Journal et

10.1177/1043986204266892
Cowell of al.
Contemporary
/ CRIMINALCriminal
JUSTICE Justice
DIVERSION
/ August 2004
PROGRAMS

The Cost-Effectiveness of Criminal Justice

Diversion Programs for People With Serious
Mental Illness Co-Occurring With Substance Abuse
Four Case Studies

ALEXANDER J. COWELL
NAHAMA BRONER
RTI International
RANDOLPH DUPONT
University of Memphis

Many cities, counties, and states have criminal justice diversion or jail diversion programs, in
which those committing low-level offenses and who have mental illness or substance abuse are
diverted from the criminal justice system into treatment. However, there is little existing evi-
dence on the cost and cost-effectiveness of such programs. This article presents the first such esti-
mates for four sites. Estimates of the impact of diversion on both costs and effectiveness varied
across the sites. This variation likely reflects heterogeneity in the structure and implementation
of the programs across the sites. Directions for future research are suggested.

Keywords: cost-effectiveness; criminal justice diversion; substance abuse; mental health

M any policy makers agree that persons with mental illness or substance
abuse disorders who commit misdemeanors should be given access to
the necessary treatment. Over time, the criminal justice system has become
the provider of first resort for many people with these health conditions
(Steadman, Barbera, & Dennis , 1994; Steadman, Deane, et al., 1999; Teplin,
1990a, 1990b; Torrey et al., 1992). Depending on research methodology and
setting, mentally ill adults represent 7% to 18% of arrested, jailed, in-prison
or under-supervision populations in the United States (Broner, Lamon,
Maryl, & Karopkin, 2003; Ditton, 1999; Lamb & Weinberger, 1998). In
Journal of Contemporary Criminal Justice, Vol. 20 No. 3, August 2004 292-315
DOI: 10.1177/1043986204266892
© 2004 Sage Publications

292
 Cowell et al. / CRIMINAL JUSTICE DIVERSION PROGRAMS 293

response, many cities, counties, and states have developed criminal justice
diversion or jail diversion programs, in which those committing misdemean-
ors or felonies and who have mental illness and substance abuse problems are
diverted from the criminal justice system into treatment (Broner, Borum, &
Gawley, 2002; Broner, Ngyun, Swern & Goldfinger, 2003; Draine & Solo-
mon, 1999; Steadman, Cocozza, & Veysey, 1999). The potential advantages
of such programs to the participant include improved mental health function-
ing, reduced substance use, and reduced criminal justice recidivism. By shift-
ing the locus of care to the community, there may also be potential benefits
for overburdened courts and crowded jails and prisons. Despite the popular-
ity of such programs and the large resource reallocations necessary to con-
duct them, no study to date has evaluated their societal costs and assessed
how these costs vary by the effectiveness of the programs. Using data from
four sites in the United States, this article presents one of the first cost-
effectiveness analyses of jail diversion programs.

BACKGROUND
Compared to the general population, inmates in the United States have
high rates of serious mental illness and substance abuse (Teplin, 1990a,
1990b, 1994). Approximately 60% of adults at arrest test positive for drug
use (National Institute of Justice, 2000), and over half of those incarcerated
are believed to be drug dependent or to have alcohol problems (Wilson,
2000). Among those incarcerated and seriously mentally ill, at least 72%
have a co-occurring substance abuse or dependence disorders (Abram &
Teplin, 1991; Abram, Teplin, & McClelland, 2003) as compared to 25% in
community samples of those with mental health problems (Department of
Health and Human Services, 2002). With such high rates of mental illness
and substance abuse among detainees, the criminal justice system has
become burdened with meeting the resultant special needs (Borum,
Swanson, Swartz, & Hiday, 1997; Steadman, Deane, et al., 1999). To address
the community needs, just under half of all communities with more than

This study was funded by a grant from the U.S. Department of Health and Human Services, The
Substance Abuse and Mental Health Administration (SAMHSA), Grant Nos. U1GSM52192
and 1UD1 SM53273-01. The authors would like to thank the principal investigators and study
teams at each of the participating sites and the funding agency, SAMHSA. Considerable assis-
tance was provided by colleagues at the coordinating center, RTI International, in particular, Wil-
liam Schlenger and Pamela Lattimore, and research assistance by Andrew Stewart and Shuwen
Ng. The opinions expressed in this article are those of the authors and are not endorsed by the
funding agency. Findings from this article were presented in a different form to the American
Psychiatric Association Committee on Jails and Prisons, Arlington, Virginia, on February 13,
2004.
294 Journal of Contemporary Criminal Justice / August 2004

100,000 people can provide specialized responses for people with mental ill-
ness (Deane, Steadman, Borum, Veysey, & Morrissey, 1999), and criminal
justice diversion programs are increasing in popularity (Steadman, Deane,
et al., 1999).
There are two broad types of criminal justice diversion. Prebooking diver-
sion occurs at the point of contact with police officers before formal charges.
Most prebooking programs require specialized training for police officers
and a 24-hour crisis center that receives persons brought in by the police. The
crisis center is then the first point of care for the patient in the treatment sys-
tem (Steadman, Stainbrook, et al., 2001).
In contrast to prebooking programs, which keep an individual out of jail
altogether, postbooking diversion programs identify and divert offenders
after they have been booked, while they are either in jail or in arraignment
court. Diversion program staff work with stakeholders in the criminal justice
system—prosecutors, public defenders, and the courts—and with stake-
holders in the community treatment system to develop and implement a plan
of treatment that may or may not include court oversight and sanctions.
Postbooking is the most prevalent type of jail diversion program.
There is a small body of literature evaluating the effectiveness of both
types of programs (e.g., Borum, Deane, Steadman, & Morrissey, 1998;
Broner, Lattimore, Cowell, & Schlenger, in press; Cosden, Ellens, Schnell,
Yamini-Diouf, & Wolfe, 2003; Dupont & Cochran, 2000; Hoff, Baranosky,
Buchanan, Zonana, & Rosenheck, 1999; Lamb, Shaner, Elliott, DeCuir, &
Foltz, 1995; Lamb, Weinberger, & Reston-Parham, 1996; Steadman, Deane,
Borum, & Morrissey, 2000; Steadman, Cocozza, et al., 1999). To date, no
study has compared prebooking and postbooking programs, and only one
(Broner et al., in press) examines effectiveness across a comprehensive set of
outcomes for a number of sites. The literature provides mixed evidence that
diversion programs are successful at achieving their goals, including
improved police responsiveness, decreased injury rates for both officers and
consumers, reduced numbers of people with mental illness in jail, reduced
recidivism, increased connection with treatment, reduced substance abuse,
improved mental health functioning, and improved quality of life.
There is little published evidence on the costs of jail diversion, and no
study to date has examined the cost-effectiveness of jail diversion programs.
Considering the growing popularity of criminal justice diversion programs
as a means of both appropriately treating clients and lightening the burden on
jails and prisons, it is important to understand the cost implications of these
programs. To appropriately allocate scarce resources, policy makers need to
know how much jail diversion costs and how that cost compares to the alter-
natives. Establishing at what cost any improvements in outcomes can be
 Cowell et al. / CRIMINAL JUSTICE DIVERSION PROGRAMS 295

achieved by jail diversion would provide further insight. This article presents
the results of one of the first cost-effectiveness studies of jail diversion.

