meth as Primary Diagnosis” column of Table 2.3, we present the
total number of primary meth treatment episodes obtained when the TEDS data are aggre-gated to reflect the same type of treatment as the average cost estimates.
7
Then, in the “Total
Service-Unit Cost” column, we multiply our average cost per modality from each instrument
by the number of meth episodes falling into that category and get a total cost of treatment for
each methodology of unit costing.
The information on the number of treatment episodes in 2005 from TEDS is reliable,
whereas the cost per average treatment episode in the specialty sector is uncertain. Therefore,
we construct our lower-, middle-, and upper-bound estimates of the cost of treatment solely
from the variation among the three alternative costing methods. The lower bound is given
using the estimated average cost per episode from the ADSS, the upper bound is given using
estimates from DATCAP, and the best estimate is given using estimates from SATCAAT.
Added to each of these average treatment cost values is the cost of detox, which, in all cases, is
estimated using the SATCAAT instrument. The choice of SATCAAT as our best estimate of
the unit cost of treatment is based primarily on the fact that the estimates from this instrument
fall within the range provided by the ADSS and DATCAP. We also have billing information
on all patients receiving treatment for meth abuse in the state of Texas, and the mean cost for
short-term inpatient care in the state is $2,292.73 for 2005, which is very close to the average
episode cost reported by the SATCAAT instrument (for short-term residential care).
8
Thus, we
have data from at least one state that suggest that the average cost estimates reported by the
SATCAAT instrument are in line with billing charges for meth treatment in that state. That
is not to say that the costs calculated from the other instruments are wrong or useless. Further
research is needed to help us understand what the appropriate unit-cost estimates are for dif-ferent types of drug treatment.
Costs of Hospital-Based Drug Treatment
Information on the delivery of hospital-based treatment for meth abuse comes from the 2005
Nationwide Inpatient Sample (NIS) data provided by the Healthcare Cost and Utilization
Project (HCUP) for 2005. Diagnosis codes (codes from the International Classification of
Diseases, ninth revision, or ICD-9s) were used to identify amphetamine-related admissions
(see Table 2.4). The two codes used to identify amphetamine abuse and dependence among
patients in the NIS are 304.4 and 305.7. The current analysis focuses on the primary diagno-sis that defines the stay and the presence of amphetamine on the patient record. For hospital-based treatment costs, the primary diagnosis must be amphetamine dependence or abuse.
We use the actual costs for each stay in constructing cost estimates for this segment. We cau-tion that, using ICD-9 codes, it is not possible to distinguish meth specifically from general
amphetamines. This is a shortcoming common to hospital and ICD-9–based data sets. In
some regions, such as California, the distinction and any potential overestimation is minor,
because meth comprises the vast majority of amphetamine abuse. However, this share is likely
to be lower in other states where meth is less common and hence result in greater potential for
7
The total number of treatment episodes in Table 2.3 is lower than that in Table 2.2 because hospital-based treatment is
excluded. We cost out hospital episodes later in this chapter.
8
Texas treatment data include information on billing charges and were made available to us for research purposes.
The Cost of Methamphetamine Treatment 15
Table 2.4
Hospital-Based Drug Treatment
Treatment
Low Estimate, Primary
Amphetamine Only
Best Estimate, Primary
Amphetamine
Upper Estimate, Primary
Amphetamine
Number Cost ($) Number Cost ($) Number Cost ($)
Detox 681 1,615,107 1,907 7,083,567 1,907 7,083,567
Treatment 1,440 4,360,334 2,790 8,893,426 2,790 8,893,426
Total 2,123 5,971,301 4,697 16,008,996 4,697 16,008,996
Adjusted for
meth only
5,559,281 14,904,376 14,904,376
SOURCE: NIS (2005).
NOTE: Due to rounding at various stages and misclassification of a small number of cases, numbers may not add
precisely.
overestimation. To address this issue in our estimates, we multiply the national estimates of
treatment costs in the hospital sector by a meth–to–total amphetamine adjustment factor that
is constructed from TEDS. In the TEDS data, we can identify treatment for meth separately
from other amphetamines, so we can use this ratio as our adjustment factor. This ratio, which
is 0.931 for 2005, may be biased if the population served in general, short-stay hospitals dif-fers in terms of its likelihood of seeking treatment for meth versus for other forms of amphet-amines. However, no other data are available on which to base such an adjustment for 2005
(and the likely variation from a total adjustment factor of 7 percent is probably within this
study’s margin of error, given other necessary approximations and omissions).
9
There were 4,697 admissions in which amphetamine abuse and dependence were identi-fied as the primary diagnosis. Of these, approximately 40 percent (1,907 admissions) were for
individuals who received detox during their stay. On average, these stays were longer (10 days)
than those that did not include detox (four days). The total costs associated with these admis-sions were $16.0 million for all amphetamines, but our best estimate is $14.9 million after dis-counting costs to isolate meth from other amphetamines. For cases in which amphetamine was
the primary and onlydrug, we obtain significantly smaller estimates of just under $6.0 million
for 2,123 cases. After our adjustment for the share due to meth (versus amphetamines), the
lower-bound estimate is $5.6 million. Stays for which amphetamine is a secondary drug are
not factored into our estimate of amphetamine-related drug-treatment costs, in order to treat
them consistently with estimates from the specialty sector.
9
When the second wave of the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC) data
becomes available, this might serve as an alternative source of information for adjusting meth to total amphetamine con-sumption in the general population.
For 2005, TEDS reports 253,836 admissions involving meth, other amphetamines, or other stimulants. There were 248,245
admissions that involved either meth or another amphetamine. With 232,205 meth admissions, this implies that 93.1 per-cent of all meth- and amphetamine-related treatment episodes are due to meth abuse. Similarly, we can use these numbers to
show that 91.1 percent of all stimulant admissions are due to meth (a figure that we will use elsewhere in the monograph).
16 The Economic Cost of Methamphetamine Use in the United States, 2005
Other Federal Treatment
Several agencies offer treatment services that are not recorded in the TEDS and NIS data sets.
These are primarily federal specialty treatment services provided by such agencies as DoD, the
IHS, Federal Bureau of Prisons, and the VA. The total treatment budgets for these four agen-cies exceeded $550 million in 2005 (ONDCP, 2006a). Unlike the data we have for TEDS and
the NIS, we do not have individual patient records for treatment services provided by these
agencies to use in constructing estimates of the share attributable to meth. Consequently, our
approach is to estimate the share of those agencies’ treatment budgets that can be reasonably
attributed to meth using drug-use and drug-treatment data from TEDS, National Survey
on Drug Use and Health (NSDUH), and a nationally representative survey of federal prison
inmates.
The share of the budgets attributable to meth is constructed by examining substance use
and treatment patterns in the populations that each agency serves. The treatment budgets for
these agencies are used to treat substance abuse other than alcohol except for the IHS budget,
which does include alcohol treatment.
For DoD, our lower-end estimate is based on the share of (nonalcohol) treatment admis-sions in TEDS that are attributable only to meth (3.8 percent). We use the rate for the full
population, since DoD covers treatment costs for service members and their dependents. The
upper bound is calculated using the share of all TEDS treatment admissions in which meth is
mentioned as one of the drugs of abuse (15.9 percent). To determine the appropriate fractions
for the VA, we again rely on TEDS, which includes a variable indicating whether the patient
is a veteran (because not all veterans are treated in the VA system). Using information on just
the veterans in TEDS, we assess what percentage of veterans report meth as their only drug of
abuse (2.4 percent) and what percentage report meth as one of their drugs of abuse (11.3 per-cent). We then apply these fractions to the VA treatment budget to allocate spending to meth.
For the IHS, we again use descriptive information in TEDS to help us identify the attribution
factors.
10
In TEDS, information is available on ethnicity, including American Indians and
Alaska Natives. To determine what fraction of American Indians identify meth as the drug of
abuse, we subset the TEDS data to this population and determine the percentage of all Ameri-can Indian admissions who have meth as the only drug of abuse (3.8 percent) and those who
have meth as one of the drugs of abuse (20 percent).
Finally, for the Bureau of Prisons budget, we estimate meth use among federal prison
inmates by using data from a nationally representative survey of that population. The lower
bound is based on the number of individuals who entered drug treatment since admission to
prison and used meth dailyin the month before arrest (11.7 percent). The upper bound is based
on the number of individuals who entered drug treatment since admission to prison and used
meth at least oncein the month before arrest (20.5 percent). The resulting dollar amounts for
each agency are shown in Table 2.5.
In almost every case, we adopt the lower bound as our best estimate of these costs because
it conservatively accounts for meth users and we really have no way of knowing the true
10
The 2005 IHS treatment budget was calculated by multiplying all IHS funds spent on alcohol and substance abuse ser-vices (prevention and treatment) in 2005 by 87.9 percent. This 87.9 percent was calculated from ONDCP’s IHS summary
for 2003 (ONDCP, 2002).
Table 2.5
Federally Provided Specialty Treatment
Agency Lower Bound ($) Best Estimate ($) Upper Bound ($)
DoD 205,824 205,824 852,316
IHS 4,681,072 24,432,017 24,432,017
Federal Bureau of Prisons 5,676,521 5,676,521 9,966,746
VA 9,101,364 9,101,364 43,539,998
Total 19,664,781 39,415,726 78,791,077
SOURCES: ICPSR (2007); IHS (2005); ONDCP (2002, 2006a); SAMHSA (2007c).
allocation. However, in the case of the IHS, there is strong evidence that meth is having a
dramatic effect on several Indian reservations, especially those in the South and West (Kronk,
2006). Thus, in this instance, we expect the true meth attribution factor to be closer to 20 per-cent (our upper-bound estimate) than to 4 percent, and we use this as our best estimate.
As noted throughout this section, there are some limitations to our estimates of treatment
costs. We capture costs for treatment in the specialty, hospital, and federal sectors. However,
we are missing costs from a couple of other relevant sectors, including those receiving care in
the general medical sector and those receiving care at a private specialty institution that does
not receive any public support. Moreover, there are no currently maintained nationally repre-sentative data sets that allow us to cost treatment in the specialty sector. Instead, we must rely
on previous studies to provide the unit-cost estimates used in our calculations. Finally, we attri-bute the costs only for individuals treated for a primary diagnosis of meth. Clearly, there will
be individuals with primary diagnoses of other drug use who suffer from meth comorbidities,
but it is difficult to determine what share of those costs should be allocated to meth without
risk of overestimation. Conversely, some individuals primarily diagnosed with meth use have
secondary and tertiary diagnoses of other drug use, and we do not, in those cases, try to sub-tract the portion of treatment costs owing to the subsidiary diagnoses.
CHAPTER THREE
The Cost of Methamphetamine-Related Health Care Among
Methamphetamine Users
A number of significant health concerns and problems beyond dependence and abuse may arise
directly from meth use. Drug-induced health issues are yet another direct cost imposed by use
of this substance. In this chapter, we provide a summary of our estimates for specific health
areas we considered as part of our health cost estimate. Information on the cost of health care
services related to meth use comes from several primary sources. Information on hospital-based
care received for amphetamine-induced and amphetamine-involved conditions comes from the
NIS. Emergency-department (ED) visits that do not result in an inpatient stay and visits to
specialty care, general, or other practitioners (e.g., dental) are not captured in the NIS. Infor-mation on suicide attempts and ED utilization comes from the Drug Abuse Warning Network
(DAWN). Information on health administration costs is drawn from the Office of National
Drug Control Policy and SAMHSA.
The scientific literature, summarized in this chapter, provides evidence of meth’s associa-tion with a wide variety of conditions but cannot always distinguish conditions or particular
cases within each condition that are causedby meth versus those that are exacerbated, acceler-ated, or otherwise influenced by meth use. After completing our literature review, we examined
information available in the NIS and separated conditions into two groups: First are those for
which we are confident the literature supports a causal relationship and in which the full cost
of treating that condition is fully attributable to meth (see Table 3.1). We refer to the hospital
stays for these conditions as meth-induced. Second are those for which the literature supports
an association but cannot definitively separate causality from selection effects or other factors.
We include only the incremental cost associated with treating these health problems. Specifi-cally, we estimate and incorporate only the additional costs that are caused by having meth
use or abuse as a comorbidity rather than the full health care cost associated with treating the
primary health problem.
In many cases, patients abuse multiple substances that may also contribute to specific
health problems. Because it can be difficult determining what fraction of these health problems
are truly due to the use of meth versus use of other substances, we provide alternative measures
when the scientific literature is not clear. For example, in our lower-bound estimate of the costs
of meth-induced cases, we include only those cases in which meth is the only substance of
abuse. These cases have no indication of other illicit drugs or alcohol in the medical record. In
our upper-bound estimate, all the cases that include amphetamine, regardless of whether alco-hol is also noted on the record, are included in our estimate of costs. Details provided in later
tables in this chapter show how these alternative assumptions significantly change the number
of cases considered for our lower- and upper-bound estimates.
20 The Economic Cost of Methamphetamine Use in the United States, 2005
Table 3.1
Health Care Costs Experienced by Methamphetamine Users (2005 $ millions)
Care Received Lower Bound Best Estimate Upper Bound
Meth-induced hospital stays 27.06 27.06 35.55
Meth-involved hospital stays, incremental only 8.20 14.29 36.70
Suicide attempts 5.51 14.24 22.96
ED visits 23.79 45.91 68.03
Outpatient care NI NI NI
Dental NI NI NI
Health administration 51.71 249.83 448.0
Total, excluding intangible cost of addiction 116.27 351.33 611.19
NOTE: NI = not assessed as part of this project. Due to rounding, totals may not sum precisely.
The cost of health care, excluding the intangible costs of addiction, is estimated at
$165.5 million, with a range from $89.4 million to $266.3 million. The combined costs of
meth-induced and meth-involved inpatient stays are estimated at $41.3 million, with a range
from $35.3 million to $72.3 million. Our best estimate of the costs of suicide attempts from the
DAWN data is $14.2 million, with a range of $5.51 million to $23.0 million. Finally, the costs
of other ED visits are $45.9 million. The lower and upper bounds for ED visits are $23.8 mil-lion and $68.0 million, respectively. Health administration costs range from $51.7 million to
$448.0 million, with a best estimate of $249.8 million.
With our data, we are unable to capture the cost of care received from general practition-ers outside the hospital and ED setting. We are also unable to include the costs associated with
dental care for meth mouth. We do, however, provide some evidence on these cost components
in Chapter Nine. These omissions imply that the totals in Table 3.1 underestimate the health
costs associated with meth use. Other limitations to these estimates, such as our inability to
identify causal versus incremental costs for meth-involved stays, are discussed in the relevant
sections in this chapter and in Chapter Nine.
This chapter focuses on the health events and costs associated with meth use. The health
care costs related to treatment of individuals injured in theproduction of meth are not cap-tured in this chapter because they cannot be readily identified in the hospital and ED data sets.
Chapter Seven uses information produced by the Centers for Disease Control and Prevention’s
(CDC’s) Hazardous Substances Emergency Events Surveillance data set to approximate the
cost of treating health problems, injuries, and trauma caused by hazardous-substance events
associated with meth production. These estimates also do not include other costs (beyond
direct health care) associated with deaths caused by meth use. Data on meth-related deaths are
available from the CDC National Center for Health Statistics multiple-causes-of-death data
file. The loss of life due to premature mortality and the associated economic costs are discussed
in Chapter Four, as are the intangible costs associated with meth abuse and dependence.
The Cost of Methamphetamine-Related Health Care Among Methamphetamine Users 21
A Literature Review
Meth use is accompanied by a rush of energy as dopamine and serotonin uptake is inhibited,
resulting in euphoria, increased wakefulness, respiration, hyperthermia, physical activity, and
decreased appetite (Anglin et al., 2000). With extended use, however, the body adjusts, so
that, when the user is not under the influence of meth, he or she experiences the mirror of
these usage effects—increased restless anxiety, irritability, fatigue, and dysphoria (Lineberry
and Bostwick, 2006). These feed the desire for more of the drug. In addition to this pernicious
dependence, there are detrimental physical effects from meth—most prominently, cardiovas-cular, pulmonary, and dental. Meth also encourages high-risk behaviors that contribute to
additional negative physical outcomes. Adverse health effects associated with meth use are not
limited to the user, in that they can affect fetal development as well. Finally, there are other
effects among nonusers, including risks to health associated with the production of meth,
which may affect producers, first-responder personnel, and bystanders. These nonuser costs are
considered in Chapter Six.
The cardiovascular effects of meth include acute and chronic damage. Meth use increases
the heart rate and blood pressure and has been associated with chest pain and palpitations
(Bashour, 1994; Furst et al., 1990; Lam and Goldschlager, 1988), hypertension (Albertson,
Derlet, and Van Hoozen, 1999), stroke (McEvoy, Kitchen, and Thomas, 1998, 2000; Perez,
Arsura, and Strategos, 1999), acute myocardial infarction (Chen, 2007), cardiovascular col-lapse, and arrhythmic, sudden death (Wermuth, 2000). There is also some evidence that the
risks of acute myocardial infarction are worse when combined with ethanol (Mendelson et al.,
1995). Some of the cardiovascular effects may be enduring.
1
For example, it is believed that
there is long-term damage from increases in blood pressure as well as cardiac toxicity (Yu,
Larson, and Watson, 2003). These chronic effects include myocardial infarction, cardiomyo-pathy (Winslow, Voorhees, and Pehl, 2007), stroke (Winslow, Voorhees, and Pehl, 2007; Ohta
et al., 2005), and cardiac lesions (Yu, Larson, and Watson, 2003; Matoba, 2001). These lesions
are associated with increased catecholamine levels (Yu, Larson, and Watson, 2003), and animal
experiments have corroborated observed human effects (Matoba, 2001). These effects on the
heart may be amplified by comorbidity with acquired immune deficiency syndrome (AIDS)
and AIDS medicines (Yu, Larson, and Watson, 2003).
Respiratory effects of meth include dyspnea (Lam and Goldschlager, 1988; Bashour,
1994), pulmonary edema (Lukas and Adler, 1996), and pulmonary hypertension (Albertson,
Derlet, and Van Hoozen, 1999). Exact incidence and prevalence of meth-induced respiratory
symptoms have not been reported (Albertson, Derlet, and Van Hoozen, 1999). Contaminants
have been suggested as the cause for the respiratory issues, although the direct effects of inhal-ing meth itself have not been eliminated (Albertson, Derlet, and Van Hoozen, 1999; Schai-berger et al., 1993). These respiratory effects may be chronic (Yu, Larson, and Watson, 2003)
or fatal, as occurs in complete respiratory failure (Winslow, Voorhees, and Pehl, 2007).