METHOD

Data

Samples

The results reported in this article are from a Substance Abuse and Mental
Health Administration (SAMHSA)-funded study of nine sites throughout
the United States. The eligible population for the study were people with co-
occurring serious mental illness—specifically, schizophrenia, depression,
and bipolar disorder—and substance abuse or dependence disorders who
were either eligible for diversion by police officers before booking or at a
later point in the criminal justice process. We estimated cost and effective-
ness of the jail diversion at four of the nine sites: Lane County, Oregon; Mem-
phis, Tennessee; New York City; and Tucson, Arizona. These programs, the
study population, and the larger study design are detailed in Broner et al. (in
press), Lattimore, Broner, Sherman, Frisman, and Shafer (2003), and in the
summary and individual site cost-effectiveness reports (Cowell, Stewart, &
Ng, 2002a, 2002b, 2002c, 2002d, 2002e). Three of the four programs—Lane
County, New York City, and Tucson—are postbooking programs; Memphis
is a prebooking program. All four programs had an 18-month intake window.
New York City alone limited intake of the diverted to a set number of avail-
able places, allocated on a first-come, first-served basis. The other three
programs placed no such additional limits on the study intake.

Cost Data

The perspective of the analysis determines the relevant cost domains and
how they should be measured and valued (Gold, Siegel, Russell, &
Weinstein, 1996). We took the taxpayer or community perspective, which
includes the costs incurred by all publicly funded agencies that were directly
involved with the jail diversion program. These agencies were categorized by
cost domains of two types, the criminal justice system and the health care sys-
tem. The estimated costs did not include costs directly incurred by study par-
ticipants. All costs included labor, materials, rent, utilities, maintenance and
supplies, capital costs, and administrative overhead.
The criminal justice system included the following cost domains: the
courts, the public defenders’ and prosecutors’ offices, the police, and the
296 Journal of Contemporary Criminal Justice / August 2004

jails. The court cost domain included costs associated with trying criminal
cases in which study participants were involved. These cases included multi-
ple court appearances as well as contacts with a public defender and
resources from the prosecutor. The jail cost domain included the cost of hous-
ing study participants in jail. The police cost domain included all costs asso-
ciated with study participants when they were arrested.
The health care system included the following cost domains: inpatient
mental health, residential substance abuse care, outpatient care (both sub-
stance abuse and mental health), emergency room (for substance abuse and
mental health visits), mental health assessment or evaluation, and case man-
agement. Costs of agencies in the health care system include all inpatient,
residential, and outpatient treatment received by study participants for men-
tal health and substance abuse in the community and in jail.
For each agency in each domain, we collected two types of cost data: (a)
utilization data, defined as the number of times a study participant came into
contact with the agency, and (b) unit cost data, defined as the cost of each con-
tact. Data were collected from the time at which each participant was
assessed for study eligibility to 1 year after the point of diversion. Because
assessment for study eligibility came before the point of diversion, the cost
data therefore include the costs of assessment for diversion. However, if staff
at the agencies had taken any special training for diversion activities (e.g.,
specially trained police officers in Memphis), the cost of that training was
largely excluded. Data were obtained using interviews with key stakeholders
and staff. For those agencies from which we could not collect this informa-
tion, we relied on available estimates in the literature and cost estimates from
other sites. Analyses results were examined for robustness to changing
assumptions such as the cost estimates used. Details on definitions and data
sources for the cost data are available from the author on request and can be
found in the summary report and individual site reports (Cowell et al., 2002a,
2002b, 2002c, 2002d, 2002e).

Effectiveness Data

Nine effectiveness outcomes were examined, including self-reported mea-

sures of criminal behavior, quality of life, substance use, and mental health
status. Criminal behavior was measured by whether the person was arrested
in the previous 30 days. This measure excluded the arrest for which the per-
son had been eligible for diversion. Measures of the effect of diversion on a
person’s quality of life included whether the respondent had been violently
and/or nonviolently victimized in the past 3 months, housing status stability,
and level of physical health. Physical health was measured by a 12-item
 Cowell et al. / CRIMINAL JUSTICE DIVERSION PROGRAMS 297

Social Function (SF-12) Physical Health Component score (PCS) (Ware,

Kosinski, & Keller, 1996). Measures of substance abuse included whether
the respondent abused alcohol at any time during the past 3 months and
whether the respondent used drugs during the past 3 months. Mental health
functioning was determined by scores on the Colorado Symptom Index
(CSI) (Shern et al., 1994) and the Mental Health Component score (MCS) of
the SF-12. The study team collected data on these measures for each person
in the sample at baseline, 3 months after baseline, and 12 months after base-
line. Broner et al. (in press) describe the main study’s procedure and
effectiveness measures in greater depth.

Estimation

Cost Estimation

For each participant in the study, the cost per unit of service and the utiliza-
tion of service were combined to obtain the total cost over all services during
the 1-year study period. To ascertain whether those diverted from jail
incurred differences in cost on average, the average cost incurred by the
diverted was compared to the average cost incurred by the not diverted. To
control for observable confounders, we conducted multivariate regression
analysis that controlled for a number of individual-level characteristics. For
person i we estimated the following specification:

costi = β0 + β1 × interventioni + ␦ × Xi + ui, (1)

where cost is the cost across all domains for person i over the 1-year study
period and intervention is an indicator variable for whether a person was
diverted. The estimated coefficient on this variable, β1, was interpreted as the
effect of jail diversion on costs, holding constant other observed characteris-
tics. The characteristics that were controlled for in the vector X (with a corre-
sponding vector of coefficients, ␦) included age, gender, race or ethnicity,
whether the individual was mentally disturbed at baseline, whether the
respondent was ever arrested as a juvenile, number of past arrests, the sever-
ity of alcohol and drug use (Michigan Alcohol Scoring Test [MAST])
(Storgaard, Nielsen, & Gluud, 1994) and Drug Abuse Scoring Test [DAST])
(Skinner, 1982), time at risk (measuring the number of days the participant is
in the community at risk of committing an offense), CSI, MCS, and PCS. The
CSI, PCS, and MCS were included because they conveniently summarize
key aspects of mental health, behavior, and physical health. The error term
is ui.
298 Journal of Contemporary Criminal Justice / August 2004

Costs were estimated separately for each site. In addition to estimating

Equation 1 for costs over all domains, we also estimated Equation 1 for each
domain separately. Because prebaseline cost data were not available at all
four sites and because the same estimation methodology was used across all
four sites so that comparisons could be made across sites, we did not adjust
for cost differences between the diverted and nondiverted at baseline. Ordi-
nary least squares estimation using Stata software was conducted for the
analyses.