1
There is evidence that someone can have a heart attack due to meth years after quitting meth use. According to Kaye
et al. (2007, p. 1209), “risk is not likely to be limited to the duration of [patients’] methamphetamine use, because of the
chronic pathology associated with methamphetamine use.” Increases in blood pressure are one factor, featuring predomi-nantly in strokes, but there are cardiac toxicity factors that extend beyond just the increased blood pressure. The cardiac
toxicity includes increased catecholine levels and causes myocardial infarction, cardiomyopathy, and cardiac lesions, as
mentioned.
22 The Economic Cost of Methamphetamine Use in the United States, 2005
There is a range of other physical effects of meth use. Seizures and convulsions are uncom-mon but potentially severe (Sommers, Baskin, and Baskin-Sommers, 2006). Meth toxicity
contributes to intracerebral hemorrhage (McGee, McGee, and McGee, 2004), hyperthermia,
renal failure, and necrotizing angiitis in the kidney, liver, pancreas, and small bowel (Callaway
and Clark, 1994; Chan et al., 1994; Sperling and Horowitz, 1994; Screaton et al., 1992; Citron
et al., 1970). Hyperpyrexia, an extreme increase in body temperature, has been observed, with
readings up to 109 degrees Fahrenheit (G. King and Ellinwood, 1997).
Meth abuse is associated with severe oral damage wherein the teeth are “blackened,
stained, rotting, crumbling, or falling apart” (ADA, 2006, cited in Klasser and Epstein, 2005,
p. 760). This is called meth-induced caries (MIC), also known as meth mouth, and is charac-terized by extremely bad caries on the outside smooth surfaces of the teeth and the adjoining
surfaces of the anterior teeth (Klasser and Epstein, 2005; Shaner, 2002; Duxbury, 1993; Don-aldson and Goodchild, 2006). There are several possible mechanisms involved in meth mouth.
It is believed that meth-induced hyposalivation, or dry mouth, minimizes the normal protec-tive capacities of saliva and that deteriorates the teeth (Klasser and Epstein, 2005; Saini et al.,
2005; Donaldson and Goodchild, 2006). Other research has considered the possibility that
the acidity of meth contributes to meth mouth, although this is far from certain (McGrath
and Chan, 2005; Klasser and Epstein, 2005; Donaldson and Goodchild, 2006). Additionally,
meth users tend to engage in behaviors that may contribute to the dental damage associated
with meth mouth, including an increased propensity for excessive chewing, tooth-grinding and
-clenching (Klasser and Epstein, 2005; McGrath and Chan, 2005), and indulging in cravings
for carbonated, sugared beverages (Brunswick, 2005; Jones, 2005; Klasser and Epstein, 2005;
Shaner, 2002). Some of these dental effects may be due to selection rather than causality. Meth
users are generally less concerned than other people are with personal or oral hygiene (Klasser
and Epstein, 2005; Shaner, 2002; McGrath and Chan, 2005; Wynn, 1997). Still, there is addi-tional evidence supporting a causal relationship between meth use and dental problems from
the meth’s prior use as a prescription medicine, which may partly address selection concerns
(Klasser and Epstein, 2005; Howe, 1995).
In addition to the direct physical effects, meth influences user behaviors, which can
lead to additional health conditions. For example, meth use is associated with skin-picking
behaviors that are common when the user is in a state known as tweaking,
2
occasionally lead-ing to abscesses (see N. Lee et al., 2005). Weight loss is prevalent (Sommers, Baskin, and
Baskin-Sommers, 2006; Winslow, Voorhees, and Pehl, 2007), accompanied by malnutrition
(Richards et al., 1999). Meth use is also associated with poor hygiene (Winslow, Voorhees, and
Pehl, 2007), increased risk-taking behavior leading to motor-vehicle accidents (Schepens et
al., 1998), and increased aggression leading to violence and even death (Baskin-Sommers and
Sommers, 2006; Kalant and Kalant, 1979; Ellinwood, 1971). To some extent, this is due to
selection effects associated with the criminal nature of meth, but animal studies suggest that
at least some of the aggression associated with meth is due to use (Melega et al., 2008; Sokolov
and Cadet, 2006; Sokolov, Schindler, and Cadet, 2004).
Meth may also influence health by contributing to a lowered immune response (Yu,
Larson, and Watson, 2003), and there is evidence of additional comorbidity with human
immunodeficiency virus (HIV) infections (Yu, Larson, and Watson, 2003). This amplifies the
2
Tweaking is the “most dangerous stage of methamphetamine abuse” and “occurs when an abuser has not slept in 3–15
days and is irritable and paranoid” (CESAR, 2005).
The Cost of Methamphetamine-Related Health Care Among Methamphetamine Users
“occurs when an abuser has not slept in 3–15
days and is irritable and paranoid” (CESAR, 2005).
additional risk of sexually transmitted diseases (STDs) due to meth-induced behaviors. Meth
use is associated with increased libido and enhanced sexual pleasure (Winslow, Voorhees, and
Pehl, 2007; Gibson, Leamon, and Flynn, 2002; Kipke et al., 1995; Meston and Gorzalka,
1992; Gorman, Morgan, and Lambert, 1995), which is believed to lead to high-risk sexual
behavior (Winslow, Voorhees, and Pehl, 2007; Shoptaw et al., 2003; Lyons, Chandra, and
Goldstein, 2006) and multiple partners (Zule and Desmond, 1999). This greater sexual risk,
as well as needle-sharing behaviors (Gibson, Leamon, and Flynn, 2002; Zule and Desmond,
1999; Schoenbaum et al., 1989) and the reduced resistance resulting from meth use, increases
the likelihood of infections (Ye et al., 2008; Yu, Larson, and Watson, 2003). Studies link meth
use with endocarditis, hepatitis, HIV, tuberculosis (Winslow, Voorhees, and Pehl, 2007; Gon-zales et al., 2006; Kipke et al., 1995; Molitor et al., 1998; Anderson and Flynn, 1997; Gorman,
Morgan, and Lambert, 1995; Des Jarlais and Friedman, 1987; Schoenbaum et al., 1989), and
methicillin-resistant Staphylococcus aureus(MRSA) (A. Cohen et al., 2007, cited in Rosenthal,
2006).
The increase in risk-taking behavior is likely to extend beyond sexual activity. There is
some evidence of an association with injuries due to thrill-seeking or aggressive risk-taking
induced by meth use as well as hallucinations, delusions, and suicide ideation among chronic
users. Sheridan et al.’s (2006) review of the literature asserts associations between meth use and
injury, including blunt trauma (usually vehicular trauma) and interpersonal trauma resulting
from gunshot wounds or stabbings and other assaults, as well as self-inflicted trauma. Schepens
et al. (1998) likewise finds an association with motor-vehicle accidents. Interestingly, a review
of Hawaiian trauma patients by Tominaga et al. (2004) found not only a link between cur-rent meth usage and incidence of injury but also higher incremental costs for the same level of
injury with co-occurring meth use. Individuals at the trauma center who were meth-positive
were older than meth-negative persons with such injuries, were more likely to have experienced
self-inflicted trauma, stayed in the hospital longer, and used trauma-center resources out of
proportion to injury severity. While previous studies found no additional costs for injury with
co-occurring meth use for life-threatening injuries, these results suggest a potential incremen-tal cost to treating injuries with co-occurring meth use at least at moderate levels of injury.
The physical effects of meth use are a particular concern with regard to expectant moth-ers because women constitute approximately 50 percent of meth users admitted for treatment.
3
Studies suggest that meth may contribute to intracerebral hemorrhage, cardiovascular collapse,
seizures, and amniotic-fluid embolism in pregnant women (Stewart and Meeker, 1997; Catan-zarite and Stein, 1995). The effects of meth extend to their children as well; there is evidence
that meth can pass through the placenta (Garcia-Bournissen et al., 2007; Smith, LaGasse, et
al., 2006; Wouldes, LaGasse, et al., 2004). Developmental effects include premature delivery
(Wouldes, LaGasse, et al., 2004; Eriksson, Larsson, and Zetterström, 1981) and smaller size
(Smith, LaGasse, et al., 2006; Smith, Yonekura, et al., 2003; Little, Snell, and Gilstrap, 1988;
Oro and Dixon, 1987). Children of meth users may remain small even through childhood
(Eriksson, Jonsson, et al., 1994; Eriksson and Zetterström, 1994). The smaller size is evident
even controlling for other factors that correlate with meth use, including polydrug use and
tobacco and alcohol use (Smith, LaGasse, et al., 2006). Research on other effects is limited,
3
Pregnant women aged 15 to 44 entering treatment were more likely than nonpregnant women to report cocaine or crack
(22 versus 17 percent), amphetamine or methamphetamine (21 versus 13 percent), or marijuana (17 versus 13 percent) as
their primary substance of abuse relative to nonpregnant women. See DASIS (2004).
24 The Economic Cost of Methamphetamine Use in the United States, 2005
but there is some evidence that prenatal meth use translates into worse neurodevelopmental
outcomes (Šlamberová, Pometlová, and Rokyta, 2007; Frisk, Amsel, and Whyte, 2002), hyper-activity, short attention span, learning disabilities (Plessinger, 1998; Woods, 1996), type 2 dia-betes and metabolic syndrome, and a collection of heart-attack risk factors, such as high blood
pressure and obesity (Wouldes, LaGasse, et al., 2004; Wouldes, Roberts, et al., 2004). There
are other observed effects in children. In isolated and rare cases, cardiac defects, cleft lip, and
biliary atresia have been observed, although the studies are not conclusive (Plessinger, 1998).
There are also cases of low activity (Smith, LaGasse, et al., 2008), poor feeding and tremors
(Oro and Dixon, 1987), and even fetal death (Stewart and Meeker, 1997). Animal experiments
show similar effects and indicate that these mental and coordination effects may last more
than a single generation (Šlamberová, Pometlová, and Rokyta, 2007). Some of the effects on
fetal development may follow from poorer parenting behaviors (Derauf et al., 2007), but such
diminished behaviors were also seen in randomized trials with mice and may therefore, to
some extent, be induced by meth (Šlamberová, Charousová, and Pometlová, 2005).
Health Service Costs Associated with Amphetamine Use
As we noted in our discussion of hospital-based treatment for substance abuse and dependence,
it is not possible to distinguish a mention of meth specifically from general amphetamine use
and abuse in the NIS data using ICD-9 codes.
4
Thus, as was done in the treatment section, we
use our 0.931 adjustment factor calculated from the 2005 TEDS data to determine the pro-portion of the hospital episodes that are likely due to meth specifically rather than all amphet-amines when generating estimates of events and costs.
Amphetamine-Induced Conditions
The literature provides evidence of meth’s association with a wide variety of conditions, some
of which are clearly generated only by meth use. Using our interpretation of the literature and
common use of ICD-9 codes, we identified nine primary conditions in the NIS that we con-sider meth-induced: fetal dependence, drug-induced neuropathy, drug-induced mental health
disorders, mental health and drug screens, poisoning by psychostimulant drugs, skin infec-tions, bacterial skin infections, other skin inflammation, and chronic skin ulcers. Although
we recognize that it is possible to suffer many of these skin conditions in the absence of meth
use, we elected to include skin conditions in our assessment of meth-induced costs because
they have been widely cited as a common consequence of meth use (e.g., N. Lee et al., 2005;
Lineberry and Bostwick, 2006; Richards et al., 1999). Meth users are particularly prone to
skin infection, lesions (e.g., excoriations and ulcers), and, consequently, cellulitis, potentially
stemming from delusion-induced scratching, needle marks, and chemical burns.
For a particular condition to be considered meth-induced, the patient record must include,
as a nonprimary diagnosis, the abuse of or dependence on amphetamine. In other words, the
costs of admissions for these nine primary diagnoses are allocated to meth only when amphet-amines are cited as an additional diagnosis. We are, however, mindful that the presence of an
amphetamine mention on the record does not necessarily imply that all the costs should be
allocated to meth when multiple substances are present and contribute to the admission. For
4
The specific ICD-9 codes used to identify these diagnoses in the hospital data are available on request.
The Cost of Methamphetamine-Related Health Care Among Methamphetamine Users 25
example, fetal dependence may result from the mother’s use of alcohol as well as amphetamine.
Allocating all the costs of these conditions to amphetamine even when other substances are
also present could overestimate the role of amphetamines. Therefore, we provide alternative
estimates of these costs.
Our lower bound attributes the cost to amphetamine only when no other drugs or alcohol
are present. Estimates are then revised downward by our adjustment factor of 0.931 to reduce
the total costs proportionately to the share of amphetamines that are due to meth. Table 3.2
shows that the health costs of inpatient hospital stays associated with these conditions are at
least $27.1 million. Our upper-bound estimates include all admissions for these diagnoses in
which amphetamine alone or amphetamine plus alcohol is present on the record. The inclu-sion of these additional cases increases costs by nearly $8.5 million. In some cases, such as
neuropathy, the numbers of cases are identical. For other diagnoses, such as fetal dependence,
the number of cases increases when alcohol is allowed on the record. However, it is not pos-sible to determine whether the diagnosis (e.g., fetal dependence) is due to the mother’s alcohol
use or to her meth use. As a result, our best estimates are based on the cases in which no other
substances (including alcohol) appear on the record.
5
Our focus on admissions that mention only amphetamine reduces the number of included
cases and the associated costs relative to admissions that include amphetamine and other
Table 3.2
Cost of Methamphetamine-Induced Hospital Stays for Which All Costs Were Attributed to
Methamphetamine in 2005
Condition
Lower Bound
Best Estimate ($)
Upper Bound
Admissions Cost ($) Admissions Cost ($)
Fetal dependence 146 426,756 426,755.99 156 442,231
Neuropathy 5 20,677 20,677.00 5 20,677
Substance-induced
psychosis
2,965 9,938,985 9,938,984.62 3,541 13,270,683
Skin, infection,
bacterial
5 25,694 25,694.16 5 25,694
Skin, infection,
other
2,366 13,895,847 13,895,846.92 2,714 16,583,612
Skin, ulcer 55 970,084 970,084.36 67 1,057,686
Skin, other
inflammation
15 84,536 84,535.85 15 84,536
Injury, mental
health or drug
screen
15 125,126 125,125.56 151 2,155,048
Injury, poison by
psychostimulant
drug
217 1,569,863 1,569,862.93 282 1,914,593
Total 27,057,567 27,057,567.39 35,554,760
SOURCE: NIS (2005).
5
Table A.2 in Appendix A measures the increase in the number of inpatient days for meth-induced stays.
26 The Economic Cost of Methamphetamine Use in the United States, 2005
substances. The size of the reduction indicates that other substances, particularly alcohol, may
prove intrinsic to understanding amphetamine abuse and its health effects.
Amphetamine-Involved Conditions
As described in our literature review, meth has also been cited as a contributing factor in that
it may exacerbate or lead to a wider range of health concerns. For these conditions, we cannot
credibly use the NIS to distinguish which cases are causedby meth use from those that are
exacerbatedby use. Nor can we identify whether meth caused the injury or is simply associated
with similar behaviors (e.g., risk-taking). Time does not permit us to thoroughly explore the
causal association between amphetamine use and each of the health conditions to which it is
related. But we can generate a potentially conservative estimate of these costs by estimating the
incrementalcosts associated with each primary diagnosis group when amphetamine is present.
That is, how much more does it cost to treat a patient with a cardiac condition who also pre-sents an amphetamine diagnosis than to treat one who presents the cardiac condition alone?
6
We do so in a regression context in order to control for other factors that might influence costs.
Failure to control for characteristics of the patient, the stay, and the patient’s health behaviors
would inappropriately attribute the costs of these complications to amphetamine. For example,
a patient with a heart condition may experience complications because of meth use or because
of tobacco use. Likewise, costs may be higher for patients who are older or whose stays com-prise a greater number of procedures.
Regression analyses were conducted to examine the incremental health care costs as indi-cated through higher costs (or longer stays; see Appendix A) for 11 general condition types:
cardiovascular, cerebrovascular, dental, fetal, injury, liver or kidney, lung, nutritional, sexually
transmitted, skin, and mental health.
7
For each of these conditions, we selected only the sub-conditions that amphetamine could influence, based on our understanding of the literature.
For example, clogged arteries are not included in the cardiovascular conditions. We also ana-lyzed the selected subconditions individually. The number of subconditions varied from zero
for dental to approximately a dozen for injuries. Focusing on subconditions can reduce the
likelihood of finding a significant relationship because conditions co-occurring with amphet-amine become rare. However, in other cases, focusing on particular subconditions sensitive
to meth use can improve our ability to isolate an effect by excluding conditions that are less
sensitive.
The specifications for inpatient costs were estimated in logs with an indicator variable
for amphetamine use. The regressions control for a variety of individual characteristics (e.g.,
age, race or ethnicity, gender), characteristics of the stay (e.g., primary payer, number of proce-dures), and hospital characteristics (e.g., size, location, region). Importantly, the regressions also
control for patient behavior, including use of alcohol, tobacco, and other substances. Regres-sions for the overall primary condition and those for the sum of the effects across subcondi-tions within each condition category produced similar estimates for most conditions. As the
estimates from these two approaches differ by less than $1 million in all but one case, we opt
6
Unobservable factors may also be generating the differences in incremental costs. However, given that the hospital
and attending physicians are the ones making decisions on procedures and LOS (i.e., the two primary drivers of costs),
the unobservable factor would have to influence medical decisions regarding treatment differentially. This seems highly
unlikely, given that we are focusing on meth as a comorbidity here, not as the primary reason for the visit.
7
The specific ICD-9 codes used to identify these diagnoses in the hospital data are available on request
for the estimate generated by the overall condition regressions rather than those for the subcon-ditions. The one exception is injuries, which we describe next.
We added an analysis for injuries because injuries can be identified using ICD-9s and
external-cause codes (E-codes). In addition to the overall and subcondition analyses, we exam-ined cases with any injury diagnosis in addition to cases with only a primary injury diagnosis.
There are 12 injury subconditions to which meth might contribute. These include injuries due
to contusions, burns, external causes, suicide, cuts, guns, fire, machine, motor-vehicle trans-port, other transport, falls, and drowning. Only the first four categories can be examined in
the same framework as our other conditions because they can be identified using ICD-9s as
well as E-codes.8
The overall injury regression (using primary condition only) showed no incre-mental costs due to meth. However, the first four subconditions (based on primary ICD-9
only) showed incremental costs totaling at least $2.3 million resulting from injuries due to
external causes and, to a lesser extent, contusions. We also use the lower bound of the con-fidence interval from this analysis to generate our lower bound of $1.6 million. However, in
our upper bound, we want to accommodate the use of E-codes. Outcomes specific to injury
classifications were re-estimated for cases in which any injury-related code (ICD-9 or E-code)
was present on the record. This approach results in significantly higher incremental costs due
to amphetamine use ($18.9 million), but the lack of consistency with the primary diagnosis
results makes these regressions suspect. Consequently, we use the point estimate of the “any
injury” results only to inform our upper bound.
With the exception of injuries, the best estimate for each meth-involved condition is gen-erated using the point estimates from regressions of the increase in costs associated with having
amphetamine present on the record for that condition. The estimates yield a total increase of
$14.3 million after our 0.931 adjustment (see Table 3.3).
9
The lower bounds for the estimates
are constructed using the lower bound of a 95-percent confidence interval around point esti-mates. Thus, the incremental costs of hospital stays complicated by meth are at least $8.2 mil-lion. The upper-bound estimate is based on the upper bound of the 95-percent confidence
interval and yields an upper bound of $36.7 million. The exception to this methodology is the
calculation of injury costs, which are discussed in the preceding paragraph. Most of the differ-ence between our best and upper estimates is due to an increase in injury costs based on our
alternative analysis. However, the alternative analysis increases our best estimate by less than
$1 million.
Table 3.3 does not include any incremental costs for skin conditions because all four rele-vant skin conditions are already included in the meth-induced admissions (Table 3.2), in which
100 percent of the costs associated with treating the cases with meth as a comorbidity were
included in our estimate of meth-induced costs. The table also excludes fetal meth dependence
for the same reason, although other meth-related fetal conditions were examined. We did not
identify any statistically significant additional incremental costs for other fetal conditions.
8
Our methodology relies mainly on primary diagnosis. But E-codes are sometimes used to identify the external cause of
injury. For example, those injured in a motor-vehicle accident have a primary diagnosis (e.g., ICD-9) indicating the type
of injury (e.g., concussion) as well as the nature of the accident (e.g., motor-vehicle accident).
9
Table A.2 in Appendix A measures the increase in the number of inpatient days in meth-involved stays.
28 The Economic Cost of Methamphetamine Use in the United States, 2005
Table 3.3
Incremental Cost of Methamphetamine-Involved Hospital Stays: Conditions for Which
Methamphetamine Use Affects the Cost of Care Received, 2005
Condition Admissions Lower Bound ($) Best Estimate ($) Upper Bound ($)
Cardiovascular 4,446 2,708,003 5,569,334 8,598,777
Cerebrovascular 1,064 3,085,548 4,676,675 6,453,663
Dental 31 0 0 0
Other fetal problems 250 0 0 0
Injury 7,538 1,599,697 2,294,411 18,867,523
Liver or kidney 1,123 0 0 0
Lung 388 0 0 0
Nutrition 73 0 0 0
STDs 922 809,015 1,745,356 2,779,855
Skin
a
Mental health 26,146 0 0 0
Total 8,202,265 14,285,777 36,699,819
SOURCE: NIS (2005).
a
Skin costs are already accounted for in the meth-induced costs (see Table 3.2)
Suicide Attempts
The DAWN data system, managed by SAMHSA, provides estimates of the number of meth-involved suicide attempts nationally each year, along with a 95-percent confidence interval
that considers sampling issues involved in generating those national estimates. According to
the 2005 DAWN report (SAMHSA, 2007b), there were 3,155 ED cases involving attempted
suicide in which meth was the primary drug and primary reason for visit. The 95-percent
confidence interval surrounding this point estimate for 2005 ranged from 1,221 to 5,088 sui-cide attempts nationally. While there is clear evidence that meth-dependent individuals have
high rates of depression and suicidal ideation (Kalechstein et al., 2000; Zweben et al., 2004;
Glasner-Edwards et al., 2008), it cannot be determined from the current science what fraction
of suicide attempts involving meth use are truly caused by meth. In the absence of information
to guide how to adjust these numbers, we simply take these estimates as our best estimates of
the number of meth-related suicide attempts (see Table 3.4). To the extent that meth is simply
used coincidentally by individuals attempting to commit suicide rather than being the factor
that caused the depression or behavior that led to the suicide attempt, our estimates of costs
using these numbers will overstate the true burden of meth in this respect. However, it is also
true that not all suicide attempts result in a visit to an ED. Some suicide attempts may simply
result in visits to a mental health professional or other medical professional or require no
medical attention. Hence, it is also likely that the DAWN ED data on meth-involved suicide
attempts under estimate the total number of meth-involved suicides in the nation. It is impos-sible to know to what extent these two biases offset each other, if at all. Future research should
investigate this association between meth use and suicide attempts more carefully.
The Cost of Methamphetamine-Related Health Care Among Methamphetamine Users 29
Table 3.4
Calculating the Medical Cost of Methamphetamine-Involved Suicide Attempts
Component of Cost Calculation Lower Bound Best Estimate Upper Bound
1 Number of meth-involved attempts 1,221 3,155 5,088
2 Median cost per episode (2005 $) 8,174 8,174 8,174
3 Average ED episode cost (2005 $) 701 701 701
4 Total cost of suicide attempts (0.51 ×
row 1 × row 2 + 0.49 × row 1 × row 3)
(2005 $)
5,509,641 14,236,622 22,959,092
To calculate the cost associated with these suicide attempts, we rely on work by Corso
et al. (2007), who show that the medical cost in 2000 dollars for a nonfatal, hospitalized sui-cide attempt is $7,234.
10
Although we do not know from DAWN-published reports whether
all meth-involved suicide-attempt cases were admitted to the hospital, we know that, in gen-eral, about half (51 percent) of all substance abuse–involved suicide attempts were admitted
for inpatient hospital care. To the extent that meth-involved suicide attempts resemble other
suicide attempts once the patient is hospitalized, we can use the median cost for all suicide
attempts (reflecting the cost of a hospital admission) to help us approximate the total cost of
these suicide attempts. Inflating Corso et al.’s estimate to 2005 dollars using the medical CPI,
we find that the medical cost associated with a nonfatal, hospitalized suicide is $8,174.
For the other half of the episodes, which do not generate a hospitalization, we use infor-mation from a recent study by the Agency for Healthcare Research and Quality (AHRQ) pre-senting findings from the Household Component of the Medical Expenditure Panel Survey
regarding the average expenditure for a hospital ED visit (Machlin, 2006). The study reports
that, in 2003, the average expenditure for an ED visit, from all sources (e.g., private insurance,
Medicaid, Medicare, out-of-pocket payments) was $560. This estimate was highly sensitive to
the services received, however. For example, the average expenditure among patients needing
surgery was $904, while that involving just special nonsurgical services (such as laboratory
tests, X-rays, and radiological services) was $637. The average expenditure for visits during
which no special services were provided and no surgery was required was only $302. Given
that it is common procedure to perform lab tests on patients who come into the ED intoxi-cated, to confirm what substances have been ingested, we use the average expenditure estimate
including special services of $637. Inflating this estimate to 2005 dollars using the medical
care–services CPI gives us an average cost per ED episode of approximately $700.
To get the estimated cost of meth-involved suicide attempts, we take the weighted aver-age cost of the meth-involved suicide attempts, for which it is assumed that 51 percent of all
attempts result in hospitalization and the other 49 percent result in release from the ED to
home. We estimate that it costs more than $14.2 million to medically treat someone having
made a meth-involved suicide attempt, although there is some uncertainty around this esti-mate, as indicated by the lower and upper bounds.
10
It should be noted that there is no reason that a meth-involved suicide attempt would cost the same as other suicide
attempts. However, in the absence of good cost data for these ED events, we rely on the average for all suicide-attempt
events.
30 The Economic Cost of Methamphetamine Use in the United States, 2005
This estimate captures only the medical costs associated with unsuccessful suicide
attempts. It does not reflect the medical costs of successful suicide attempts nor the productiv-ity losses associated with them.
11
Because it is not possible to know the extent to which medi-cal care was involved in the emergency response to meth-involved deaths, we do not attempt
to approximate these costs here. Instead, we consider the additional cost of a full emergency
response to successful suicide attempts in a later chapter, in which we consider other direct
costs associated with meth-involved death.
Emergency-Department Care
According to the 2005 DAWN data report (SAMHSA, 2007b), there were an estimated
108,905 meth-related ED visits in 2005 (note that DAWN reports estimates for general stim-ulants and then breaks them out by all amphetamines and meth separately). The 95-percent
confidence interval reported for meth-specific ED episodes ranges from 58,469 to 159,340
visits. However, included in these ranges are meth-involved suicide attempts that were con-sidered in the previous section. To make sure we do not double-count these cases, we subtract
these suicide attempts from our estimates of ED visits, to give us a count of nonsuicide meth-involved ED episodes (see Table 3.5, row 3).
Although we can identify these cases as being meth-involved, we do not know what frac-tion of these are due purely to meth use. DAWN tracks a number of alternative reasons for the
ED visit, but, after 2003, reports separate estimates only for those having attempted suicide
and those seeking detox. As shown in row 4 of Table 3.5, only a small number of the non-suicide meth-involved ED mentions involve detox. However, one can confidently presume that
at least these cases can be causally attributed to meth use.
That leaves, in the “other” category, a number of indications, including overdose,
un expected reactions, and withdrawal, that could also be reasonably attributed to meth use.
Table 3.5
Methamphetamine-Involved Emergency-Department Episodes, 2005
Episode Type Lower Bound Best Estimate Upper Bound
1 Total number of meth-involved ED
episodes
58,469 108,905 159,340
2 Meth-involved suicide attempts 1,221 3,155 5,088
3 Nonsuicide meth-involved ED episodes 57,249 105,750 154,252
Detox 4,744 15,088 25,432
Other meth-involved cases 52,505 90,662 128,820
4 Presumed number of meth-caused
episodes (0.556 × other cases + detox
cases)
33,937 65,496 97,056
5 Unit cost ($) 701 701 701
6 Total ($) 23,786,783 45,906,852 68,027,521
SOURCE: SAMHSA (2007b).
11
Successful suicide attempts are captured as premature mortality in Chapter Four.
The Cost of Methamphetamine-Related Health Care Among Methamphetamine Users 31
In an attempt to capture at least some of these cases, we went back to the DAWN report for
the third and fourth quarters of 2003 (SAMHSA, 2004) and examined the extent to which the
conditions represented in the “other” category are clearly drug related. According to the 2003
report (SAMHSA, 2004, p. 58, Table 19), out of a total of 225,345 patients in what would
now be the “other” category, 125,351 (55.6 percent) had medical diagnoses clearly attributable
to drug use (i.e., abuse, dependence, overdose, toxic effects, and withdrawal). Unfortunately,
it does not break these down by specific drugs involved. Because no further information can
be obtained, we use this ratio (55.6 percent) to adjust down the number of meth-involved ED
cases falling into our “other” category to come up with the probable number of meth-caused
ED episodes for that category. We then add this to the total number of meth-involved detox
cases to generate our estimates of the presumed number of meth-caused ED episodes, as shown
in row 4 of Table 3.5.
As done previously, we use the AHRQ estimate of a hospitalized ED episode as our
best estimate of the cost of an ED visit. Multiplying this unit cost per ED visit by our esti-mate of meth-caused ED visits generates a best estimate of the cost of ED-related care of
$45.9 million.
Health Administration and Support
Table 3.6 shows how we calculate the share of the health administration and support dollars
that can be attributed to meth use. These costs include federal and state prevention efforts,
Table 3.6
Health Administration and Support
Category
and Source
of Budget
Information
Total Cost
(2005
$ millions)
Attributable Share (%) Meth-Related Costs (2005 $ millions)
Lower Upper Lower Bound Best Estimate Upper Bound
Federal
prevention
(ONDCP, 2006b)
1,530.1 0.50 12.42 7.65 98.84 190.04
State
prevention
(ONDCP, 2006a)
108.3 0.50 12.42 0.54 6.99 13.45
Prevention
research
(ONDCP, 2006b)
621.2 0.50 12.42 3.11 40.13 77.15
Treatment
research
(ONDCP, 2006b)
422.0 3.00 12.42 12.66 32.54 52.41
Training
(ONDCP, 2004a)
74.9 3.83 15.86 2.87 7.37 11.88
Insurance
administration
(Mark, Levit, et
al., 2007)
649.6 3.83 15.86 24.88 63.95 103.02
Total 3,406.06 51.71 249.83 447.95
SOURCES: ONDCP (2006a, 2006b, 2004a); Mark, Levit, et al. (2007).
NOTE: Due to rounding, numbers may not sum precisely
prevention and treatment research, training, and administration (ONDCP, 2006a, 2006b,
2004a; Mark, Levit, et al., 2007). To calculate our lower-bound estimates of federal and state
prevention efforts as well as federal prevention research, we use the annual prevalence of meth
use in the general household population from the 2005 NSDUH (0.5 percent) (NSDUH,
2006). We base our upper-bound estimate on the percentage of people entering treatment in
2005 who reported meth use (12.42 percent).
12
The same upper bound is used to allocate treat-ment research dollars to meth. The lower-bound estimate for treatment research is the share
of admissions in which meth is the only substance mentioned (3.0 percent). We also use the
TEDS data to construct fractions for the training and insurance administration budgets, but
these shares are constructed by excluding admissions in which alcohol is the only substance
mentioned (thus raising both the lower and upper attributable percentages to 3.83 percent and
15.86 percent, respectively).
Limitations
The health care costs appear quite low, in part because it is not possible to determine causality
for many conditions and thus our approach is to err in a conservative fashion. For example,
we include information on incremental hospital costs alone in cases in which causal relation-ships are not definitive. For some conditions we explored, we did not find a statistically sig-nificant effect of meth use for conditions in which the scientific literature suggests we should
find an effect. Mental health is a good example. This is not entirely surprising, as there may be
factors offsetting a meth-involved result. We might reasonably expect that substance use exac-erbates mental health conditions. But efforts on the part of addicted mental health patients
to exit the hospital more quickly, for example, can negatively influence our ability to identify
or quantify additional costs associated with their stays. Similarly, when considering national
data regarding ED visits, we do not include all meth-involved mentions, as we cannot be cer-tain that these ED visits were caused by meth use directly.
Of course, a significant omission in our estimate of the cost of meth-related health con-ditions is the lack of accounting of health-related costs for conditions treated outside the hos-pital setting. For example, the lack of a finding for dental costs in a hospital setting is perhaps
not very surprising, as most people are likely to be seen in a dental office for these types of
conditions. However, we are unaware of any national data that would enable us to estimate
the number of meth-induced dental visits in the general population, let alone information
on the average cost of those dental visits. The omission of such care as this, as well as other
urgent care received in physicians’ offices and private urgent-care clinics, is a major limita-tion of our results and the primary reason that we interpret these estimates as a lower bound
of the actual health-associated costs of meth use. We do, however, provide some thoughts on
the potential magnitude of these omitted costs in Chapter Nine.
12
Treatment for alcohol is included in the denominator. This percentage assumes that all amphetamine mentions from
Oregon are for meth.
33
CHAPTER FOUR
Premature Death and the Intangible Health Burden of Addiction
A major criticism of previous COI studies evaluating the economic cost of substance abuse
is that they ignore the intangible costs associated with drug abuse (Moore and Caulkins,
2006; Cook and Moore, 2000). Cook and Moore (2000) explain that this is because the COI
accounting framework is production-based rather than consumption-based. Hence, the subjec-tive value that individuals place on their health and on consuming these substances is ignored.
From a practical perspective, the primary justification for ignoring these costs is that adequate
data for estimating them have been difficult to come by. Methods have been developed in
other literatures to capture them, however. In the crime literature, for example, monetary esti-mates of fatality-related losses in quality of life (QoL) are constructed using information on the
amount that people regularly spend reducing their risk of death (Viscusi, 1993). In the case of
nonfatal injuries, estimates of the cost of pain, suffering, and fear as well as reduced QoL are
based on jury awards in trials involving nonfatal injuries (M. Cohen, 1990; M. Cohen and
Miller, 2003). Estimates of the economic cost of crime that employ these methods to account
for intangible costs are substantially larger than estimates ignoring intangible costs. For exam-ple, T. Miller, Cohen, and Wiersema (1996) find that the intangible costs of violent crime are
nearly twice the total tangible losses.
In this chapter, we break from previous COI studies and consider the intangible costs
associated with both premature death and the health burden of addiction. We present these
losses in their natural units (specifically, the number of deaths and quality-adjusted life-years,
or QALYs), so that our estimates of premature mortality and QALYs can be easily compared
with those from other studies that might put different monetary valuations on these out-comes. The cost of premature mortality is generated using an estimate of the value of life that
is based on a review of the literature employing the revealed preference approach rather than
the human capital approach (described below) (Viscusi and Aldy, 2003; Aldy and Viscusi,
2008). Although the human capital approach is the more commonly used method for valuing
premature mortality in the substance abuse literature, recent international studies are more
frequently adopting alternative methods that capture some of the consumption value or intan-gible value of life (Collins and Lapsley, 2002, 2008; Godfrey et al., 2002; Priez et al., 1999).
We estimate the intangible health burden associated with being addicted, which represents
the reduced welfare experienced by individuals who must live with addiction. In doing so, we
employ a dollar conversion of QALYs that is based on the same statistical value of life used in
the valuation of premature mortality so that the two are internally consistent.
Our attempts to capture these nontraditional intangible costs generate estimates that
are, in fact, more consistent with unit-cost estimates being applied in other chapters of this
monograph, including our estimates of the cost of crime and child endangerment, which both
34 The Economic Cost of Methamphetamine Use in the United States, 2005
similarly incorporate the intangible burden of these problems. However, by incorporating these
costs, particularly the costs considered in this chapter, our estimates become less comparable to
previous estimates of the cost of illicit drugs in the United States.