Effectiveness Estimation

For each of the nine outcomes in each of the four sites, we assessed
whether diversion was associated with improvements in the outcome. As
noted for costs, a comparison of unadjusted average outcomes between the
diverted and the nondiverted would fail to account for a number of confound-
ing variables. Unlike the cost data, the outcome data at all four sites contain
outcome measures both prebaseline and for the two follow-up periods after
the point of diversion. Therefore, we were able to control for potential con-
founders that were unobservable in addition to those confounders that were
observable. For each outcome, a regression of the following functional form
was conducted (using the CSI score as an example):

CSIit = γ0 + γ1 × groupi + γ2 × followup3it + γ3 × followup12it +

(2)
γ4 × intervention3it + γ5 × intervention12it + ␮ × ⌮it + eit.

In Equation 2, group equals 1 if person i is in the diverted group, and group

equals 0 if the person is not in the diverted group. Including group in the
equation controls for selection bias due to unobserved variables. If people
were randomly assigned into groups, then the associated coefficient, γ1,
equals 0. The variable followup3 equals 1 if the observation at time t is at the
3-month follow-up and 0 otherwise. Followup12 is similarly constructed for
the 12-month follow-up. Including the followup variables controls for matu-
ration bias. For example, if average CSI was not improving between baseline
and 3 months regardless of diversion, then γ2 equals 0.
The most important variables in Equation 2 are denoted intervention3 and
intervention12. These terms are the interaction between the follow-up vari-
ables and group and capture the effect of the jail diversion program itself at
the 3- and 12-month follow-up points. Thus, the diverted simultaneously
have two comparison groups, the nondiverted and themselves prebaseline.
Intervention3 equals 1 if the person is both in the diverted group and
observed at the 3-month follow-up; otherwise, it equals 0. The corresponding
 Cowell et al. / CRIMINAL JUSTICE DIVERSION PROGRAMS 299

coefficient, γ4, measures the degree to which jail diversion is associated with
CSI as measured at 3 months. Similar reasoning applies to intervention12
and its associated coefficient, γ5.
Z contains those covariates also included in the cost regressions. For the
outcome measuring whether the person had been arrested in the previous 30
days, age of first arrest is also included as an additional variable because that
would directly affect the probability, frequency, and severity of future
offenses. Mu is the corresponding set of coefficients for Z. SAS and Stata
software were used for the analyses. Most models were estimated in a gener-
alized estimating equations framework with an unstructured correlation
structure, which allows for unequally spaced observations within each panel.

Cost-Effectiveness Analysis

Cost-effectiveness analysis combines the cost associated with jail diver-

sion with each effectiveness measure. The main tool of cost-effectiveness
analysis is the cost-effectiveness ratio, which expresses the additional cost
associated with a 1-unit change in effectiveness for an intervention. The ratio
is the effect of jail diversion on cost divided by the effect of jail diversion on
effectiveness.
Following Gold et al. (1996), a cost-effectiveness ratio was estimated only
if either diversion or its alternative was not dominated; one alternative is said
to dominate the other if it is both more effective and less expensive. When-
ever one alternative dominates another, the cost-effectiveness ratio becomes
irrelevant to the analysis and it need not be calculated. Because the cost-
effectiveness ratio is a nonlinear combination of two random variables, con-
fidence intervals errors were obtained using nonparametric bootstrap
techniques.

RESULTS
Rather than combine the data from the four sites, results are presented sep-
arately for each site. Sites were not combined, because the sites differ mark-
edly in the characteristics of both the diversion program and the population
served (Lattimore et al., 2003).

Descriptive Statistics

Table 1 presents for each site unadjusted baseline means for effectiveness
measures. The mean proportion of participants who have been arrested var-
ied considerably from 4% of the not-diverted in New York City to 23% of the
not-diverted in Lane County. Both violent and nonviolent self-reported vic-
300
TABLE 1
Unadjusted Means of Effectiveness Measures at Baseline (standard deviation in parenthesis)
Lane County, OR (n = 185) Memphis, TN (n = 609) New York, NY (n = 231) Tucson, AZ (n = 90)
Effectiveness Measure D (n = 93) ND (n = 92) D (n = 301) ND (n = 308) D (n = 111) ND (n = 120) D (n = 65) ND (n = 35)

Criminal behavior
Whether arrested in the past 30 daysa 0.17 0.23 0.081*** 0.16*** 0.07 0.04 0.17 0.08
(0.38) (0.42) (0.27) (0.37) (0.26) (0.20) (0.38) (0.28)
Quality of life
Whether seriously victimized in the past 3 monthsa, b 0.39 0.41 0.38 0.41 0.44 0.42 0.52 0.36
(0.49) (0.50) (0.49) (0.49) (0.50) (0.50) (0.50) (0.49)
Whether nonviolently victimized in the past 3 monthsa, c 0.33 0.33 0.31 0.34 0.42* 0.32* 0.35 0.44
(0.47) (0.47) (0.46) (0.47) (0.50) (0.47) (0.48) (0.51)
Whether homeless in the past 30 daysa, d 0.28 0.25 0.20 (0.40) 0.23 0.33 0.29 0.15
(0.45) (0.44) (0.42) (0.47) (0.46) (0.36) (0.37) 0.16
Physical Health Component scoree, f 44.67 47.64 46.02 44.71 50.47 49.87 48.32 44.00
(12.76) (13.44) (11.55) (11.66) (10.60) (10.76) (11.68) (11.43)
Substance use
Whether abused alcohol in the past 3 monthsa, g 0.47 0.40 0.37*** 0.56*** 0.54 0.57 0.65 0.52
(0.50) (0.49) (0.48) (0.50) (0.50) (0.50) (0.48) (0.51)
Whether used drugs in the past 3 monthsa, h 0.78 0.71 0.45*** 0.62*** 0.80 0.78 0.69 0.60
(0.41) (0.46) (0.50) (0.49) (0.40) (0.41) (0.47) (0.50)
Mental health status
Colorado Symptom Index scoree 44.86* 41.87* 46.94*** 41.91*** 46.04 44.09 46.06 45.44
(12.92) (10.31) (15.38) (14.36) (11.10) (11.87) (11.12) (13.37)
 Mental Health Component scoree, f 34.29** 30.36** 39.01*** 33.42*** 34.66 33.04 33.78 33.37
(12.53) (11.03) (12.25) (12.46) (12.69) (12.60) (12.29) (12.24)