Our findings in terms of the number of lives affected and their monetized value are sum-marized in Table 4.1. In terms of premature mortality, we estimate that 895 deaths in 2005
were caused by meth use and abuse, although the number is estimated imprecisely because of
uncertainty about the role that meth played in cases in which other substances (e.g., alcohol)
were also present or other conditions contributed (e.g., heart problems). Placing a monetary
value on these lives is a difficult and controversial thing, as discussed in the next section, but
if we take a conservative value of a statistical life from the revealed preference approach of
$4.5 million, we estimate that the dollar value of premature death is on the order of $4 billion.
Although substantial, even these costs are dwarfed by what we estimate to be the dollar value
of the intangible health burden associated with meth use, which is nearly $12.6 billion. The
intangible health cost of living with addiction is significantly greater than that associated with
death because substantially more people are affected by addiction. We conservatively estimate
that 44,313 QALYs were lost in total by individuals living with addiction in 2005.
In the next section, we describe how we reached our estimates of lost lives, lost QoL
for those living with addiction, and the monetary values placed on these lives. Clearly, the
monetization of these values is necessary for them to be included in a full cost estimate but
is something that is highly debatable and open to critique. We do our best to make the esti-mates as transparent as possible so that others who prefer alternative monetary values for life
years and QALYs may insert their preferred values into our calculations and generate their own
estimates.
Table 4.1
Summary of the Cost of Premature Mortality and Intangible Health Burden of Methamphetamine
Cost Lower Bound Best Estimate Upper Bound
Premature mortality
Lives lost to meth-caused premature death 723 895 1,669
Value of one statistical life ($ millions) 4.50 4.50 4.50
Value of premature death ($ millions) 3,253.50 4,027.50 7,510.50
Health burden of living with meth addiction
Lost QALYs associated with living with addiction 32,574 44,313 74,004
Value of one QALY ($) 284,283 284,283 284,283
Value of intangible health burden ($ millions) 9,260.23 12,597.43 21,038.08
Total lost well-being due to health and mortality
($ millions)
12,513.73 16,624.93 28,548.58
Premature Death and the Intangible Health Burden of Addiction 35
Premature Death
We begin our assessment of the burden of premature death by identifying the number of
deaths that can reasonably be attributed to meth use or abuse. Once we have identified our
estimate of these deaths, we discuss how we arrive at our valuation of premature mortality.
Number of Methamphetamine-Related Deaths
Our definition of meth-involved death is based in large part on the World Health Organi-zation (WHO) definition of drug-related death, relying on identification using WHO and
WHO Collaborating Centers for the Family of International Classifications for North Amer-ica (1992–1994). A case is counted as meth use–related when its underlying cause of death was
mental or behavioral disorders due to psychoactive-substance use or poisoning of accidental,
intentional, or undetermined intent.
The WHO identifies substance-related harmsas harmful use, dependence, and other
mental and behavioral disorders resulting from use of opioids (F11), cannabinoids (F12),
cocaine (F14), other stimulants (F15), or hallucinogens (F16); multiple drug use (F19); or acci-dental poisoning (X41, X42), intentional poisoning (X61, X62), or poisoning by undetermined
intent (Y11, Y12) by opium (T40.0), heroin (T40.1), other opioids (T40.2), methadone (T40.3),
other synthetic narcotics (T40.4), cocaine (T40.5), other unspecified narcotics (T40.6), canna-bis (T40.7), lysergide (T40.8), other unspecified psychodysleptics (T40.9), or psychostimulants
(T43.6). The T-codes are selected in combination with the respective X-codes and Y-codes
shown in Table 4.2 (EMCDDA, 2007).
1
The multiple-cause-of-death file attempts to identify those who died of the immediate
consequence of use (e.g., overdoses) as well as those who died from complications of long-term
substance abuse. There may be individuals for whom only the immediate consequence was
recorded (e.g., cardiac failure, fatal accident) without the history of substance use. Those deaths
would not be identified. The cause-of-death file also does not include the deaths of nonusers
who were victims of meth production or trafficking. Some of these deaths are captured later in
this monograph. For example, in Chapter Eight, we include deaths related to hazardous events
due to meth production, and deaths due to drug-related violence would be considered under
crime in Chapter Six.
Table 4.2
Codes for Underlying Cause of Death, by WHO Definition
Cause of Death Selected ICD-10 Code
Disorders F11–F12, F14–F16, F19
Poisoning, accidental X42
a
, X41
b
Poisoning, intentional X62
a
, X61
b
Poisoning, undetermined intent Y12
a
, Y11
b
a
In combination with the T-codes T40.0–9.
b
In combination with T-code T43.6.
1
X-codes identify various types of intentional self-harm or assault that is related to the injury. Y-codes indicate that the
external event causing the injury is undetermined.
36 The Economic Cost of Methamphetamine Use in the United States, 2005
As we cannot strictly identify meth (or even amphetamine) in the ICD-10 codes used
to classify deaths, we first sum up all deaths involving psychostimulants. Our upper bound
involves all cases for T43.6 (psychostimulants) in which just the X- and Y-codes specified in
Table 4.2 are included along with the I-codes (those for cardiovascular conditions). This cap-tures all F15 cases that result in death as well. Our lower-bound estimate captures only T43.6
cases in which no other substance and no alcohol is included and just our X- and Y-codes are
included. Cross-tabulations created for us by staff at the CDC result in a lower-bound estimate
of 771 and an upper bound of 1,779 (see Table 4.3). Our middle estimate of 954 is based on
a total number of T43.6 cases in which no other substance and no alcohol was included but
additional codes using the X-, Y-, and I-codes were captured (for example, cardiovascular con-ditions). Consultation with staff at the CDC suggests that this is likely an accurate reflection
of meth-induced deaths even though these additional codes are not captured in the WHO
definition of meth-related death.
In order to isolate meth’s role, we use information from the 2005 TEDS, which shows
that meth was the only psychostimulant reported in 93.8 percent of psychostimulant-only
treatment admissions. Thus, to adjust the CDC estimate to account only for meth-related
deaths, we take 93.8 percent of these totals (see the second row of Table 4.3). Thus, our best
estimate of the number of meth-induced deaths in 2005 is 895 deaths, but there is some
uncertainty about this estimate as indicated by the range given by our lower- and upper-bound
estimates. This uncertainty stems from our attempt to reduce the potential effect of alcohol on
these deaths, as both our lower and best estimates exclude cases also involving alcohol.
Placing a Value on Premature Mortality
There are three primary methods for estimating the value of a human life: the human capi-tal approach, the contingent-valuation approach, and the revealed preference approach. The
human capital approach bases the value of an individual life on the individual’s earnings for-gone due to the premature mortality. While this approach has the advantage of being based on
readily observable data for an individual, group of individuals, or even a society, it places more
weight on the lives of wealthy individuals than on those of poorer individuals. Moreover, it
does not capture the additional welfare that people gain (beyond production) from being alive.
The contingent-valuation approach, often referred to as willingness to pay (WTP), is one of two
alternative methods that attempt to measure the additional welfare gain associated with living.
The contingent-valuation method infers an individual’s willingness to pay to avoid death or
risk of death through answers to hypothetical questions and trade-offs presented to the indi-vidual and uses this to construct a value of life. Critics of this approach argue that answers
Table 4.3
Calculation of Methamphetamine-Involved Deaths and Costs
Cost Lower Bound Best Estimate Upper Bound
Psychostimulant deaths 771 954 1,779
Adjustment to capture meth-only cases 723 895 1,669
Assumed value of one life ($ millions) 4.5 4.5 4.5
Total economic value ($ millions) 3,253.5 4,027.5 7,510.5
SOURCE: MCOD (2005).
obtained from hypothetical questions are unreliable and biased because people may not answer
truthfully if not actually faced with that decision. Thus, the revealed preference approach has
emerged as yet a third alternative for valuating life. In the revealed preference approach, infor-mation on the value of life is inferred from actual behaviors people take to reduce their risk
of death or injury. It overcomes the problem of contingent valuation by using real economic
decisions, such as paying for a burglar alarm, moving to a safer neighborhood, and extra com-pensation for riskier jobs.
There are strengths and limitations to each approach, and scientists have not developed
a consensus on a single preferred approach. None of these approaches generates a single point
estimate for the value of life, as each is influenced by population characteristics, variation
in population income, risk preferences, and life expectancy. A study by Hirth et al. (2000)
provides a review of studies adopting a variety of approaches and shows that estimates of the
value of life vary substantially both within and across methods. Estimates of the value of a life
using the human capital approach range from $460,511 to $611,151, while estimates from the
revealed preference approach, which includes WTP, range from $679,224 (using safety risk)
to $19,352,894 (using job risk), all measured in 1997 dollars. While it is clear that valuations
using the WTP approach are generally higher than those using the human capital approach,
substantial variation in the value of a life remains even when using the same basic approach.
For this study, we base our estimate of the value of a statistical life on findings from a
review conducted by Viscusi and Aldy (2003). They reviewed the economics literature examin-ing the value of a statistical life based on revealed preference decisions and conclude that cur-rent estimates of the value of life fall within the range of $4 million and $9 million per statis-tical life in 2000 dollars. This range far exceeds the $1 million valuation that has been widely
adopted in previous studies, but many studies employing this $1 million valuation have failed
to adjust for inflation since the original study year from which the $1 million valuation arose.
Because we wish to include the consumption value of life (and intangible cost) but still wish to
take a conservative approach, we adopt the lower-bound estimate of $4 million per statistical
life identified by Viscusi and Aldy (2003). We inflate this value of a statistical life to 2005 dol-lars to get $4.537 million but then round the estimate to the nearest hundred thousand to get
our final estimate of $4.5 million.
Using this somewhat conservative value of $4.5 million for a human life, we estimate that
the economic value of premature mortality associated with meth in 2005 exceeds $4 billion
($4,027 million). To demonstrate the variability in this estimate due solely to our uncertainty
regarding the actual number of meth-induced deaths, we show in Table 4.3 upper- and lower-bound estimates of the economic value of premature death, which ranges from $3.2 billion
to $7.5 billion. Thus, even when we assume a constant value of a lost life, we still see a rather
substantial range in estimates of the value of these lost lives because of the enormous value that
each life has in this calculation.
The decision to use $4.5 million as the value of life has considerable influence on our esti-mates. If we instead use the midpoint range provided by Viscusi and Aldy (2003) ($6.5 mil-lion) and inflate it, we get a 2005-dollar estimate of $7,057,700. Naturally, when multiplied by
the lower-, best-, and upper-bound estimates of meth-related deaths, this life value generates
an even larger total range of economic values of lost productivity ranging from $5.1 billion to
$11.8 billion. Alternatively, if we employ an extremely conservative value of $1 million (which
has been applied in COI studies for 20 years without adjusting for inflation), we get a smaller
overall economic value of $895 million but still with a considerable range of $723 million to
38 The Economic Cost of Methamphetamine Use in the United States, 2005
$1.7 billion. These examples demonstrate the sensitivity of our estimates to alternative assump-tions regarding the value of a statistical life, which remains a highly contentious issue in the
scientific field with no general consensus (Hirth et al., 2000; J. King et al., 2005; Aldy and
Viscusi, 2008).
The Cost of the Health Burden Associated with Being Addicted
Addiction and drug dependence reduce the QoL of those suffering from the condition, indepen-dent of its potential effects on productivity, employment, or health-service utilization. Health
improvements (recovery from addiction) translate into direct welfare gains for those affected by
the illness as well as indirect gains for those who care for or live with the individuals afflicted.
It is difficult to place a monetary value on the burden that addiction creates, but failing to do
so significantly underestimates the full burden of the disease. We now attempt to place bounds
on the probable loss associated with the health burden of addiction as a disease. We provide
estimates of this burden in terms of lost QALYs and in terms of the economic (monetary) value
of this health burden. Additional welfare losses are also borne by those addicted, their family
members, and other nonusers. Some of these losses, such as reduced employment and increased
involvement in crime, are captured elsewhere in this monograph, but others, such as family
burden, remain unmeasured.
A number of methods have been used to try to quantify the loss in well-being associated
with various health conditions, including cancer, multiple sclerosis, liver disease, hypertension,
and HIV/AIDS. By far, the most common method that has been adopted in the international
health literature is QALYs.
2
The QALY approach presumes that the impact of health problems
on the overall QoL can be quantified through trade-offs that people would be willing to make
between alternative health states, given variations in the length of time they would live with
each. In one formulation, respondents are asked the amount of lifetime they would be willing
to sacrifice in order to be relieved from a health problem (i.e., time trade-off ). In another for-mulation, respondents are asked about whether they would prefer some imperfect health state
with certainty, or a gamble between life and death with varying weights on each (i.e., standard
gambles). QoL measures are then constructed using weights obtained from these questions and
believed to measure a person’s own valuation of his or her current or alternative health states.
QoL, therefore, represents the valuation an individual places on his or her QoL when living
with a particular disease state or health condition; it is equivalent to 1 minus the loss in QoL
caused by having the disease.
Several health state–classification systems, such as the EuroQol (EQ-5D), SF-36® health
survey, the Multidimensional Index of Life Quality (MILQ) instrument, and the Quality
of Well-Being Scale (QWB), have been developed by researchers to assist in the translation
of health functioning into numerical scales (Drummond et al., 1997; Gold et al., 1996). The
results elicited can differ substantially based on who is making the trade-off (J. King et al.,
2005; Dolan et al., 1996; Read et al., 1984) or the time units in which they are calculated (Sin-2
QALY is a subset of a full class of quality-adjusted life indexes (QALIs) that have been developed to try to measure loss
in quality of life. What is unique about QALYs is that they measure QoL in terms of both the degree of the disability and
the survival probability of living with the illness. So the index is measured in terms of years of quality life. Other QALIs
can measure changes in well-being in terms of functioning or disability (disability-adjusted life-years, or DALYs).
Premature Death and the Intangible Health Burden of Addiction 39
delar and Jofre-Bonet, 2004). As is the case with valuations of human life, individual prefer-ences related to time preference, risk aversion, income, and wealth effects can all influence the
weights, as can the current health status of the individual.
Although QALYs are used broadly in the health literature, very little work has been done
applying QALYs to the burden of addiction. Most of the U.S. studies that have been done on
drug abuse focus on heroin addiction and the relative benefits of methadone maintenance as
an alternative form of treatment (Barnett, 1999; Barnett and Hui, 2000; Zaric, Barnett, and
Brandeau, 2000). Given the availability of a pharmaceutical therapy for heroin, one cannot be
sure that estimates of QALYs would apply to other drugs for which pharmacotherapies are not
widely available. Furthermore, addiction researchers have harshly criticized the main classifica-tion systems used to develop QALYs in the literature so far (e.g., the EQ-5D and the QWB)
because they do not capture the improvements in well-being that result from substance abuse
treatment that are nonmedical in nature, such as improved employment outcomes, obtaining
a supportive peer group, or reduced involvement in crime.
3
Only the Addiction Severity Index
(ASI) has attempted to capture some of these non–health-related aspects of the burden of the
disease, but its use for these purposes is just now becoming more common (Pyne et al., 2008;
Sindelar and Jofre-Bonet, 2004).
Given the debate regarding appropriateness of scales and their applicability to addiction,
we draw our estimates of the effect of dependence on QALYs from a recent study that set out
to investigate the construct validity of two generic preference-weighted measures for substance
use disorders (Pyne et al., 2008). Examination of preference-weighted scores for substance use–
disorder patients and other patient groups suggests that those suffering with a lifetime substance
use disorder and currently experiencing symptoms have a reduction in well-being of 0.126 or
0.141 depending on which preference-weighted index was used (Pyne et al., 2008). The self-administered QWB (QWB-SA), which is constructed from responses to questions on the ASI,
generated a reduction of 0.126, while the SF-12® standard gamble weighted (SF-12 SG), which
constructs the index based on responses to the SF-12, generated a reduction of 0.141. The dif-ference in lost QoL between these two alternative measures is fairly small, but both suggest a
somewhat smaller reduction than the 0.20 reduction in QALY based on samples of heroin users
(e.g., Zaric, Barnett, and Brandeau, 2000).
While the difference in QALYs in the Pyne et al. (2008) study suggests that the varia-tion across scales is fairly small, we cannot be certain that the small differences in lost QALYs
generated by the Pyne et al. study are not just a function of the preference-weighted scales
used or of the population surveyed (the full population, not heroin users). Thus, we take an
agnostic approach regarding which of these numbers is a better indicator of lost well-being and
simply use the Zaric, Barnett, and Brandeau (2000) estimate for our upper-bound estimate,
the QWB-SA for our lower-bound estimate, and the SF-12 SG as our best estimate, because it
falls within the range given by the other two (see row 1 of Table 4.4).
To calculate the total health burden caused by dependence, we now need to multiply
the reduction in QALYs due to dependence by the total number of people suffering from
meth addiction in 2005. Using TEDS, we reported in the previous chapter that there were
3
Indeed, even being an addict in recovery via nonmedicinal treatment is not the same as never having been dependent.
Having cravings that are held in check is still worse than not having cravings at all, even if it is much, much better than the
outcome of giving in to the cravings. Likewise, recovering addicts who change their lifestyles to avoid cravings (e.g., not
seeing friends who continue to use) still suffer some disutility from their condition
Table 4.4
QALY Approach to Estimating the Health Burden Associated with Methamphetamine Dependence
Cost Element Lower Bound Best Estimate Upper Bound
1 Reduction in QALY due to meth
dependence
0.126 0.141 0.20
2 Dependent users (TEDS) 160,101 160,101 160,101
3 Dependent users (NSDUH) 155,243 243,173 331,102
4 Percentage of meth-dependent users
who were not in treatment in the past
year (NSDUH)
63.4 63.4 63.4
5 Nontreated meth-dependent users
(NSDUH)
98,424 154,172 209,919
6 Estimated total number of dependent
users (TEDS from row 2 + NSDUH from
row 5)
258,525 314,273 370,020
7 Total QALYs lost due to meth dependence 32,574 44,313 74,004
8 Value of one QALY in 2005 284,283 284,283 284,283
9 Total value of the intangible health
burden per year ($ millions)
9,260.23 12,597.43 21,038.08
160,101 individuals in treatment in 2005. We need to add to this the number of people in the
general population experiencing dependence and not getting treatment. We need to rely on
another data source to capture information on individuals in need of treatment, so we turn to
the NSDUH. Each year, the NSDUH reports the number of people meeting Diagnostic and
Statistical Manual for Mental Disorders (DSM-IV) criteria for abuse or dependence for meth as
well as other drugs, in addition to reporting information on whether those individuals received
any drug treatment in the past year. According to the 2005 NSDUH, the population-weighted
average number of people meeting DSM-IV criteria for dependence was 243,173 people (with
a 95-percent confidence interval given by 155,243–331,102; see row 3 of Table 4.4). However,
more than one-third (36.6 percent) have had a previous treatment episode in the past year. To
avoid double-counting individuals already captured through TEDS, we want to subtract these
individuals from the number in NSDUH. In other words, we want to include in our total only
the 63.4 percent of individuals who met DSM-IV criteria for dependence but had not been
in treatment (see row 4 of Table 4.4). In row 5, we report the number of people in NSDUH
meeting DSM-IV criteria for dependence but who did not report going to drug treatment in
the past year. We then combine the total untreated dependent population reported in NSDUH
(row 5) with the total number of individuals receiving treatment for meth dependence in TEDS
(row 2) to get our total estimated number of meth-dependent individuals in 2005 (row 6).
Note that this number is almost certainly an underestimate, because there are most likely
some out-of-treatment meth-dependent users who fall outside the household-survey sampling
frame or who underreport conditions of their use in the survey (e.g., homeless users would fall
outside a household-based survey’s sampling frame). We have no empirical basis for estimating
the extent of this undercount. However, it is noteworthy in that the number of people known
to be receiving meth treatment (160,101) substantially exceeds the number of household-survey
respondents who report being meth dependent and receiving treatment.
Premature Death and the Intangible Health Burden of Addiction 41
We multiply the number of meth-dependent individuals by the reduction in QALYs asso-ciated with meth dependence to get our estimate of the number of QALYs lost due to meth
dependence. As reported in row 7 of Table 4.4, our calculations suggest that those dependent
on meth lost 44,313 QALYs. We have a lower- and upper-end estimate of the total number of
QALYs lost based on the range of reductions in QALYs provided by the different studies and
the estimated total number of dependent individuals. The range is pretty broad, from a low of
32,574 QALYs to a high of 74,004 QALYs, a breadth created by the near doubling in estimated
reduction in QALYs (from 0.13 to 0.20) as well as the significant variation in estimates of the
number of dependent users in the household population.
To assess monetarily the health burden associated with dependence, we start with esti-mates provided by French and his colleagues (French, Mauskopf, et al., 1996; French, Salomé,
et al., 2002). French, Salomé, et al. (2002) report that the dollar value of a quality-adjusted
life-day (QALD) for a 38-year-old white male with an average life expectancy and a $1 million
value-of-life estimate (assuming a 5-percent discount rate of future income) is $173.
4
If we mul-tiply this dollar value per QALD by 365 days, we get a dollar value per QALY of $63,174. This
dollar value per QALY, or $QALY, presumes a statistical value of a life of $1 million, which
was based on a literature review of the value of a life conducted in the early 1990s. As discussed
previously, more recent reviews of the literature place the value of a statistical life much higher,
in the range of $4 million to $9 million in 2000 dollars (Aldy and Viscusi, 2008; Viscusi and
Aldy, 2003). Thus, we need to scale up the $QALY estimate used by French, Salomé, et al.
(2002) to reflect this higher value of a statistical life. To generate a $QALY value assuming
a statistical value of life of $4.5 million, we can simply multiply the $63,174 by 4.5 to get a
$QALY of $284,270. This estimate is substantially larger than that suggested by Cutler and
Richardson (1997) and Tolley, Kenkel, and Fabian (1994) but well within the range of plau-sible values of a QALY based on the review conducted by Hirth et al. (2000). The larger value
merely demonstrates the sensitivity of these estimates to alternative assumptions regarding the
value of a statistical life.
5
We can now calculate the dollar value of a reduction in QALYs by simply multiplying the
reduction in QALYs associated with meth dependence (row 7) by the dollar value of a reduc-tion in QALY (row 8). As can be seen by the final row of Table 4.4, the estimated cost of the
intangible health burden of living with meth addiction in the United States in 2005 is quite
substantial. Our best estimate of these costs is $12.6 billion, with a range from $9.3 billion
to $21.0 billion. The QoL lost due to dependence, therefore, represents by far the biggest cost
component considered thus far.
The substantial variation between our low and high estimates implies that these estimates
are imprecise and merit further research. Indeed, our upper-bound estimate is more than twice
as large as our lower-bound estimate, and the difference between the bounds exceeds $10 bil-lion. The variation in cost estimates is driven by two factors: (1) the estimated number of
4
French, Salomé, et al. (2002) selected the 38-year-old white male as a reasonable characterization of the modal person in
treatment for the time period they examined. According to our analysis of the 2005 TEDS, the typical (median) individual
in treatment with a primary diagnosis of meth was male, white, and between the ages of 30 and 34, so this representation
is not too far off that for the median meth-dependent individual in treatment.
5
It is interesting to note that our $QALY of $284,270 is very close to the statistical value of a life year estimated by Aldy
and Viscusi (2008) for a 38-year-old, using a cohort-adjustment model (value of a statistical life year is approximately
$310,000, assuming a 3-percent discount factor).
42 The Economic Cost of Methamphetamine Use in the United States, 2005
people who are dependent and (2) $QALY, which is measured imperfectly. While both sources
of uncertainty are important and generate substantial variation by themselves, the uncertainty
related to the proper value for a QALY is an area in significant need of further research (J. King
et al., 2005).
Limitations
While these estimates represent our best attempt to quantify the costs associated with prema-ture death and the intangible burden of addiction, the numerous assumptions and qualifica-tions clearly influenced the estimates we obtain. Perhaps the assumption having the largest
effect on both of these estimates is the assumed value of a statistical life ($4.5 million), which
affects both the valuation of premature mortality as well as our dollar conversion of QALYs.
While we believe that this is a fairly conservative estimate of the value of a life, we recognize
that others, particularly those in favor of the more traditional human capital approach, might
deem this too large. Still others might consider this value too low. Given the significant debate
surrounding the issue of the value of life, we did our best to make our calculations as clear as
possible so that others can generate estimates with their own preferred values.
It is important to note that precision in estimating the number of meth-induced deaths
is also critical to reducing the variability in our estimates. Our current estimates, while vari-able because of alternative assumptions regarding the inclusion of cases involving alcohol, still
exclude numerous other conditions in which meth might have played a critical role in the
death, such as highway fatalities and explosions and physical assaults ending in death. Thus,
future work should consider other external causes of death that might be highly associated
with meth use to get a better sense of the extent to which meth plays a role.
The tremendous size of the intangible health burden suggests that careful attention and
consideration should go into future construction of this estimate. This value is sensitive to
assumptions regarding the loss in QALYs due to meth dependence, the dollar value of that
loss in QALY, and the estimated number of individuals addicted to meth. Each of these could
be explored in greater detail to reduce the variability. More accurate attempts to estimate the
number of people living with addiction need to be considered, as household populations clearly
omit arrestees and marginalized populations that are likely to exhibit higher rates of depen-dence. Thus, future work that can help in the estimation of addiction in these missing popula-tions would be tremendously useful for all of these estimates. And clearly, the magnitude of
these costs will be dramatically influenced by the dollar value per QALY. Future work that
can help narrow the range of plausible estimates for these values may improve the precision of
future estimates of these costs.
Finally, we should note that there are other components of lost welfare caused by addic-tion that are not captured in these estimates, including the burden placed on family members
or friends who live with a loved one suffering with meth addiction. These intangible costs,
which have yet to be reflected in any of the estimates provided thus far, may prove to be quite
substantial and, consequently, may represent a fruitful area for further research.
Productivity Losses Due to Methamphetamine Use
Although there is a widespread belief that drug use reduces productivity, the scientific litera-ture examining the relationships between drug use and wages is inconclusive (French, Zarkin,
and Dunlap, 1998; Kaestner, 1998, 1994a, 1994b). Several economists have offered theories as
to why, but difficulties in measuring true differences in individual’s abilities, selection effects
related to job choice and early job entry, and the importance of nonmonetary compensation
(e.g., health benefits, flexible hours) make it difficult to truly identify the reason for the differ-ent findings. Thus, efforts have increased to consider alternative ways of measuring productivity
effects, including decisions related to labor supply, schooling, and absenteeism. Our estimate
presented in this chapter follows these efforts and attempts to quantify the costs associated
with meth-related productivity losses in four key areas: (1) higher probability of unemployment
(and the lost earnings associated with that reduced employment), (2) absenteeism from work,
(3) lost productivity associated with incarceration, and (4) other employer costs. Although this
last category does not really represent a reduction in productivity of workers, it does capture
the value of resources that companies expend to deal with costs caused by meth-using employ-ees and hence represents resources that might otherwise be dedicated to increasing production
or production efficiency. Thus, we include them here.
Table 5.1 summarizes our lower, upper, and best estimates of the cost associated with lost
productivity. In looking at the costs included in our estimate, the biggest factors in our mea-sure of lost productivity come from general absenteeism (missed workdays) and incarceration,
Table 5.1
Summary of Productivity Losses Associated with Methamphetamine Use ($ millions)
Loss Lower Bound Best Estimate Upper Bound
Lower probability of working 15.26 62.53 93.46
Absenteeism
Time in treatment 25.37 25.37 25.37
Other missed workdays 40.02 249.73 422.06
Incarceration 274.75 305.35 336.11
Additional employer costs
Drug testing 24 44 177.92
Higher health care and benefit costs NI NI NI
Total 379.40 686.98 1,054.92
44 The Economic Cost of Methamphetamine Use in the United States, 2005
which, according to our best estimate, together represent more than half a billion dollars. The
range provided for these numbers is equally insightful, however, and indicates the relatively
better data we have available on the processing of drug offenders than on the productivity
effects of meth users. The considerably larger variation observed for missed workdays due to
absenteeism is driven by uncertainty in a number of the components that make up these costs,
and future research in this area would dramatically improve these estimates and reduce the
uncertainty.
Most of the estimates provided in Table 5.1 were constructed through our own analysis of
key measures from available data sets. Information on lost productivity due to unemployment
and absenteeism at work comes from the NSDUH, while information on absenteeism related
to time in treatment comes from TEDS. Information on lost productivity due to incarceration
of sales-related offenses comes from the Bureau of Justice Statistics (BJS) and State Court Pro-cessing Statistics. Information on weekly wages and earnings, which is our primary method
for valuing the lost work time, comes from the Bureau of Labor Statistics (BLS). Information
on the number of people subject to employee drug testing comes from the U.S. Department
of Transportation and the NSDUH, while information on the cost of a drug test comes from
summary information provided by the Office of National Drug Control Policy (ONDCP).
Full details regarding these estimates and what is considered in them are provided next.
Literature Review
There are no published peer-reviewed articles examining the effects of meth use specifically on
wages or other direct measures of productivity on a nationally representative or probabilistic
sample. Quite an extensive literature exists examining the impact of substance use more gener-ally on worker productivity and labor market participation. Substance use is believed to dimin-ish productivity and lead to poor labor market outcomes for several reasons. First, it may delay
initiation into the work force, thereby reducing experience and human capital accumulation
associated with on-the-job training. Second, it may decrease the probability of being employed,
which, again, may interfere with human capital accumulation (Gill and Michaels, 1992; Reg-ister and Williams, 1992). Third, it may increase absenteeism, which directly influences the
productivity of not only the drug user, but also those individuals who work with that user
(French, Zarkin, and Dunlap, 1998; Bass et al., 1996). Finally, substance abuse may reduce an
individual’s productivity on the job, which should translate directly into lower wages if wages
are indeed a good indicator of marginal productivity (Hoyt, 1992).
Empirical studies that analyze the direct effect of substance use and abuse on earnings
have generated very mixed findings, however. Even after accounting for the endogeneity of
substance use, earnings of substance users are found to be higher by some researchers (Kaest-ner, 1991, 1994a; Gill and Michaels, 1992; Register and Williams, 1992; French and Zarkin,
1995), lower by others (Burgess and Propper, 1998; Hoyt, 1992), and either statistically insig-nificant or not determinable by others (Kaestner, 1994b; Zarkin et al., 1998). The lack of a
robust finding has led many economists to focus on other measures of productivity, such as the
probability of being employed or unemployed (Bray, Zarkin, Dennis, and French, 2000; Reg-ister and Williams, 1992; Kandel and Davies, 1990). Here, too, the evidence is mixed. Using
the 1984 and 1985 waves of the National Longitudinal Survey of Youth (NLSY), Kandel and
Davies (1990) find that use of marijuana and cocaine in the past year is positively associated
Productivity Losses Due to Methamphetamine Use 45
with the total number of weeks unemployed. However, using data from the 1984 wave of the
NLSY, Register and Williams (1992) find that use of marijuana on the job in the past year and
long-term use of marijuana both have a positive impact on the probability of being employed.
Higher frequency of marijuana use in the past month, however, did lower the probability of
being employed.
A number of factors contribute to the lack of a robust finding. First, studies examine the
impact of substance use on earnings and labor market outcomes for populations of varying
ages. While some studies focus on young adults (Kandel and Yamaguchi, 1987), others focus
on mature young adults (Kandel and Davies, 1990; Register and Williams, 1992), and still
others focus on the full adult population (Bray, Zarkin, Dennis, and French, 2000; Zarkin
et al., 1998). It is quite possible that the nature of the relationship between substance use and
labor market outcomes changes over the life cycle as job-market experience and job tenure
begin to dominate the effects of other individual determinants of labor market outcomes.
Indeed, a few studies have explicitly considered this fact and noted the differential effects of
substance use on wages conditional on age (Mullahy and Sindelar, 1993; French and Zarkin,
1995), but it is not a factor that has been consistently considered in the literature.
A second factor complicating the interpretation of findings from the literature is the
inconsistent treatment of indirect mechanisms through which substance abuse could affect
earnings—for example, through educational attainment, health, fertility, or occupational
choice. Given that these inputs have been established as important determinants of labor
market participation and wages (Becker, 1964; Mincer, 1970; Willis and Rosen, 1979) and
that there are strong associations between these measures and drug use (Chatterji, 2006; Bray,
Zarkin, Ringwalt, and Qi, 2000; Kenkel and Wang, 1998; Cook and Moore, 1993; Mullahy
and Sindelar, 1994), it is important to consider whether analyses looking at the impact of sub-stance abuse on earnings consider the indirect effects as well.
Finally, the literature is inconsistent in its definition of substance use. Current use has been
defined as daily use (Kandel and Yamaguchi, 1987), use in the past month (Chatterji, 2006;
Cook and Moore, 1993), and use in the past year (Kandel and Davies, 1990; Register and Wil-liams, 1992; Mullahy and Sindelar, 1993). A few studies attempt to differentiate the effects of
chronic use from casual use (Roebuck, French, and Dennis, 2004; Kenkel and Ribar, 1994) or
to proxy chronic use with measures of early initiation (Bray, Zarkin, Ringwalt, and Qi, 2000;
Ringel, Ellickson, and Collins, 2006). Given all the different ways in which substance use can
be operationalized, with some representing more chronic or persistent use and others more
casual use, it is not surprising that findings vary across the studies.
It is clear that the relationship between substance use and abuse and labor market out-comes is dynamic and can be potentially influenced by the relationship between early sub-stance use and human capital production. The potential for reverse causality, however, is also
real. Just as substance use and abuse can lead to job separation and other poor labor market
outcomes, job separation may lead to increased substance use and abuse. In light of the poten-tial for feedback loops, it is important to use appropriate statistical methods that can isolate
the true nature of the relationship. Our own analyses attempt to do this through the use of
instrumental variable techniques (Greene, 2003; Angrist, Imbens, and Rubin, 1996; Bound,
Jaeger, and Baker, 1995).
In the sections that follow, we examine the impact of self-reported meth use in the past
year on a variety of labor market outcomes using data from the 2005 NSDUH. The NSDUH
is a nationally representative household survey of individuals 12 years and older conducted
46 The Economic Cost of Methamphetamine Use in the United States, 2005
by SAMHSA. The primary purpose of the survey is to accurately estimate the prevalence of
illicit drug, alcohol, and tobacco use in the United States among the household population.
The NSDUH sample is drawn from a clustered, multistage sampling design, resulting in a
nationally representative sample of noninstitutionalized civilians each year. Interviews occur
continuously throughout the calendar year and take roughly one hour to complete. To ensure
confidentiality, respondent names are not used, interviews are conducted in private, and sen-sitive questions about drug use are completed through audio computer-assisted self-interview
(ACASI) technology, with which the respondent keys answers directly into a laptop computer
in response to prerecorded instructions. Further information on survey methodology is pro-vided in the annual NSDUH findings report (NSDUH, 2008).
In recent survey years, respondents have been asked whether they have ever used meth
and the timing of use (i.e., in the past month, in the past year, more than a year ago). Table B.1
in Appendix B shows the unweighted and weighted average frequency of responses. It is clear
that self-reported use of meth is somewhat low, especially when compared with that of other
substances, such as alcohol or marijuana use. Given that the data are self-reported, there may
be significant underreporting, despite attempts to ensure protection of privacy and truthful
reporting. The impact of underreporting on the analyses would bias the estimated relationships
toward zero. Thus, to the extent that underreporting occurs, our estimates should be viewed
as conservative estimates of the true relationship. Unfortunately, no other externally validated
data (i.e., drug tests) are available on a national scale.
Lower Probability of Being Employed
We begin with an examination of the likelihood of being unemployed among individuals
between the ages of 21 and 50. By restricting our sample to individuals within this age range,
we hope to reduce the confounding effects caused by part-time students and those who are
retired or partially retired. In the 2005 NSDUH, respondents are asked whether, in the past
12 months, there was ever a period during which they did not have at least one job or business.
Respondents who reported an unemployment spell during the past year and did not previously
state that they were a full-time student or out of the labor force were then asked how many
weeks in the past year they were without a job. These two questions provide us with informa-tion for our main assessment of the impact of meth use on employment.
Substantial consideration was given to the proper specification of the empirical model.
Through a series of diagnostic tests, we identified the bivariate probit model as the most appro-priate methodology for estimating the joint probability of being unemployed in the past year
and reporting meth use in the past year. The bivariate probit explicitly considers the unobserved
correlation that is believed to exist between the decision to use meth and the probability of
being employed and explicitly models this correlation in the error terms using an assumption
of joint normality. Empirical tests of that assumption support the adoption of this model.