NOTE: D = diverted; ND = not diverted.

a. Binary response.
b. Serious victimization includes assault, rape, mugging, or robbery.
c. Nonviolent victimization includes burglary, theft of property, and being cheated.
d. Being homeless means not having a regular place to stay at least half the time and whenever the person pleases.
e. The higher the score, the more functional the individual, or the healthier the behavior.
f. Taken from the 12 item Social Function (SF-12) survey instrument.
g. Alcohol abuse is if the person used more than four drinks for females or more than five drinks for males at any one sitting.
h. If used any marijuana/hashish, cocaine/crack, sedatives, stimulants, opiates, psychedelics, or inhalants.
i. Number of days at risk of committing a new crime or offense in the community in the past 3 months.
*Significant at the 10% level. **Significant at the 5% level. ***Significant at the 1% level.

301
302 Journal of Contemporary Criminal Justice / August 2004

timization was prevalent among the sample, ranging between approximately

30% and 50% of the diverted and not-diverted groups across the sites. There
was also a high percentage of homelessness among the sample, ranging from
15% for the diverted in Tucson to 33% for the diverted in New York City.
Because the population is substance using, the self-reported drug and alcohol
use rates are high. Typically, more than half of those across sites abused alco-
hol in the previous 3 months and between 50% and 70% abused illegal drugs.
Demographic and other personal characteristics at baseline are summa-
rized as follows.1 Respondents were typically in their mid- to late 30s. Most
groups had an average of 50% to 60% men with the exception of the not-
diverted sample in Memphis, which is 88% men. The majority of participants
in Lane County and Tucson were White (94% and 65%, respectively), the
majority of participants in Memphis and New York City were African Amer-
ican (72% and 62%, respectively), and both New York City and Tucson had
sizeable proportions of participants who were Hispanic (23% and 21%,
respectively). Measures of criminal justice history varied across sites and by
diversion history. For example, 52% of the diverted sample in Tucson were
ever arrested as a juvenile, a significantly higher proportion than the 20% of
the not-diverted in that city. In contrast, in Memphis 31% of the diverted were
ever arrested as a juvenile, a significantly lower proportion than the 50% of
the not-diverted ever arrested as a juvenile. As would be expected among a
population with serious mental illness, the proportion that experienced psy-
chosis at the time of initial arrest was high, ranging from 24% of the non-
diverted for the Lane County study to 50% of the diverted in Memphis.
Table 2 presents for each site unadjusted mean costs for each domain. For
all sites the bulk of criminal justice costs consisted of jail costs, and for all
sites the average cost of staying in jail was higher for the not-diverted than the
diverted. For example, in Lane County the average jail cost incurred by the
not-diverted was $9,925, whereas the average jail cost incurred by the
diverted was $5,886. In contrast, health care treatment costs were typically
higher for the diverted. In Memphis, for example, the average total health
care treatment costs incurred by the diverted were $9,013, compared to the
$2,192 incurred by the not-diverted. New York City was the exception of the
four sites, where the not-diverted group’s treatment costs were $9,185, which
is $107 greater than the average of the diverted group’s treatment costs
($9,078). Table 2 also shows that the standard deviation of the major cost
domains is very large. For example, the mean of jail costs for the diverted in
Tucson was $1,535 and the standard deviation was $4,047. This feature of the
data is perhaps to be expected, given the sample have serious mental illness
and substance abuse problems, and thus, many had high rates of use of
services.
 TABLE 2
Unadjusted Means of Cost Domains (in dollars, standard deviation in parenthesis)
Lane County, OR (n = 129) Memphis, TN (n = 609) New York, NY (n = 231) Tucson, AZ (n = 90)
Cost Domain D (n = 93) ND (n = 36) D (n = 301) ND (n = 308) D (n = 111) ND (n = 120) D (n = 65) ND (n = 35)

Total criminal justice costs 8,779** 11,595** 665** 1,682** 4,288*** 9,807*** 2,090 3,904
(4,357) (6,501) (1,682) (2,019) (7,944) (13,897) (4,103) (4,233)
Jail 5,886** 9,925** 373*** 863*** 2,527*** 8,673*** 1,535 3,100
(8,260) (8,500) (1,559) (2,115) (8,533) (14,634) (4,047) (4,077)
Other criminal justice costs 2,892 1,669 292** 819** 1,516*** 622*** 555 803
(3,477) (2,080) (453) (1,020) (345) (907) (58) (23)
Total treatment costs 7,385 4,148 8,075*** 2,003*** 9,078 9,185 9,886 7,215
(9,450) (15,827) (14,791) (8,059) (15,975) (12,197) (14,457) (9,841)
Inpatient and outpatient substance abuse and N/A N/A N/A N/A N/A N/A 9,837 7,193
mental health services (14,417) (9,847)
Inpatient N/A N/A 7,663*** 1,826*** 2,475 1,299 N/A N/A
(14,541) (8,026) (13,636) (7,278)
Jail residential N/A N/A N/A N/A 4,720 5,199 N/A N/A
(8,352) (8,137)
Other treatment costs N/A N/A 412 177 1,883 2,687 N/A N/A
(394) (125) (2,033) (4,230)
Total costs 16,164 15,743 8,740*** 3,685*** 13,366** 18,480** 11,976 11,119
(13,245) (17,498) (14,911) (8,352) (17,114) (17,629) (15,048) (2,155)

NOTE: D = diverted; ND = not diverted; N/A = Not available.

**Significant at the 5% level. ***Significant at the 1% level.

303
304 Journal of Contemporary Criminal Justice / August 2004

For two sites, Memphis and New York City, we were able to disaggregate
health care costs into their component domains. For Memphis, the majority
of health care costs incurred were inpatient mental health costs. For New
York City, in addition to inpatient mental health treatment, health care costs
were driven by residential substance abuse treatment both in and outside the
jail.