For the productivity models, all the models include as additional controls the individual’s
gender, race or ethnicity, educational attainment, age bracket, marital status, general health
status, number of children in the household under the age of 18, number of prior jobs, and
population density. The instruments used for identification of the causal association between
meth use and unemployment status include indicators of religious attachment (specifically,
the number of times in the month the respondent attends religious services), whether the
respondent has been offered drugs in the past 30 days, and the age of first use of marijuana, a
substance commonly believed to be a gateway drug to other illicit substances, including meth
Productivity Losses Due to Methamphetamine Use 47
(Yamaguchi and Kandel, 1984; Kandel and Yamaguchi, 1993). All of these instruments were
assessed in terms of their validity, exogeneity, and exclusion criteria using standard assessments
(see Appendix B).
The first row of Table 5.2 shows the results from our estimates of the impact of meth use
on the probability of being unemployed. Estimates reflect the percentage-point increase in the
probability of being unemployed, conditional on meth use in the past year. As the analyses in
Appendix B show, we consistently find that self-reported meth use in the past year has a posi-tive and statistically significant association with unemployment in the past year. These findings
hold even when the use of other substances is accounted for. Results from our preferred models
show that the additional probability of a meth user being unemployed in the past year is 0.97
percentage points, which is about a 10-percent increase in light of the baseline unemployment
rate of 9.72 percent. The inclusion or exclusion of other variables does very little to change the
magnitude of the effect. The low and high values for the change in predicted probability are
relatively close to our best estimate.
In the second row of Table 5.2, we provide information on the population-weighted
number of people between the ages of 21 and 50 who reported meth use in the past year (i.e.,
our best estimate) in the 2005 NSDUH. The 95-percent confidence interval surrounding this
population estimate is used to generate the low and high estimates.
To construct an estimate of the cost of the unemployment spell, we must know the average
number of weeks for which meth users are unemployed. An examination of weighted means for
meth users and non–meth users in the past year reveals that meth users do not have, on aver-age, longer unemployment spells, although there is a lot more variability around their mean.
Additional regression analyses confirm that there is no statistical difference in average weeks
unemployed. In light of the lack of a statistically significant difference in weighted means, we
use the average number of weeks unemployed for the entire population in our estimate of the
Table 5.2
Impact of Methamphetamine Use on Income Due to Unemployment, Population Ages 21 to 50
Effect Low Estimate Best Estimate High Estimate
1 Increased likelihood of being unemployed
(from regression analysis) (%)
0.92 0.97 1.03
2 People using meth in past year (weighted
counts from the NSDUH)
647,195 870,205 1,093,216
3 Total number of meth-induced
unemployed (row 2 × row 1)
5,954 8,441 11,260
4 Average number of weeks unemployed
(weighted mean from the NSDUH)
12.75 12.75 12.75
5 Median weekly wage (from the BLS) ($) 201 581 651
6 Lost income per average spell (row 4 ×
row 5) ($)
2,563 7,408 8,300
7 Total lost income due to meth-related
unemployment (row 3 × row 6) ($)
15,260,102 62,530,928 93,458,000
effects of meth on unemployment (12.75 weeks). We multiply this average number of weeks by
the average wage
1
to generate the average loss per user attributable to his or her meth use.
Information on the median weekly wage comes from the U.S Department of Labor,
which collects information on wage and salaries through the National Compensation Survey
(NCS). In light of the significant skewness that exists in wage data, we use the median
weekly wage as our measure of lost income rather than the mean weekly earnings, as we
think that the former better approximates the typical wage lost by the average employee.
The median weekly earnings of full-time wage and salary workers for individuals 16 years
and older in 2005 is $651. The median weekly earnings of part-time workers in the same
age group in 2005 is $201. To know which estimate should be used as our measure of lost
income for meth users, we need to know whether meth users are more or less likely to be
employed full time or part time. Using information available in the 2005 NSDUH, we esti-mated the effect of self-reporting meth use in the past year on the probability of working full
time, conditional on working. We found that meth use had no statistically significant effect
on the probability of working full time. In light of this, we decided that we would use the
part-time median wages as a lower-bound estimate of the cost of unemployment due to meth
use and the median weekly wage for full-time and salaried workers as an upper-bound esti-mate. Our best estimate takes a weighted average of the full-time and part-time wages, where
the weights are given by the fraction of the meth-using population who work full and part
time $. .$ .$. 58080 0 156 201 0 844 651 r§
©
·
¹
r§
©
·
¹
We then calculate the typical lost income
associated with the average-length unemployment spell by multiplying results in rows 4 and
5 (results are rounded to the nearest dollar). The total lost income associated with meth-induced unemployment is the product of the number of meth users unemployed because of
meth (row 3) and the typical income lost (row 6).
The estimated range reflects the significant uncertainty in the effect of meth on the
probability of a spell, number of meth users, and weekly wages. The best estimate is around
$62.5 million, but the bounds on our estimates range from $15.2 million to $93.5 million.
Absenteeism
The literature suggests that employee drug use is associated with increased absenteeism (Bass
et al., 1996; Bross, Pace, and Cronin, 1992; Normand, Salyards, and Mahoney, 1990), though
the causality of the relationship is not established. The only economics study examining the
relationship between illicit drug use and absenteeism that explicitly dealt with the issue of
causality found no statistically significant effect between past-year drug use and absenteeism
(French, Roebuck, and Alexandre, 2001). However, the study did not explicitly examine meth
use. Given the lack of research on causality, we conducted our own analyses and identified two
potential sources of absenteeism: missed work due to time in treatment and missed work due
to other reasons.
We begin by estimating the lost productivity associated with time spent in drug treat-ment. We know from the 2005 TEDS that there were 28,055 individuals in treatment with a
primary diagnosis of meth abuse who worked full time at the time of their admission. Another
1
We have no information in the NSDUH that would enable us to estimate whether meth users have lower wages than do
non–meth users. Furthermore, there are inconsistent findings in the general literature on substance use. Given the lack of
better information, we presume that the average wage for the population is a suitable approximation for the loss associated
with increased unemployment.
Productivity Losses Due to Methamphetamin
12,689 patients reported being employed part time at the time of admission. Although not all
of these individuals were receiving therapy in a manner that would preclude them from work-ing, those participating in residential or intensive outpatient therapy would most certainly miss
work. This time away from work spent dealing with the drug problem represents a real cost to
the employer in terms of lost productivity and can be cleanly attributed to drug use itself.
We attempt to capture these losses by using information on the number of full-time and
part-time employees engaged in intensive drug treatment (either inpatient or intensive out-patient) and use information on average LOS by type of service setting to derive the average
time spent away from work. If we assume that it is only those individuals in residential treat-ment (hospital-based or free standing) or those receiving intensive outpatient therapy who are
unable to work during their time in treatment (largely because they are physically unable to
go to work), we find that there were 6,832 meth patients who were full-time employees miss-ing work and 2,894 part-time employees missing work because of their drug treatment in
2005. However, the typical course of treatment in terms of LOS varies quite substantially even
among these more intensive forms of drug treatment. Thus, instead of assuming that all full-time and part-time employees missed the same number of days independent of service setting,
we consider the LOS by service modality and construct a weighted average number of days of
missed work.
In Table 5.3, we show the breakdown of full-time and part-time primary meth clients
admitted to TEDS and receiving care in a residential treatment facility or intensive outpatient
treatment facility. The median LOS for each modality, from the 2005 TEDS Discharge Report
(SAMHSA, 2008a), is shown in the final column. Given the significant skewness in the distri-bution of LOS even within modalities, we use the median LOS rather than the mean LOS as
an estimate of the actual time spent in treatment in the third column of Table 5.3.
If we take a weighted average of the number of employees times the median LOS (mea-sured in days), we have an estimate of the total number of days missed by full-time and part-time employees while they attended drug treatment. Full-time employees missed 239,501 days
of employment, and part-time employees missed 107,805 days of employment (see Table 5.4,
row 1).
The first step in our calculation involves converting the number of days in treatment into
workweeks so that we can use median weekly wages for full-time and part-time employees
to generate our estimate of lost productivity in dollars. The second row of Table 5.4 divides
Table 5.3
Full-Time and Part-Time Employed Methamphetamine Patients in Treatment Facilities, 2005
Service Setting Full-Time Employees Part-Time Employees Median LOS (days)
Rehab, residential, hospital (nondetox) 117 41 3
Rehab, residential, short term 1,524 558 21
Rehab, residential, long term 1,271 660 46
Ambulatory, intensive outpatient 3,540 1,562 42
Total
a
6,832 2,894
SOURCE: SAMHSA (2008a).
a
Totals are slightly lower than stated in the text because information on where the patient is treated is not
identified for all individuals in the TEDS data.
50 The Economic Cost of Methamphetamine Use in the United States, 2005
Table 5.4
The Value of Lost Work Time Spent in Treatment for Methamphetamine
Time Lost Full-Time Employed Part-Time Employed All Employed
Total number of days absent due to treatment 239,501 107,805
Weeks absent due to treatment (divide first
row by 7 days per week)
34,214.43 15,400.71 49,615.14
Median weekly salary (from the BLS) ($) 651 201
Economic cost of time spent in treatment ($) 22,273,594 3,095,543
Total lost productivity caused by time spent in
treatment ($)
25,369,137
these days by 7 to generate the number of weeks absent. In the third row, we again report the
median weekly wage for full-time and part-time employees, as reported by the BLS. We mul-tiply these median weekly wages by the number of missed workweeks to get the total value
of lost work time for full-time and part-time employees separately (row 4). Summing these
together generates our total estimated cost associated with absenteeism of $25.4 million.
A major assumption in the construction of this estimate of absenteeism associated with
treatment is that the earnings of meth users are similar to those of the average person, as we use
the median wage of all employees, which includes meth users and non–meth users. Meth users
may have systematically lower earnings, so the use of a measure like the median weekly wage
might overstate the actual value of lost productivity. Unfortunately, we do not have data avail-able that will allow us to empirically test whether meth users who work earn less than nonusers
do. Thus, in the absence of any empirical support for this notion, we simply rely on aggregate
statistics for the population of workers as a whole. The implication of this assumption, if it is
true, is that our estimate of lost productivity associated with absenteeism due to treatment may
overstate the actual value of lost productivity.
A second major limitation of this calculation is that it assumes that everyone entering
treatment retains his or her job and cannot work if in intensive or residential therapy. Thus, it
assumes that the time spent in treatment represents productivity that is actually lost, and this
is an assumption that cannot be validated in our data. If the assumption is not true, then we
may be overstating the productivity losses of those in treatment. However, we also do not cap-ture those who lose their jobs because of treatment, either due to missing too much work or as
a penalty for being identified as a drug user. Thus, one could also argue that our estimates of
lost productivity are understated. We do not have data that could adequately inform us as to
which of these is more likely to be true, so our estimates must be viewed with these alternative
interpretations unanswered.
To consider the effect of meth use on absenteeism beyond days lost to treatment, we
examine self-reported days of missed work from the NSDUH. The 2005 NSDUH asks respon-dents about missed workdays through two specific questions:
How many days in the past month did you miss work because of injury or illness (i.e., t
“sick/injury days”)?
How many days in the past month did you miss work because you just did not feel like t
going to work (i.e., “blah days”)?
In the first and second rows of Table 5.5, we report the estimated effects (obtained from
regressions in Appendix B) for the number of workdays missed due to feeling blah and due to
sick and injury days, respectively. The low and high estimates in the first row (for blah days)
represent the limits of the 95-percent confidence interval for the estimated co efficient, which
represents our best estimate. In the case of sick and injury days (second row), the estimated
coefficient was not statistically significant at conventional levels of significance in all models.
Hence, in this case, we use 0 as our lower and best estimate (reflecting the lack of a statistically
significant result from the main instrumental variables [IV] model), and the high estimate is
based on the estimated coefficient from using the treatment regression model approach and
including substance abuse and income measures. In the third row, we just sum up the total
number of missed workdays, on average, per month due to meth use.
In the fourth row, we convert the missed workdays to total number of weeks missed. To
generate these measures, we multiply the estimate of workdays missed in the past month by
12 months and divide by 5 days. We use 5 here instead of 7 because most people work only
five days per week. In the case of our high estimate, the estimated number of weeks using this
method (11.66 weeks) exceeds that which a company would allow the typical worker to miss
unless the person was out on disability. Hence, for our upper-bound estimate, we impose the
restriction that the employee would not remain employed if he or she missed more than four
weeks of work in the year. This also ensures that we reduce the potential effect of double-counting individuals who are out due to illness related to their drug dependency, which was
captured in the previous table.
Table 5.5
The Value of Lost Work Time Due to Other Methamphetamine-Related Absences
Loss Low Estimate Best Estimate High Estimate
1 Average effect of meth use on days
absent because feeling blah in the past
30 days
0.92 1.42 1.93
2 Average effect of meth use on days
absent because of illness or injury in the
past 30 days
002.93
3 Average effect of meth use on days
absent per month (row 1 + row 2)
0.92 1.42 4.86
4 Total number of weeks per year
missed due to meth ([row 3 ×
12 months]/5 workdays per week)
2.21 3.41 4.0
a
5 Median weekly salary (from the BLS) ($) 201.00 580.80 651.00
6 Number of working meth users who
worked all year, ages 21–50 (from NSDUH)
250,284 350,253 450,221
7 Adjustment for percentage of working
meth users who use meth at least once a
week (36% of row 6)
90,102 126,091 162,080
8 Total cost of lost productivity caused by
other absenteeism ($)
40,024,209 249,726,756 422,056,320
a
Upper bound of four weeks was imposed due to the implausibility of a person retaining his or her job after
missing more than four weeks of work in a year without going on disability leave.
52 The Economic Cost of Methamphetamine Use in the United States, 2005
We again use the BLS measures of median weekly salary (row 5) to calculate the value of
this lost work time. The best estimate for the median weekly salary is constructed as a weighted
average of the full-time and part-time salaries, based on the fraction of the population work-ing each as reported in the 2005 NSDUH. We use population weighted estimates, and the
95-percent confidence interval surrounding these, for our best, low, and high (respectively)
estimates of the number of people between the ages of 21 and 50 employed and using meth in
the past year. However, not all of these working meth users report using meth at a frequency
that would cause them to miss two or more weeks of work. In fact, more than one-third of
the employed meth users (35.0 percent) report use at a rate of only once a month, and another
29 percent report using less than once a week, making it highly unlikely that their meth use
would cause them to miss two weeks of work per year (our low-end estimate). Therefore, in
row 7 of Table 5.5, we adjust these population estimates of employed meth users by the fraction
of that population that reports using meth at least once a week (36.0 percent).
Finally, we multiply the value of the lost work time (the product of rows 4 and 5) by the
number of meth users who were employed during the full year, as reported in the NSDUH.
Given these calculations, our best estimate of the lost productivity associated with sick and
blah days is about $249.7 million.
Lost Work Due to Incarceration
It is standard practice in previous studies of the cost of drug abuse to consider the value of
lost productivity associated with the incarceration of nonviolent drug-market participants (i.e.,
simple possession and sales offenders) (Harwood, Fountain, and Livermore, 1998; ONDCP,
2001). It is reasonable to argue that these costs are best considered outside of an evaluation of
the cost of drug use or abuse because these costs are a function of the policy response to the
problem (i.e., prohibition) rather than a direct consequence of meth use itself. Even if meth
use continues, the costs associated with lost productivity could change just by changing our
policy response to the meth problem. Thus, these costs truly are more a reflection of our policy
response than of drug use itself. Nonetheless, so that we can construct an estimate in a fashion
that is comparable to other studies on the cost of drug abuse, we consider these costs here.
We begin by trying to estimate the total number of meth-involved offenses that could
generate a prison or jail sentence and then cost out the expected sentence served. As discussed
in greater detail in Chapter Six, it is extremely difficult to determine the number of simple
meth-possession arrests that result in a prison or jail sentence. The vast majority of simple pos-session arrests, which are misdemeanor charges, result in probation or some other community
service. However, there are people in jail and prison with possession offenses. Some of these
may be the result of a plea bargain, however. Given the difficulty of trying to approximate the
fraction of possession arrests that end in incarceration, we focus our attention here on the lost
productivity associated with incarceration for meth-related sales. Unlike possession offenses,
there are very good data on the number of meth-related sales arrests from the BJS and the State
Court Processing Statistics. According to data from the BJS, there were 52,584 total meth-related sales arrests made in 2005 (row 1 of Table 5.6). Information on the disposition of these
sales arrests comes from the State Court Sentencing of Convicted Felons, 2004 Statistical
Tables (BJS, 2007), which show that about 73 percent of arrests end in conviction. Of those
convicted, 37 percent of those get sentenced to prison (row 3), while 38 percent are sentenced
to jail (row 8). Lower and upper bounds come from confidence intervals around these point
estimates.
Productivity Losses Due to Methamphetamine Use 53
Table 5.6
The Value of Lost Work Time Due to Methamphetamine-Related Incarceration
Loss Lower Bound Best Estimate Upper Bound
1 Total number of meth-related sales
arrests
52,584 52,584 52,584
2 Probability of conviction for sales offense 0.71 0.73 0.75
3 Probability of being sentenced to prison
for sales offense
0.35 0.37 0.39
4 Total number of meth-related sales
arrests resulting in prison term (row 1 ×
row 2 × row 3)
13,067 14,203 15,381
5 Average prison-sentence length served
for a sales offense (years)
1.74 1.74 1.74
6 Average annual income (based on
minimum wage) ($)
10,712 10,712 10,712
7 Total cost in lost productivity caused by
time spent in prison (row 4 × row 5 ×
row 6) ($)
243,554,245 264,728,013 286,684,613
8 Probability of being sent to jail for sales
offense
0.30 0.38 0.45
9 Total number of meth-related sales
arrests resulting in jail term (row 1 ×
row 2 × row 8)
11,200 14,587 17,747
10 Average jail-sentence length served for
sales offense (years)
0.26 0.26 0.26
11 Total cost in lost productivity by time
spent in jail (row 9 × row 10 × row 6) ($)
31,183,344 40,626,545 49,427,525
12 Total cost in lost productivity caused by
incarceration (prison and jail) ($)
274,747,589 305,354,558 336,112,138
Information on the average prison and jail sentence served (not given) comes from
analyses of the 2002 National Corrections Reporting Program, the most recent year of data
available. According to these data, the typical meth-sales offender served 1.74 years of his
or her prison sentence and 0.26 years of his or her jail sentence (thus we use the actual time
served as opposed to time given to calculate our estimate of lost productivity). To value the
opportunity cost of the time spent in prison and jail, we use information on the 2005 federal
minimum wage to construct an estimate of the annual salary for a full-time worker receiv-ing minimum wage. We use information on minimum wage instead of the median weekly
salary used in previous tables because it is well known that people who have been convicted
of a criminal offense have a harder time finding a job (and hence being employed) than
those who have not and generally are able to get only low-level jobs because of their criminal
record. In 2005, the federal minimum wage was $5.15, so assuming that a person could work
40 hours per week 52 weeks per year, the annual income associated with a minimum-wage
job would be $10,712. To generate our total loss in productivity due to prison (row 7), we
multiply this estimate of the value of lost employment time by the number of years served
(1.74) and the number of people convicted to prison. Similarly, to get our total loss in pro-ductivity due to jail (row 11), we multiply this average annual income by the number of years
54 The Economic Cost of Methamphetamine Use in the United States, 2005
served (0.26) and the number of people sentenced to jail. Our total cost of incarceration,
therefore, is given as the sum of lost productivity due to prison and jail (row 12). The total
cost estimate is $305.4 million with a range from $274.7 million to $336.1 million. The
majority of costs, as expected, are generated by lost productivity due to prison rather than
jail time.