Cost Results

For each site, regression analysis was used to compare the estimated cost of
the diverted to those not diverted. Because Lane County suffered from
nonrandom attrition in the comparison county, Heckman (1979) selection
correction methods were used to estimate the effect of diversion in Lane
County.2 The nonrandom attrition affected collateral utilization data only,
and not self-reported data. Because the cost estimates use collateral data, the
number of observations in the comparison county reported in Table 2 (36) is
below that reported in the baseline means (92).
Table 3 presents the results of the models for the main cost domains as well
as all domains combined. The estimates presented are the additional costs per
domain (or over all domains) during the 12-month study period that would be
incurred if someone were diverted from jail, holding constant other observed
characteristics of the person.
The results confirm the finding from the unadjusted means presented above
that diversion was associated with lower criminal justice costs. The reduction
in jail costs associated with diversion was very large in the three postbooking
diversion sites. In Lane County, for example, a person diverted from jail
incurred $3,775 (p = .05) lower jail costs over the 12 months of study; in New
York City, the difference was even greater (coefficient = –7,038, p < .01). In
the prebooking site, Memphis, the estimated difference in jail costs, $384
lower costs for the diverted, was not statistically significant.
Turning to health care costs, the association between diversion and health
care costs varied greatly across the four sites. Whereas the overall treatment
costs were higher for diversion in the prebooking site, Memphis, no statisti-
cally significant difference was found for the three postbooking sites. A
diverted person in Memphis incurred $6,577 higher treatment costs during
the 12-month study period. In Lane County, the incremental treatment cost of
being diverted was positive but comparatively small, and far from statisti-
cally significant (coefficient = 940.76, p > .1). In Tucson, because of the
small sample size the difference was not statistically significant (coefficient
= 2,395.663, p > .1), despite the positive incremental cost being large. In New
York City, the incremental health care cost was negative, small, and only mar-
ginally significant (coefficient = –316.3, p < .1).
 TABLE 3
Adjusted Incremental Costs of Jail Diversion (in dollars, standard error in parenthesis)
Cost Domain Lane County, OR (n = 187) Memphis, TN (n = 609) New York, NY (n = 231) Tucson, AZ (n = 90)

Jail costs –3,770* –318* –7,038*** –1,735

(1,956) (184) (1,819) (1,071)

Total treatment costs 941 6,557*** –316 2,430

(2,798) (1,141) (2,052) (3,412)

Total costs –1,796 –5,855*** –6,260** 447

(3,492) (1,158) (2,594) (3,551)

*Significant at the 10% level. **Significant at the 5% level. ***Significant at the 1% level.

305
306 Journal of Contemporary Criminal Justice / August 2004

Because Memphis had particularly detailed treatment cost data, we con-

ducted two additional analyses of the treatment costs in Memphis. For the
first we examined outliers for each modality of treatment. Perhaps surpris-
ingly, the results suggested that outliers did not disproportionately influence
the results. For the second additional analysis we examined the profile of the
effect of diversion on inpatient costs from month to month. We found large
differences in inpatient costs in the first month immediately following diver-
sion, but then the difference diminished and leveled off for the remainder of
the study period.
After combining costs over all domains, there was considerable variation
across sites in the association between diversion and costs over all domains
(or total cost). In Memphis, the greater inpatient mental health care costs
exceeded the jail costs associated with diversion, resulting in diversion being
associated with an increase in costs over all domains of $6,576 (p < .01). The
New York City jail diversion program was associated with total costs being
$6,260 lower (p < .05). Neither the greater total costs in Tucson (coefficient =
857, p > .1) nor the lower total costs in Lane County (coefficient = –1,796, p >
.1) associated with diversion was statistically significant.

Effectiveness Results

Table 4 presents the effectiveness results. The estimates presented in the

table are robust to alternative correlation structures. Odds ratios (ORs) are
presented for binary outcomes and coefficient estimates are presented for
continuous outcomes. The effect of diversion associated with the outcomes
was estimated relative to baseline both at the 3-month and 12-month follow-
up points.
The results indicate that there is considerable variation between sites as to
which effectiveness measures were significantly associated with diversion.
In Lane County, diversion was associated with a large but marginally signifi-
cant increase in the odds of being arrested at 3 months (OR = 3.24, p < .1) and
a large but marginally significant increase in the odds of being nonviolently
victimized at 3 months (OR = 3.81, p < .1). However, at the 12-month follow-
up, diversion was also associated with a reduction in the odds of drug use in
the previous 3 months of 80% (OR = 0.20, p < .05). There was no association
between drug use and diversion at 3 months.
In New York City, diversion was associated with significant reductions in
nonviolent victimization (OR = 0.266, p < .05) at 3 months. The association
between diversion and reductions in serious victimization approached statis-
tical significance (OR = 0.353, p < .1).
Diversion in both Memphis and Tucson was associated with improvements
in psychiatric symptoms as measured by the CSI. In Memphis, diversion is
 TABLE 4
Effectiveness of Jail Diversion (standard error in parenthesis)
Lane County, OR (n = 187) Memphis, TN (n = 609) New York, NY (n = 231) Tucson, AZ (n = 90)
3-Month 12-Month 3-Month 12-Month 3-Month 12-Month 3-Month 12-Month
Effectiveness Measure Effect Effect Effect Effect Effect Effect Effect Effect

Criminal behavior
Whether arrested in the past 30 daysa 3.24* 1.28 1.03 1.42 1.24 0.54 0.41 0.25
(1.94) (2.25) (1.48) (1.60) (2.88) (3.02) (3.66) (3.39)
Quality of life
Whether seriously victimized in the past 3 monthsa 2.21 0.84 1.01 1.14 0.37* 1.13 0.65 0.51
(1.84) (2.09) (1.34) (1.38) (1.73) (2.00) (2.42) (2.31)
Whether nonviolently victimized in the past 3 monthsa 3.81* 3.87 0.71 0.55 0.27** 0.36 5.01* 0.83
(2.17) (3.36) (1.40) (1.50) (1.90) (2.26) (0.98) (0.73)
Whether currently homelessa N/A 0.63 N/A 0.75 N/A 1.25 N/A 1.95
(2.02) (1.53) (1.93) (2.07)
Physical Health Component score 1.56 0.92 –0.50 0.21 3.26* –0.75 1.44 –0.40
(0.43) (0.70) (1.15) (1.33) (0.09) (0.81) (2.77) (3.19)
Substance use
Whether abused alcohol in the past 3 monthsa 1.22 0.35 0.83 0.96 3.89 0.86 1.67 0.39
(1.74) (2.06) (1.28) (1.33) (3.48) (2.08) (2.06) (1.96)
Whether used drugs in the past 3 monthsa 0.81 0.21** 0.96 1.11 0.42 0.59 1.29 0.84
(1.64) (1.90) (1.33) (1.36) (0.31) (0.41) (2.14) (2.11)
Mental health status
Colorado Symptom Inventory score 1.05 –0.80 2.39* 2.14 –0.46 –1.72 6.01* 5.32*
(0.58) (0.75) (1.34) (1.56) (0.78) (0.54) (3.48) (3.26)
Mental Health Component score 1.25 –0.53 1.20 0.45 0.59 –0.09 –0.34 0.04
(0.54) (0.86) (1.33) (1.39) (0.77) (0.98) (3.65) (3.25)

NOTE: N/A = Not available.

a. Odds ratio reported.