In Table 5.6, there is not much variation in the lower- and upper-bound estimates of lost
productivity associated with incarceration. This is due to the fact that the only variation cap-tured in this calculation is in the number of meth-related sales arrests that result in a prison
term (i.e., conviction and sentencing variation).
Employer Costs of Hiring Methamphetamine Users
Employers may also incur costs trying to screen out meth users from employment. Although
screening costs comprise a relatively small fraction of the total burden, it is possible to approxi-mate these costs with existing data, as shown next.
Although workplace drug testing is becoming more common, accurate assessment of
the extent to which it occurs and the fraction that can be attributed to meth use specifically
is not really known. Quest Diagnostics, a national firm that specializes in diagnostic testing,
information, and services, reports that it conducted more than 7.3 million drug tests in 2005.
According to Quest Diagnostics (2006), more than 12 million employees in safety-sensitive
positions, such as truck drivers, airline pilots, mass-transit operators, and mariners, are sub-ject to mandatory drug testing under the U.S. Department of Transportation guidelines.
Data from the NSDUH, however, suggest that the number may be even larger, as 32.02 mil-lion employees work for companies that conduct random drug testing of current employees
(NSDUH, 2006). Nearly 50 million workers report that their employers require a drug test
prior to hiring. This suggests that the total number of drug tests conducted by and financed
by employers could easily exceed 30 million (assuming that companies do not test everyone
in the company every single year but that they do test potential new hires). But how much of
the cost of this testing can be attributable to meth specifically? Testing of a single urine sample
or hair specimen often involves the identification of multiple substances. For example, Quest
Diagnostic’s standard urine-specimen tests use a five- or nine-drug panel screen, checking for
the presence of a range of drugs or drug classes, including cocaine, marijuana, opiates, amphet-amines, and phencyclidine. The marginal cost of testing for meth in a urine sample, therefore,
is not the full cost of the test but rather the cost of the extra work required to identify the
meth in the specimen. According to information from SAMHSA on workplace drug testing,
additional effort is exerted to identify each chemical in a urine specimen even in a multisub-stance panel test (SAMHSA, undated). While no information is available on the marginal cost
of conducting these subtests, we can approximate the marginal cost using information on the
average cost of a drug test.
According to information available from the ONDCP related to student drug testing,
individual tests for a range of target drugs can cost anywhere from $10 to $50, depending on
the type of test used (e.g., urine, hair, fluids, sweat patch). If we assume that the lower price
would pay for the urine test including only a five-drug panel, then the cost of identifying each
of the five drugs would be $2. The $2 estimate is less than an average cost, however, because
it does not attribute any share of the employee’s or testing company’s time to meth directly. In
Productivity Losses Due to Methamphetamine Use 55
effect, the cost is marginal, because we assume that those costs are sunk because testing would
have been done in any case for other substances. Similarly, if we used the higher cost estimate
of $50 and presumed that this represented the cost of a nine-drug panel, we would get a mar-ginal cost of $5.56 per meth-specific test.
Table 5.7 reports the employer’s cost of drug testing under varying assumptions about the
cost approach for screening employees. We chose as a lower bound the number of employees
subject to mandatory drug testing under the Department of Transportation guidelines and, as
an upper bound, the number of people who report that their employer conducts random drug
testing of all employees. Our best estimate is simply given as the average of this upper- and
lower-bound number of employees tested. Using the conservative estimate of $2 per meth test
for both the lower-bound estimate and our best estimate, we get a lower and best estimate of
$24 million and $44 million, respectively. In the upper-bound estimate, we instead assume
that the cost of a drug test is $50 per test, reflecting a nine-drug panel, which suggests a mar-ginal cost per drug tested of $5.56. The higher cost per drug test in addition to higher number
of employees tested generates our much larger estimate for the upper bound of about $178 mil-lion paid by employers nationwide.
Limitations
There are a number of limitations of the calculations and assumptions made that should be
noted when examining the numbers just presented. First, our work developing an estimate of
the value of lost productivity associated with unemployment, absenteeism, and jail and prison
time assumes that median wages in the population (or minimum wage, in the case of the incar-ceration losses) are a good approximation of the wages of meth users. This is an assumption
that we had to make because there is little empirical evidence on the effect of meth on wages
directly and we had no way to estimate the effect ourselves. Thus, our estimates of the value of
time not spent working may, in fact, be biased. Another major limitation of our productivity
calculations is the lack of information on other employer costs for employing meth users. To
the extent that meth users make greater use of health care services or employee-assistance pro-gram (EAP) services that are partially paid for by employers, these meth users may impose a
greater cost on employers beyond those calculated thus far. Furthermore, we have not captured
the cost to employers of replacing meth employees who are fired because of their meth use. We
are unaware of any data from which we can attempt to bound these costs, however, so we are
left simply recognizing the fact that they are missing and should be considered in future work
if possible.
Table 5.7
Employer Costs of Drug Testing Attributable to Methamphetamine
Cost Lower Bound Best Estimate Upper Bound
Employees tested (millions) 12 22 32
Cost per test ($) 2 2 5.56
Total cost of testing ($ millions) 24 44
The Cost of Methamphetamine-Related Crime
There is a strong belief that meth use causes crime and imposes serious costs to the criminal
justice system and the rest of society. For example, in 2005, the National Association of Coun-ties released results from a survey of law-enforcement officials from 500 counties in 45 states
suggesting that meth-induced crime was increasing, and more than half reported that meth
was their county’s greatest drug problem. The media have also raised concerns that the expan-sion of meth use has increased crime (Butterfield, 2004; Suo, 2004) and possibly contributed
to the rise of new types of crime, such as identity theft (e.g., Sullivan, 2004).
This chapter focuses on the economic costs associated with what we refer to as meth-specific
offenses (possession, sales, and meth-related community corrections violations) and meth-induced offenses (property and violent crimes) that occurred in 2005. We review the scientific
literature on the meth-crime relationship, analyze criminal justice data from several sources,
and present a range of cost estimates that we believe to be conservative but credible.
1
The costs
we consider fall into three categories:
costs associated with new arrests for meth possession and sales t
costs associated with meth-specific parole and probation revocation t
costs associated with property and violent crimes caused by meth use. t
These components do not encompass all potential costs. We do not consider the effect of
meth use on every type of offense (e.g., vandalism) or community corrections violation, nor do
we consider all the costs associated with meth-related convictions (e.g., denial of some welfare
benefits, denial of student aid, removal from public housing) (GAO, 2005). Quantifying these
consequences is important for understanding the full costs of meth use, but data limitations
prevent their inclusion here. Even our focus on the three primary cost categories just listed
requires significant assumptions to generate estimates of the prevalence of events and their
costs.
Table 6.1 summarizes the cost categories considered as well as a range of possible cost
estimates. Our best estimate is that meth generated approximately $4.2 billion in crime and
criminal justice costs in 2005. The greatest share of costs is due to arrests for meth possession
and sales, at $2.4 billion.
1
We do not consider the costs associated with those who committed a meth-related offense before 2005 (e.g., someone
who was incarcerated for meth sales in 2004 and still serving the sentence in 2005). As noted in Chapter One, we treat
criminal justice costs in a manner consistent with the losses associated with premature death, by incorporating the full net
present value of future expenditures associated with an offense in 2005.
58 The Economic Cost of Methamphetamine Use in the United States, 2005
Table 6.1
Cost of Methamphetamine: Crime and Criminal Justice, 2005 ($ millions)
Cost Category Lower Bound Best Estimate Upper Bound
Meth-specific offenses
State and local possession offenses 332.2 737.5 2,400.7
State and local sales offenses 757.9 892.5 1,186.8
Federal possession and sales offenses 729.5 745.2 761.0
Meth-related parole and probation revocation for
technical violations
14.8 70.4 125.9
Other parole and probation violations NI NI NI
Meth-induced crime
Index crimes 743.6 1,764.2 11,266.5
Identitytheft NININI
Other crimes NI NI NI
Total 2,578.0 4,209.8 15,740.9
There is a tremendous amount of uncertainty surrounding these estimates, as evidenced
by the large difference between our low and high estimates (i.e., the range). It is noteworthy
that neither the best estimate nor the range includes consideration of meth-induced identity
theft. We discuss this issue in Chapter Nine.
Cost of Methamphetamine-Specific Arrests
Methamphetamine Offenses at the State and Local Levels
Possession and sale of meth is illegal in all 50 states and the District of Columbia, although the
statutory penalties vary considerably (ImpacTeen, 2002). Because federal law also prohibits the
possession and sale of meth, any meth offense can be adjudicated at the federal level; however,
federal law-enforcement agencies typically focus on trafficking and high-volume cases. This
means that state and local law-enforcement agencies handle the bulk of the possession and sales
cases in the United States.
Possession. We begin by calculating the number of arrests for meth at the state and
local levels. Because Federal Bureau of Investigation (FBI) statistics lump meth with other
dangerous, nonnarcotic drugs (e.g., other types of amphetamine, benzodiazepines), we apply
the ratio of meth to the total number of dangerous, nonnarcotic mentions from the treatment
population (SAMHSA, 2007c) to generate our meth-specific count.2
As shown in Table 6.2,
the estimated number of meth arrests in 2005 is then approximately a quarter of a million.
The lower and upper bounds for the cost of these arrests come from a drug-court cost study
(for one jurisdiction; NIJ, 2006) and an analysis of the marginal cost of a drug arrest in
2
Using TEDS 2005 that was last updated on April 25, 2008, and assuming that all amphetamine mentions for Oregon in
2005 are actually meth mentions (based on personal communication with treatment officials in Oregon), we put this figure
at 71.2 percent.
The Cost of Methamphetamine-Related Crime 59
Table 6.2
State and Local Possession Offenses, Misdemeanor Arrests
Cost Element Lower Bound Best Estimate Upper Bound
Cost per arrest, police ($) 477 1,085 1,693
Cost per arrest, court ($) 680 1,287 1,895
Probability of conviction 0.31 0.61 0.61
Average probation-sentence length (years) 2.10 2.10 3.17
Probation served (%) 50 75 100
Probation served (years) 1.05 1.58 3.17
Cost per year ($) 701 730 3,450
Net present value of probation sentence ($) 735 1,137 10,478
Number of arrests 240,572 240,572 240,572
Total cost ($) 332,234,780 737,549,339 2,400,713,222
SOURCES: FBI (2006); Carlson (undated); DCJS (2008); NIJ (2006); Fischer (2005); analyses of SAMHSA (2007c).
Washington state (Aos et al., 2004), respectively. Each study has its strengths (e.g., one figure
is more recent and was published in an NIJ report, and the other is a true marginal cost based
on multivariate regressions), and, for lack of better information, we simply take the midpoint
of these estimates for our best estimate.
Because there is no nationally representative database that will allow us to track convic-tion rates and sentences for misdemeanor offenses, we pull these estimates from a variety of
sources and make some simplifying assumptions.
3
We assume that all possession arrests are
charged as misdemeanors and that the only punishment is probation. Obviously, some offend-ers are incarcerated for possession (before or after trial), so this is a conservative assumption
that will underestimate the true cost of a possession arrest. The probability of conviction given
a possession arrest is set to equal the conviction rate published for all misdemeanors in the state
of New York in 2005 (61 percent; DCJS, 2008). This number could be high or low if drug
crimes have lower or higher rates of conviction than do other misdemeanor charges. Regardless
of whether it is high or low on that account, the number is likely an overestimate because it is
based on those cases that actually make it to court, thus excluding cases in which the charge
was not filed. That factor is difficult to quantify,4
so, for a low estimate, we assume that the
conviction rate for meth is 30.5 percent—half the all-misdemeanor, complaint-only rate for
New York. Sensitivity analyses reveal that using this more conservative figure reduces the total
crime costs by only about 3 percent.
The next step is to determine the sentence length. The lower and upper bounds for proba-tion sentences come from a meth-specific report from Wisconsin (Fischer, 2005) and the aver-age probation sentence for those convicted of felony sales in state courts (BJS, 2007), respec-3
There is tremendous variation in state and local policies with respect to what constitutes misdemeanor versus felony
possession.
4
For example, for California, we can identify the share of misdemeanor arrests for which law enforcement sought a com-plaint, but this does not mean that the prosecutor actually filed the complaint. Calculating this figure for misdemeanor
drug arrests is also complicated by court diversion programs and how they are handled in each substate jurisdiction.
60 The Economic Cost of Methamphetamine Use in the United States, 2005
tively. In the eyes of the law, selling is more harmful than possession, so this seems like a
credible upper bound. Because the Wisconsin figure is specific to meth, we use it as our best
estimate. It is also important to note that we cannot assume that the full probation sentence
will be served. Technical violations and new arrests are not uncommon for probationers, espe-cially those with substance use problems. For the cost estimates, our best guess assumes that,
on average, 75 percent of the sentence will be served.
5
For the annual cost of probation, we begin with a report from Minnesota suggesting that
annual probation costs for a meth offender were $701 (Carlson, undated). Probation costs are
likely to differ within and across states, but this figure is consistent with our belief that county
probation generally costs about $2 per day. Data from the Georgia Department of Corrections
(undated) suggest that “Standard probation supervision costs $1.43 per probationer per day.
Intensive or Specialized Probation Supervision costs $3.46 per day.” So for lack of a precise esti-mate or a breakdown for regular and intensive cases, our best estimate is that probation costs
are $2 per day, for an annual cost of $730. The upper bound is the annual cost for federal pro-bation for all offenses (U.S. Courts, 2006). This is an upper bound because one would expect
federal probation to be more resource intensive because it is dealing with more serious offend-ers. We use a discount rate of 4 percent to generate the net present value of the sentence.
This approach generates a best estimate of $737.5 million attributable to state and local
arrests for meth possession. The uncertainty in our factors is substantial, as is indicated by the
range of $332.2 million to $2,400.7 million.
Sales and Trafficking. Table 6.3 reports the costs associated with sales and trafficking
arrests made at the state and local levels. Many of the estimates and sources are the same as
those used for the possession calculation, so this section will highlight only those that differ.
We assume that all arrests for sales are felonies, but we do not assume that all subsequent con-victions are for felonies (e.g., the defendant could plea-bargain down to a misdemeanor posses-sion offense and serve a jail sentence or probation).
Unlike misdemeanor possession, there are good data available about felony sales arrests
in state and local courts. There are two major sources for information about felony conviction
rates and type of sentence: Felony Defendants in Large Urban Counties, 2004 (Kyckelhahn and
Cohen, 2008) and State Court Sentencing of Convicted Felons(Durose, 2007). Unfortunately,
neither of these sources is perfect for our purposes. First, neither source includes information
specific to meth—we must rely on general information about drug-trafficking convictions.
Second, the information from urban counties may not reflect the sentencing practices for
meth, a drug that is popular in rural areas. Third, the information for convicted felons does not
account for those who were charged with a felony but convicted of a misdemeanor. Since the
information from Durose (2007) generates the largest upper-bound costs, we use these figures
for the conviction rate and distribution of sentences conditional upon conviction (prison with
parole, jail, or probation) for our upper bound.
6
We use the information from Kyckelhahn and
Cohen (2008) for the lower bound and, for lack of better information, use the midpoint as our
best estimate.
5
Some offenses are more likely to be detected when someone is being supervised in the community (e.g., drug use via drug
testing). Since we do not include these costs in our calculations, we err on the side of being conservative.
6
This is attributable to the higher prison and parole rates and the fact that, in the upper bound, the net present value for
probation is larger than the net present value for jail.