307
*Significant at the 10% level. **Significant at the 5% level.
308 Journal of Contemporary Criminal Justice / August 2004

associated with an average increase in CSI of 2.4 points at 3 months (coeffi-

cient = 2.4, p < .05). The effect of diversion at 12 months is of a similar mag-
nitude but is not statistically significant (coefficient = 2.2, p > .1). These
results indicate a significant improvement in psychiatric symptoms 3 months
after being diverted from jail. In Tucson, the 3-month improvement in CSI
associated with diversion was 5.5. However, the estimate only approached
significance (p < .1). The estimated 12-month effect for CSI was also large,
but was larger than standard levels of significance (coefficient = 4.8, p > .1). It
should be noted that the CSI results reported here control for a large number
of observed factors. The raw score differences are even greater, often exceed-
ing 6 to 8 points on the CSI. Raw score differences as low as 3 points have
been associated with major differences in psychiatric functioning, such as the
use of crisis services within the past month (Lee, Stern, Coen, Bartsch, &
Wilson, 2002).

Cost-Effectiveness Results

For each effectiveness outcome at each of the two follow-up periods in

each site, we assessed whether one alternative (jail diversion or its alterna-
tive) dominated the other and whether both the effectiveness and cost esti-
mates were statistically significant. These assessments determined whether
we estimated a cost-effectiveness ratio (CER). In Memphis alone there was a
detectable improvement in an effectiveness measure associated with diver-
sion, as well as an increase in costs incurred.
Before estimating the CER, we reestimated incremental costs to be consis-
tent with the 3-month period used for the effectiveness measure. The result
was that incremental costs associated with jail diversion in Memphis were
$2,955 (p < .01) over the first 3 months of the study. This estimate was the
numerator in the CER; the effectiveness estimate was the denominator.
The estimate of the CER indicated that for every 1-point improvement in
the CSI, an additional $1,236 of resources was used by the jail diversion pro-
gram. The lower bound of the bias-corrected 95% confidence interval was
$492.41 and the upper bound was $17,727.97. These estimates suggest that
although the variation in the estimate was large, the estimate is statistically
significant at standard levels.

DISCUSSION AND CONCLUSIONS

The result that diversion was associated with lower jail costs is intuitively
appealing, because the core defining feature of any criminal justice diversion
program is to keep those diverted out of jail. This is perhaps seen most clearly
in the results from the New York City site, where diversion results in $7,038
 Cowell et al. / CRIMINAL JUSTICE DIVERSION PROGRAMS 309

lower jail costs on average (p < .01). In New York City, it is this difference in
jail costs that drives the finding that jail diversion resulted in $6,260 lower
costs over all domains (p < .05).
The treatment domain greatly influenced cost differences. Although in
Memphis diversion was associated with significant criminal justice cost sav-
ings, the amount and unit cost of treatment drove the finding that jail diver-
sion was associated with higher costs. Total costs over all domains were
$6,575 (p < .01) higher on average if someone was diverted. The main factor
in that difference is that diversion leads to $6,335 higher inpatient mental
health costs over the 12-month study period.
The finding of an association between diversion and a 2.4-point improve-
ment in the CSI as measured at 3 months in Memphis is perhaps remarkable.
This finding represents a short-term improvement in a population that has a
high incidence of psychosis and that typically comprises low-functioning
individuals (Lee et al., 2002). Further analysis of the outcome data in Mem-
phis (Dupont, 2002) also suggests that in contrast to the cost results, the
effectiveness results are not entirely driven by access to inpatient care. The
differences in the CSI may be attributable to factors related to diversion itself,
including intervention by a specially trained police team, a specialized
receiving facility, and avoidance of the negative experience of being in jail.
Thus, the availability of alternative, less costly treatments may be critical in
controlling the costs of prebooking jail diversion.
The four sites discussed here vary considerably in the way that diversion
was implemented and in the amount and timing of treatment received by both
those diverted and those not diverted. Perhaps the most important difference
was between prebooking and postbooking. Prebooking programs target indi-
viduals based on behaviors indicating that potential criminal justice events or
potential mental health events may follow. Thus, a person who appears to a
police officer to be decompensating would likely be diverted in a prebooking
setting. The mental health system would then solely determine the appropri-
ate following steps for that person. In contrast, postbooking programs target
people based on their actual diagnosis and once a criminal justice event has
actually occurred. The act of diversion thus requires the cooperation of both
the mental health system and the criminal justice system.
The broad differences in program characteristics imply that prebooking
programs typically serve more people, because diversion is based on a broad
spectrum of behaviors. Postbooking programs tend to serve fewer people but
typically marshal resources to the specific needs of their clients. The implica-
tion for differences in costs is seen perhaps most clearly early on in the diver-
sion process. In the first few months of diversion, prebooking programs, such
as Memphis, typically provide more treatment in the community, which is
often expensive compared to the jail treatment received. Postbooking pro-
310 Journal of Contemporary Criminal Justice / August 2004

grams, such as New York City, provide treatment services to both the diverted
and the nondiverted in jail.
For certain outcomes and in certain sites, the lack of significance across
sites is also notable. A low number of observations certainly mitigated against
finding any significant findings in Tucson. However, in Lane County, New
York City, and Memphis, the number of observations was high enough to
detect differences at either follow-up. Thus, the lack of any significant
increase in rearrest, alcohol use, or drug use can be interpreted as encouraging.
This study faced potential limitations in the cost analysis, the effectiveness
analysis, and the overall design. All three limitations of the cost analysis were
a consequence of the limited study resources available. The first was that the
cost data were not complete enough to allow us to include prebaseline costs
in all sites. Controlling for prebaseline differences in inpatient costs in Mem-
phis, for example, may well have diminished the association between jail
diversion and increased treatment costs. The second limitation to the cost
analysis was that two categories of costs were not explicitly included in the
analyses. First, medication costs were not separately tracked from inpatient
and outpatient treatment in the data. To collect medication cost data in future
studies would likely require access to billing or claims data. Second, training
and setup costs were excluded. Thus, costs should be interpreted as the costs
of ongoing operations. Third, state and federal prison costs were altogether
excluded from the analyses. Tracking movements of our sample into prisons
proved beyond the study resources. In the absence of accessing departments
of corrections’ arrest, release, and prisoner data from several states, this
challenge is likely to remain in future studies.
The limitations faced by the effectiveness study were also a consequence
of the limited resources available to the study. The effectiveness data were
self-reported, and thus, measurement error in outcomes may be introduced.
One alternative to using self-reported data for the alcohol and drug measures
is to collect measures directly. However, this alternative would require con-
siderable study resources. An additional concern with self-reported data is
that they tend to be more susceptible to the influence of sample attrition.
Those who leave the study may be systematically different from those who
remain, thus compromising external validity. Also, if the characteristics of
those who leave are systematically different by diversion status, there may be
selection bias. However, fully testing for and correcting for selection bias on
a site-by-site basis is outside the scope of this study.
Resources also limited the study team to tracking the sample for 12
months. For longer term outcomes, such as mental health functioning, this
period could be too short to see any sizeable effects. However, in an observa-
tional study, collecting data over a longer period of time can also present
challenges. First, sample attrition is unlikely to improve. Second, as the par-
 Cowell et al. / CRIMINAL JUSTICE DIVERSION PROGRAMS 311

ticipants in the comparison groups progress from jail into treatment, they
receive similar or the same treatment as the diverted.
The fact that this study is an observational study can be seen both as a
design limitation and as a strength. The limitation is that identifying the
effects relies on the comparability of the comparison group. The strength is
that the findings can have direct relevance to policy makers and those who
want to construct appropriate interventions.
There are two clear directions for future research on this topic. First, the
research presented here ignored differences in delivering diversion within
programs. For example, Tucson and New York City have multitiered pro-
grams, and the Lane County program embeds diversion within a drug court.
Future research should exploit these within-site differences in diversion pro-
grams. Second, the success of criminal justice diversion programs may ulti-
mately depend on whether participants receive appropriate treatment while
incarcerated as well as whether they receive appropriate treatment in the
community. Future research should thus focus on the intersection of three
treatment characteristics: what the diverted are being diverted from, what the
diverted are being diverted to, and the timing of that diversion in the criminal
justice process.

NOTES
1. Details can be found in the site reports (Cowell, Stewart, & Ng, 2002a, 2002b,
2002c, 2002d) and are available from the authors on request.
2. Study participants signed an agreement allowing agencies to release data to
study staff. However, for 56 participants we were unable to obtain complete cost data
because this agreement expired before data were collected from key agencies.
Because the date when the agreement expired was correlated with whether we
obtained data for a participant, but was not separately correlated with the impact of
diversion on cost, this expiry date served as an instrument for the Heckman selection
method.

REFERENCES
Abram, K. M., & Teplin, L. A. (1991). Co-occurring disorders among mentally ill jail
detainees: Implications for public policy. American Psychologist, 46(10), 1036-
1045.
Abram, K. M., Teplin, L. A., & McClelland, G. M. (2003). Comorbidity of severe psy-
chiatric disorders and substance use disorders among women in jail. American
Journal of Psychiatry, 150(5), 1007-1010.
Borum, R., Deane, M. W., Steadman H., & Morrissey, J. (1998). Police perspectives
on responding to mentally ill people in crisis: Perceptions of program effective-
ness. Behavioral Sciences and the Law, 16, 393-405.
312 Journal of Contemporary Criminal Justice / August 2004

Borum, R., Swanson, J., Swartz, M., & Hiday, V. (1997). Substance abuse, violent
behavior and police encounters among persons with severe mental disorder.
Journal of Contemporary Criminal Justice, 13, 236-250.
Broner, N., Borum, R, & Gawley, K. (2002). Criminal justice diversion of individuals
with co-occurring mental illness and substance use disorders: An overview. In
G. Landsberg, M. Rock, L. Berg, & A. Smiley (Eds.), pp. 83-106. Serving men-
tally ill offenders and their victims: Challenges and opportunities for mental
health professionals. New York: Springer Publishing.
Broner, N., Lamon, S. S., Mayrl, D. W., & Karopkin, M. G. (2003). Arrested adults
awaiting arraignment: Mental health, substance abuse, and criminal justice char-
acteristics and needs. Fordham Urban Law Journal, 30(2), 663-721.
Broner, N., Lattimore, P. K., Cowell, A. J., & Schlenger, W. (in press). Diverting indi-
viduals with co-occurring serious mental illness and substance use disorders:
Outcomes from a national multi-site study. Behavioral Sciences and the Law.
Broner, N., Nguyen, H. Q., Swern, A., & Goldfinger, S. (2003). Adapting a substance
abuse court diversion model for felony offenders with mental illness and sub-
stance use disorders: Initial implementation. Psychiatric Quarterly, 74(4), 361-
385.
Cosden, M., Ellens, J., Schnell, J., Yamini-Diouf, Y., & Wolfe, M. (2003). Evaluation
of a mental health treatment court with assertive community treatment. Behav-
ioral Sciences and the Law, 21, 415-427.
Cowell, A. J., Stewart, A. M., & Ng, S. W. (2002a). Assessment of the cost-effective-
ness of Eugene’s jail diversion program (Report prepared for U.S. Department of
Health and Human Services, Substance Abuse and Mental Health Administra-
tion). Research Triangle Park, NC: RTI International.
Cowell, A. J., Stewart, A. M., & Ng, S. W. (2002b). Assessment of the cost-effective-
ness of Memphis’s jail diversion program (Report prepared for U.S. Department
of Health and Human Services, Substance Abuse and Mental Health Administra-
tion). Research Triangle Park, NC: RTI International.
Cowell, A. J., Stewart, A. M., & Ng, S. W. (2002c). Assessment of the cost-effective-
ness of New York’s jail diversion program (Report prepared for U.S. Department
of Health and Human Services, Substance Abuse and Mental Health Administra-
tion). Research Triangle Park, NC: RTI International.
Cowell, A. J., Stewart, A. M., & Ng, S. W. (2002d). Assessment of the cost-effective-
ness of Tucson’s jail diversion program (Report prepared for U.S. Department of
Health and Human Services, Substance Abuse and Mental Health Administra-
tion). Research Triangle Park, NC: RTI International.
Cowell, A. J., Stewart, A. M., & Ng, S. W. (2002e). Cost-effectiveness of four criminal
justice diversion programs: Summary report (Report prepared for U.S. Depart-
ment of Health and Human Services, Substance Abuse and Mental Health
Administration). Research Triangle Park, NC: RTI International.
Deane, M. W., Steadman, H., Borum, R., Veysey, B., & Morrissey, J. (1999). Emerg-
ing partnerships between mental health and law enforcement. Psychiatric Ser-
vices, 50, 99-101.
 Cowell et al. / CRIMINAL JUSTICE DIVERSION PROGRAMS 313