The Cost of Methamphetamine-Related Crime 61
Table 6.3
State and Local Sales Offenses, Felony Arrests
Cost Element Lower Bound Best Estimate Upper Bound
Costs per arrest, police ($) 477.00 1,084.93 1,692.85
Costs per arrest, court ($) 680.00 1,287.25 1,894.50
Percent convicted
a
75 73 71
Probability sentenced to prison
a
0.35 0.37 0.39
Average prison-sentence length (years) 4.25 4.25 4.25
Prison sentence served (%) 41 41 41
Prison sentence served (years) 1.74 1.74 1.74
Cost per year, prison ($) 23,579.53 23,579.53 24,977.67
Net present value of prison sentence ($) 40,357.27 40,357.27 42,750.24
Average parole-sentence length (years) 2.17 2.33 2.50
Cost per year, parole ($) 1,062.16 3,031.08 5,000.00
Net present value of parole sentence ($) 2,250.41 6,870.37 12,119.08
Probability sentenced to jail
a
0.45 0.38 0.30
Average jail-sentence length (years) 0.50 0.50 0.50
Jail sentence served (%) 52 52 52
Jail sentence served (years) 0.26 0.26 0.26
Cost per year, jail ($) 21,845.66 21,845.66 24,977.67
Net present value of jail sentence ($) 5,679.87 5,679.87 6,494.19
Probability sentenced to probation
a
Average jail-sentence length (years) 0.50 0.50 0.50
Jail sentence served (%) 52 52 52
Jail sentence served (years) 0.26 0.26 0.26
Cost per year, jail ($) 21,845.66 21,845.66 24,977.67
Net present value of jail sentence ($) 5,679.87 5,679.87 6,494.19
Probability sentenced to probation
a
0.19 0.24 0.28
Average probation-sentence length (years) 3.17 3.17 3.70
Probation sentence served (%) 50 75 100
Probation sentence served (years) 1.58 2.38 3.70
Cost per year, probation ($) 700.00 730.00 3,450.00
Net present value of probation sentence ($) 1,090.38 1,688.39 12,103.95
Number of arrests 52,584 52,584 52,584
Total cost ($) 757,938,084 892,501,455 1,186,831,103
SOURCES: Aos et al. (2004); BJS (undated, 2004, 2006b, 2007); FBI (2006); Kyckelhahn and Cohen (2008); Carlson
(undated); NIJ (2006); DCJS (2008); analyses of SAMHSA (2007c).
a
We use figures from Durose (2007) for the upper-bound estimates of the conviction rate and distribution of
sentences conditional upon conviction (prison with parole, jail, or probation). We use the information from
Kyckelhahn and Cohen (2008) for the lower bound and, for lack of better information, use the midpoint as our
best estimate/
The length of prison, jail, and probation sentences imposed is based on the mean sen-tences for all felony drug offenses reported by BJS (2007; BJS, undated, Table 5.48.2004). The
one exception is the upper-bound estimate for the average probation sentence: This is specific
to meth sales in Wisconsin and is considered the upper bound simply because it is higher (3.7
years) than the average from the state-court statistical tables.
Actual time served for dangerous-drug convictions was calculated to be 41 percent and
52 percent for state prison and local jail sentences, respectively, from our analyses of 2002
National Corrections Reporting Program (NCRP) data.7
The lower-bound estimate of the cost
of incarceration (prison and jail, separately) is based on the marginal resource operating costs
for one year of supervision in a Washington State Department of Corrections institution (Aos
et al., 2004). The upper bound is based on the average annual operating cost per state inmate
published by BJS (2004). The lower bounds are calculated separately for jail and prison; the
upper bound is based only on prison and applied to both. The low and high estimates are very
close for prison, and we use the lower bound as the best estimate for jail and prison, since it
captures the marginal cost. Once again, we use a discount rate of 4 percent to generate the net
present value of the sentence.
Finally, we also consider that many of those released from prison will be subject to parole
supervision. If we assume that those leaving prison serve the rest of their time on parole, then
the average parole sentence would be nearly 30 months (4.25 years minus 1.74 years). This is
close to the 26-month average reported by the BJS for parolees released after spending at least
one year in state prison (BJS, 2004). We use these figures as the upper and lower bounds,
respectively, and use the midpoint (28 months) as the best estimate. For the lower-bound
costs, we use the annual rate from Georgia ($1,062) (Georgia State Board of Pardons and
Paroles, undated), and, for the upper bound, we use the annual rate reported by the Massachu-setts Executive Office of Public Safety and Security ($5,000) (EOPSS, undated). Taking the
midpoint as the best estimate seems reasonable, since these seem like reasonable bounds and
$3,000 is consistent with some of the other estimates we have uncovered.
8
The economic cost of parole includes more than government expenditures on supervision,
drug testing, and possibly programming. There is also the risk that a parolee will be caught for
a criminal offense, which would result in another round of adjudication and sanction costs to
the criminal justice system. While this monograph does explore the costs of parole and proba-tion violations directly related to meth (see “Cost of Community Corrections Revocations,”
later in this chapter), we do not consider the other costs.
The best estimate for the total cost of felony arrests for state and local sales offenses is
approximately $892.5 million, with a range from $757.9 million to $1,186.8 million.
Methamphetamine Offenses at the Federal Level
Table 6.4 reports the figures used to calculate the costs associated with federal drug arrests.
The number of federal meth-related arrests in 2005 is based on a Drug Enforcement
7
Based on NCRP offense code categories 350 and 390. Time served was calculated by dividing time served on current
admission by total prison sentence.
8
Parole cost estimates for other states appear to fall in this range (in 2005 dollars)—e.g., New York, $3,000 (Prisoner
Reentry Institute, undated), and Washington, $3,542 (Aos et al., 2004).
The Cost of Methamphetamine-Related Crime 63
Table 6.4
Federal Methamphetamine Offenses
Cost Element Lower Bound Best Estimate Upper Bound
Number arrested and booked 6,090 6,090 6,090
Cost per arrest ($) 477 1,693 1,693
Adjudication and trial costs per arrest ($) 680 1,895 1,895
Number incarcerated 5,035 5,035 5,035
Average prison-sentence length (years) 8.03 8.03 8.03
Average prison sentence served (%) 85 85 87
Average prison sentence served (years) 6.83 6.83 6.99
Cost per year, prison ($) 23,432 23,432 23,432
Net present value of prison sentence ($) 143,117 143,117 146,080
Number sentenced to probation 347 347 347
Average probation-sentence length (years) 3.17 3.17 3.17
Average probation sentence served (%) 50 75 100
Average probation sentence served (years) 1.58 2.38 3.17
Cost per year, probation ($) 3,450 3,450 3,450
Net present value of probation sentence ($) 5,374 7,979 10,478
Total ($) 729,505,478 745,210,370 760,996,105
SOURCES: Aos et al. (2004); BJS (2006a, 2006b); NDIC (2007a); Carlson (undated); USSC (2006); U.S. Courts (2006).
Administration (DEA)–sourced table published in the National Drug Intelligence Center’s
National Drug Threat Assessment 2008(NDIC, 2007a, Appendix C, Table 4).
The low and high estimates for arrest and court costs are the same as those used for the
state courts. However, the best estimate now reflects the high marginal cost estimate, because
we would expect, on average, that the federal cases would be more expensive: They often
involve more serious offenders. The cost of incarcerating someone in a federal prison in 2005
was calculated by the Bureau of Prisons and the Administrative Office of the U.S. Courts (U.S.
Courts, 2006).
The U.S. Sentencing Commission (USSC) publishes detailed information on the number
of individuals incarcerated in a federal prison for meth (these figures do not distinguish between
sales and possession offenses) as well as the average sentence length. The data for January 12,
2005, through September 30, 2005, suggest that 3,614 individuals were sent to prison for
meth for an average of 96.4 months each. Annualizing this total to estimate the full year
3 614 262 365 ,
§
©
·
¹
r generates a figure of 5,035 federal defendants sent to prison for meth in
2005. Truth-in-sentencing laws mandate that at least 85 percent of the sentence be served, and
Sabol and McGready (1999) estimate that an average of 87 percent of each sentence is served.
To estimate how many individuals were sentenced to probation, we take the number of
drug offenders sentenced to probation and divide it by the number of suspects arrested for drug
offenses in fiscal year (FY) 2004 1 879 32 980 0 057 ,, . §
©
·
¹
(BJS, 2006b). We then multiply
64 The Economic Cost of Methamphetamine Use in the United States, 2005
this rate by the number of individuals arrested for meth at the federal level in 2005 (6,090) to
generate an estimate of 347 individuals sentenced to probation.
This approach generates a best estimate of $745.2 million attributable to federal arrests for
meth. The range is fairly narrow, from $729.5 million to $761.0 million.
Cost of Community Corrections Revocations
It is not uncommon for a parolee or probationer to be incarcerated before completing his or
her term of community supervision. While some of these individuals are sentenced to prison
or jail for committing a new crime, others are incarcerated for violating a condition of their
supervision (e.g., do not use drugs, do not associate with gang members). A publication from
the BJS (2008b) reports the percentage of parolees who were reincarcerated because of a posi-tive drug test. Table 6.5 presents these results for the seven states listed in the report. The
shares range from 3 percent in Florida to 16 percent in South Dakota. It is important to note
that these re incarceration rates would likely be higher if all possible technical violations were
considered.
The percentages shown in Table 6.5 are probably smaller than the percentages failing
a drug test. Probationers and parolees usually do not have supervision revoked after the first
technical violation, especially if it is a positive drug test (see, e.g., Deschenes et al., 1996; Klei-man et al., 2003). Instead, revocation usually occurs after there have been several technical
violations or the individual commits a new crime. This section focuses on the costs associated
with revocation only for a technical violation that leads to imprisonment—we do not consider
the costs and consequences associated with other violations.
In 2005, almost 192,000 individuals were discharged from parole and returned to incar-ceration for a technical violation or a new offense (BJS, 2006b). Furthermore, more than
350,000 probationers, or 16 percent of the total,
9
left probation because they were incarcerated.
Table 6.5
Adult Parolees Returned to Prison Because of a Positive Drug Test
State Parolees (%)
Florida 2.9
Hawaii 9.7
Michigan 6.3
Pennsylvania 3.6
South Dakota 15.8
Utah 9.4
Wyoming 5.4
SOURCE: BJS (2008b).
NOTE: This figure was reported for only seven states.
9
Sometimes, probation is revoked but the individual is not incarcerated. This accounts for 13 percent of the probation
exits in 2005.
The Cost of Methamphetamine-Related Crime 65
To determine what share of these individuals were returned to prison or jail for a meth-related technical violation (e.g., failed a drug test or did not show up to drug treatment),
we turn to the BJS inmate surveys. The BJS conducts nationally representative surveys of
inmates within the state and federal correctional systems, as well as those in jails. The Survey
of Inmates in State Correctional Facilities (SISCF) and Survey of Inmates in Federal Cor-rectional Facilities (SIFCF) were most recently collected in 2004, and the Survey of Inmates
in Local Jails (SILJ) was collected in 2002.
10
(We assume no change in percentages returned
for a meth violation between the 2002 and 2004 surveys and the 2005 year of interest.)
Our analyses of the 2004 state and federal prison inmate surveys suggest that 253,736
and 324,088 of all prison inmates were on parole and probation, respectively, when they started
serving their time in prison. They could have been incarcerated for a new crime or a techni-cal violation. Our analysis of the 2002 jail inmate survey suggests that 24,247 and 107,566 of
these inmates were on parole and probation, respectively, when they started serving their time
in jail. To determine the share of these revocations that are attributable to meth-related techni-cal violations, we constructed slightly different criteria for our low and high estimates, based
on the information available in the inmate surveys.
low t
used meth regularly –
did not use other illegal drugs regularly –
was not incarcerated for a new offense –
had one of the following technical violations: –
failed drug test 0
drug possession 0
not taking drug test 0
failing to report to treatment 0
high t
used meth regularly –
was not incarcerated for a new offense –
had one of the following technical violations: –
failed drug test 0
drug possession 0
not taking drug test 0
failing to report to treatment 0
failing to report to counselor 0
failing to report to officer 0
leaving jurisdiction. 0
While the violations considered in the low definition are drug-specific, the additional vio-lations considered for the high definition could be drug related if the individual was trying to
avoid a drug test or the consequences of testing positive, but we cannot be certain of this. The
individual could be a regular meth user who absconded for another reason. For lack of better
information, we use the midpoint between these figures for our best estimate.
10
While the jail survey includes those who are being held for trial, we limited the sample to those who were currently con-victed offenders serving sentences in local jails or awaiting transfer to prison.
The typical sentence for parole revocation depends on the state. In California, the maxi-mum sentence for a technical violation is one year (CDCR, undated). In other states, parolees
can be sent back to prison to serve out the remainder of their term, which can exceed one year.
11
To generate a range, we use data from the 2002 National Corrections Reporting Program on
actual prison time served for those whose were incarcerated for a parole or probation viola-tion.
12
The 95-percent confidence intervals are 8.25–10.1 months and 4.01–5.46 months for
parole and probation violations, respectively. Data from California show that parolees returned
to prison for a drug-related technical violation serve, on average, four months in prison (Travis,
2003). Because California accounts for a large share of all prisoners nationally who are returned
to prison for a technical parole violation, we will consider four months to be the lower bound
for parole. We use the 95-percent confidence intervals for the best estimate and upper bound.
We do the same for the probation numbers, again adopting the California parole figure for the
lower bound. For lack of better information, we use these same sentences for parolees and pro-bationers sent to jail. We use the court costs discussed in the preceding section for processing
new arrests to approximate the cost of a revocation hearing (see Table 6.6).
Table 6.6
Cost of Methamphetamine-Related Parole and Probation Revocations
Parole and Probation Revocations Lower Bound Best Estimate Upper Bound
Parolees incarcerated before completing sentence 191,800 191,800 191,800
Share of these parolees returned to prison for meth-related technical violation
0.0030 0.0110 0.0190
Share of these parolees returned to jail for meth-related technical violation
0 0.0001 0.0003
Probationers incarcerated before completing sentence 353,552 353,552 353,552
Share of these probationers returned to prison for
meth-related technical violation
0.0027 0.0051 0.0075
Share of these probationers returned to jail for meth-related technical violation
0.0006 0.0009 0.0013
Parole served for technical violation (years) 0.33 0.76 0.84
Probation served for technical violation (years) 0.33 0.33 0.46
Cost per year, prison ($) 23,579.53 23,579.53 24,977.67
Cost per year, jail ($) 21,845.66 21,845.66 24,977.67
Cost per revocation hearing ($) 680.00 1,287.25 1,894.50
Total ($) 14,817,119 70,370,070 125,923,022
SOURCES: Aos et al. (2004); BJS (2006b); CDCR (undated); Travis (2003); author analyses of SISCF, SIFCF, and SILJ
(see Appendix D for more information on the calculations).
11
For example, the Delaware Board of Parole (2007) notes, “Offenders on parole or conditional release are under the juris-diction of the Board of Parole and may be returned by the Board to prison to serve the balance of their sentence if they fail
to abide by the conditions of supervision.”
12
Based on the item inquiring about the offense with the longest sentence.
The Cost of Methamphetamine-Related Crime 67
These costs are, by far, the smallest of our included categories. The best estimate of meth-related parole and probation violations totals $70.4 million. The upper and lower bounds are
$14.8 million and $125.9 million, respectively.
Cost of Methamphetamine-Induced Crime
The meth-specific crime costs just discussed reflect the control regime. In contrast with all the
other costs estimated in this monograph, the costs of the control regime would disappear if
the regime were removed. The cost of meth-specific crime is thus not a cost of meth use in the
same sense as all the others are. In this section, however, we discuss crime costs that are not
specific to the control regime and that are presumably kept in check by it.
Background: Methamphetamine’s Link to Property and Violent Crime
The academic literature supports an association between meth use and a variety of prop-erty and violent crimes. A survey of 655 meth users in Queensland found that a substantial
number had committed property and violent crimes; these users often cited meth use or the
need to pay for meth as the cause (Lynch et al., 2003). A substantial share of respondents also
reported that their meth use had caused them to be violent toward partners (35 percent), close
friends (29 percent), friends or acquaintances (33 percent), family members (27 percent), or
strangers (29 percent) at least once.
Similarly, Sommers, Baskin, and Baskin-Sommers’s (2006) survey of meth users found
that users were substantially involved in criminal behavior, including drug use and sales, non-violent (economic) crime, and violent crime. Cartier, Farabee, and Prendergast (2006) found
that 20 percent of parolees reported having used meth in the 30 days prior to the interview.
They were also significantly more likely than nonusers to have been returned to custody or
to self-report having committed a violent act (including robbery) in the previous 30 days.
However, there was no difference in the probability of being returned to prison for a violent
offense. Past-30-day use was significantly associated with self-reported violence regardless of
controls for involvement in drug trade. Logistic regressions suggest that use was associated with
a greater likelihood of return to prison in general (with and without controlling for drug-trade
involvement) but not for return due to a violent offense. Furthermore, our tabulations of posi-tive drug tests among arrestees in the Arrestee Drug Abuse Monitoring data show that meth
use is significantly higher among those arrested for property crimes and violent crimes relative
to the general population, with 13.53 percent of violent offenders and 21.68 percent of property
offenders reporting use of meth in the past year (IPCSR, 2004).
But correlation is not the same as causality, and the limited literature on this topic pro-vides little support for a causal link between meth use and crime. A forthcoming article by
Dobkin and Nicosia leverages exogenous variation from a significant supply shock for meth
precursors to examine this issue. The instrumental-variable regressions do not provide strong
evidence of a significant causal link between meth use (proxied by hospital admissions for
meth use) and property and violent crimes. There are two potential explanations. The first is
that there is not a causal relationship. The second is that the supply shock did not reduce meth
68 The Economic Cost of Methamphetamine Use in the United States, 2005
use among those users who commit property and violent crime.13
Unpublished dissertation
work by Rafert (2007) likewise finds no association with violent crime but a potential link to
property crime.
Another approach to understanding the relationship between meth use and other crimes
relies on offender self-reports. The BJS inmate survey data sets include information about the
type of crime committed, the inmate’s use of meth and other substances, whether the inmate
was under the influence of meth at the time of the offense, and whether the inmate was in need
of money for drugs at the time of the offense. However, there are several limitations to using
the inmate surveys to determine the effect of meth on crime:
Individuals incarcerated for a crime may be different from those who commit the same t
type of offense but are not incarcerated.
Recall and social-acceptability biases may reduce the accuracy of the self-reported t
information.
Being under the influence does not necessarily mean that the inmate would not have t
committed the offense if he or she were not using meth.
It is hard to establish meth causality if the inmate was a polysubstance user. t
That being said, these surveys represent the best available data for understanding the relation-ship between meth use and crime.
The goal is to estimate the number of crimes that are caused by meth and then generate
an estimate of their costs. We focus primarily on the seven index crimes from the FBI’s Uni-form Crime Report (UCR): murder and nonnegligent manslaughter, forcible rape, aggravated
assault, robbery, burglary, motor-vehicle theft, and larceny.
First, we estimate the attribution factor—that is, the percentage of inmates whose actions
can be attributed to their meth use. We conservatively define an offense as meth-induced if the
inmate meets the following criteria:
not a regular user of other illicit drugs 1.
not under the influence of other illicit drugs at time of offense 2.
not under the influence of alcohol at time of offense 3.
(regular meth user 4. andcommitted crime in order to obtain drugs or money for drugs)
or(under the influence of meth alone at time of offense).
We generate a best and low estimate of the attribution factors for each offense category
based on the mean and the lower bound of the 95-percent confidence interval associated with
the inmate survey data. For the high estimate, we drop criteria 1 through 3 and focus only
on cases that meet the fourth and use the upper bound of the