Department of Health and Human Services (DHHS). (2002). Results from the 2001
National Household Survey on drug abuse: Vol. I. Summary of national findings
(National Household Survey on Drug Abuse Series H-17, DHHS Publication No.
SMA 01-3758). Rockville, MD: DHHS, Substance Abuse and Mental Health
Services Administration, Office of Applied Studies.
Ditton, P. M. (1999). Mental health and treatment of inmates and probationers
(Bureau of Justice Statistics Special Report NCJ 174463). Washington, DC: U.S.
Department of Justice.
Draine J., & Solomon, P. (1999). Describing and evaluating jail diversion services for
persons with serious mental illness. Psychiatric Services, 50, 56-61.
Dupont, R. (2002). Final report: Criminal justice diversion project. Rockville, MD:
Department of Health and Human Services, Substance Abuse and Mental Health
Services Administration.
Dupont, R., & Cochran, S. (2000). Police response to mental health emergencies—
Barriers to change. Journal of the American Academy of Psychiatry and the Law,
28, 338-344.
Gold, M. R., Siegel, J. E., Russell, L. B., & Weinstein, M. C. (1996). Cost-effective-
ness in health and medicine. New York: Oxford University Press.
Heckman, J. (1979). Sample selection bias as a specification error. Econometrica, 47,
153-161.
Hoff, R., Baranosky, M. V., Buchanan, J., Zonana, H., & Rosenheck, R. A. (1999).
The effects of a jail diversion program on incarceration: A retrospective cohort
study. Journal of the American Academy of Psychiatry and Law, 27(3), 377-386.
Lamb, H. R., Shaner, R., Elliott, D. M., DeCuir, W. J., & Foltz, J. T. (1995). Outcome
for psychiatric emergency patients seen by an outreach police-mental health
team. Psychiatric Services, 46(12), 1267-1271.
Lamb, H. R., & Weinberger, L. E. (1998). Persons with severe mental illness in jails
and prisons: A review. Psychiatric Services, 49(4), 483-492.
Lamb, H. R., Weinberger, L. E., & Reston-Parham, C. (1996). Court interventions to
address the mental health needs of mentally ill offenders. Psychiatric Services,
47(3), 275-281.
Lattimore, P. K., Broner, N., Sherman, R., Frisman, L., & Shafer, M. (2003). A com-
parison of pre-booking and post-booking diversion programs for mentally ill
substance using individuals with justice involvement: A multi-site study. Journal
of Contemporary Criminal Justice, 19(1), 30-64.
Lee, B. Stern, D., Coen, A., Bartsch, B., & Wilson, N. (2002). The Colorado Symptom
Index: A measure of distress for persons with severe and persistent mental illness
(Revision of manuscript #R-6192). Unpublished manuscript.
Shern, D. L., Wilson, N. Z., Cohen, A. S., Patrick, D. C., Foster, M., Bartsch, D. A.,
et al. (1994). Client outcomes II: Longitudinal client data from the Colorado
Treatment Outcome Study. Milbank Quarterly, 72(1),123-143.
Skinner, H. A. (1982). Drug abuse screening test. Addictive Behavior, 7, 263-371.
Steadman, H. J., Barbera, S. S., & Dennis, D. L. (1994). A national survey of jail
diversion programs for mentally ill detainees. Hospital and Community Psychia-
try, 45(11), 1109-1113.
314 Journal of Contemporary Criminal Justice / August 2004

Steadman, H. J., Cocozza, J. J., & Veysey, B. M. (1999). Comparing outcomes for
diverted and nondiverted jail detainees with mental illnesses. Law and Human
Behavior, 23(6), 615-627.
Steadman, H. J., Deane, M. W., Borum, R., & Morrissey, J. P. (2000). Comparing out-
comes of major models for police responses to mental health emergencies. Amer-
ican Journal of Public Health, 51(5), 645-649.
Steadman, H. J., Deane, M. W., Morrissey, J. P., Westcott, M., Salasin, S., & Shapiro,
S. (1999). A SAMHSA research initiative assessing the effectiveness of jail
diversion programs for mentally ill persons. Psychiatric Services, 50(12), 1620-
1623.
Steadman, H. J., Stainbrook, K. A., Griffin, P., Draine, J., Dupont, R., & Horey, C.
(2001). A specialized crisis response location as a core element for police-based
diversion programs. Psychiatric Services, 52, 219-222.
Storgaard, H., Nielsen, S. D., & Gluud, C. (1994). The validity of the Michigan Alco-
holism Screening Test (MAST). Alcohol & Alcoholism, 29(5), 493-502.
Teplin, L. A. (1990a). Detecting disorder: The treatment of mental illness among jail
detainees. Journal of Consulting and Clinical Psychology, 58, 233-236.
Teplin, L. A. (1990b). The prevalence of severe mental disorder among urban male
detainees: Comparison with the Epidemiologic Catchment Area Program. Amer-
ican Journal of Public Health, 80(6), 663-669.
Teplin, L. A. (1994). Psychiatric and substance abuse disorders among male urban jail
detainees. American Journal of Public Health, 84, 292-293.
Torrey, E. F., Steiber, J., Ezekiel, J., Wolfe, S. M., Sharfstein, J., & Flynn, L. M.
(1992). Criminalizing the seriously mentally ill: The abuse of jails as mental hos-
pitals. Innovations & Research (Vol. 2, pp. 11-14). Washington, DC: Public Citi-
zen’s Health Research Group.
Ware, J. E., Kosinski, M., & Keller, S. D. (1996). A 12-item short-form health survey:
SF-12 construction and preliminary tests of reliability and validity. Medical
Care, 34, 220-233.
Wilson, D. J. (2000). Drug use, testing, and treatment in jails (Bureau of Justice Sta-
tistics special report NCJ 179999). Washington, DC: Bureau of Justice Statistics.

Dr. Alexander J. Cowell is an economist at RTI International, Research Triangle Park, North
Carolina. His research interests include the economics of mental health, substance use, and
crime.
Nahama Broner, Ph.D., a clinical psychologist, is senior research psychologist at RTI Interna-
tional and adjunct associate professor at New York University. Her research and evaluation
interests have focused on criminal justice involving populations with serious mental illness, co-
occurring substance use, and trauma disorders and their service use, as well as diversion and
reentry intervention and technology transfer of evidence-based practices.
Dr. Randolph Dupont is a professor and clinical psychologist at the University of Memphis
School of Urban Studies and Public Policy, Department of Criminology and Criminal Justice.
His research interests include crisis intervention strategies, jail diversion studies, and addictive
disorders.