0
We're unable to sign you in at this time. Please try again in a few minutes.
Retry
We were able to sign you in, but your subscription(s) could not be found. Please try again in a few minutes.
Retry
There may be a problem with your account. Please contact the AMA Service Center to resolve this issue.
Contact the AMA Service Center:
Telephone: 1 (800) 262-2350 or 1 (312) 670-7827  *   Email: subscriptions@jamanetwork.com
Error Message ......
Original Article |

Effect of Mental Health Courts on Arrests and Jail Days:  A Multisite Study FREE

Henry J. Steadman, PhD; Allison Redlich, PhD; Lisa Callahan, PhD; Pamela Clark Robbins, BA; Roumen Vesselinov, PhD
[+] Author Affiliations

Author Affiliations: Policy Research Associates Inc, Delmar (Drs Steadman and Callahan and Ms Robbins); School of Criminal Justice, University at Albany, Albany (Dr Redlich); and Department of Economics, Queens College, City University of New York (Dr Vesselinov), New York.


Arch Gen Psychiatry. 2011;68(2):167-172. doi:10.1001/archgenpsychiatry.2010.134.
Text Size: A A A
Published online

Mental health courts (MHCs) are an increasingly popular postbooking jail diversion program. While there is some disagreement about which was the first MHC,1,2 there is no debate about the robust expansion of these courts during the past decade from 1 or 2 courts in 1997 to approximately 250 today.3,4 Mental health courts have the laudable goal of moving persons with serious mental illness out of the criminal justice system and into community treatment without sacrificing public safety.5,6 Mental health courts share some common features4 but their implementation widely varies by jurisdiction, by judge, and across time. Consequently, single-site evaluations of the effectiveness of MHCs in meeting their primary objective of enhanced public safety are limited by the idiosyncrasies of the particular court.

In general, potential clients are referred to the MHC staff by jail personnel, defense attorneys, and others who become familiar with the defendant.7 If the potential enrollee meets eligibility criteria and chooses to participate in the MHC, he or she then follows the specific procedures for enrollment into that court, such as having a hearing before the MHC judge, at which time the individual may enter a guilty plea and agree to the terms established by the MHC team and to the disposition of the criminal charges. Most MHCs require participation in treatment as a term of enrollment. The individual is then released into the community under MHC supervision with a subsequent status hearing date, usually weekly at the beginning.4 Courts can use the “power of the gavel” to sanction participants who violate the terms of their release through bench warrants, temporary reincarceration, or outright revocation, while also facilitating treatment options for these often difficult clients.8,9

Most research on MHCs to date has been case studies, pre-enrollment/postenrollment studies, or treatment-as-usual (TAU) comparison studies involving a single court. Overall, the studies are equivocal. Two of the most ambitious, 1 with a well-chosen comparison group10 and 1 randomized controlled trial,11 found no difference in subsequent arrests between the MHC enrollees and the comparison/control subjects. The other 2 studies with a control group12,13 found the MHC enrollees to be about one-third less likely to be subsequently arrested. Two single-site studies using pre-enrollment/postenrollment designs14,15 found that MHC enrollees were much less likely to be arrested in the year following enrollment than in the year before.

What is missing from the MHC literature to this point is an experimental design that includes treatment and comparison samples from multiple locales. Because of MHCs' notorious idiosyncrasies,16 it is important to study more than 1 court using the same methodology. For innovative interventions to become evidence-based practices, research must progress from studies of single courts to those involving multiple courts. Additionally, many of the studies on MHCs have had methodological limitations such as comparison groups that were purposely selected by the MHC judge, comparisons made across inconsistent points in time, and inclusion of only retrospective observations. In this study, we attempted to overcome many of these limitations.

This study is a 4-site, prospective, longitudinal, quasiexperimental study. The MHC and TAU samples were interviewed and followed up for 18 months at each site. The core research questions addressed here are (1) is participation in an MHC associated with more favorable criminal justice outcomes than processing through the regular criminal court system? and (2) for what types of defendants do MHCs produce the most favorable criminal justice outcomes?

SITE SELECTION AND PARTICIPANTS

The 4 MHCs included in this study are San Francisco County, CA, Santa Clara County, CA, Hennepin County (Minneapolis), MN, and Marion County (Indianapolis), IN. These courts were selected based on a national survey included in an earlier phase of the study.4 To be included, the courts were required to be large enough to have a substantial caseload from which to draw a sample, have operated long enough to have stability, and represent a range of types of courts from level of sanctioning to types of defendants such as both misdemeanor and felony cases. In addition, the courts had to be in jurisdictions with large county jails to ensure sufficient sampling for the TAU group.

The treatment group in each site comprises newly enrolled MHC participants (MHC group; n = 447). Data from the MHCs were reported on a weekly basis to the research team as to the sex, age, criminal charge, race, and diagnosis of the enrollees. The comparison group consists of similar subjects who were eligible for the MHC but were never referred to it or were never rejected from the MHC (TAU group; n = 600). Newly booked jail detainees identified by jail mental health staff as having mental health problems were matched as closely as possible to the MHC enrollees, first for sex and criminal charges, and then for race, age, and diagnosis. The actual sample characteristics are seen in Table 1. Subjects were interviewed at baseline/study enrollment, and 70% were interviewed again at 6 months. We conducted analyses of variance and χ2 comparison analyses and determined that the interviewed and noninterviewed subjects did not differ in sex, race, age, or whether they had received treatment in the prior 6 months. They differed in diagnosis, with a larger proportion of subjects with schizophrenia and a smaller proportion with depression being interviewed at 6 months. As part of their participation in the study, subjects provided informed consent allowing access to their mental health and criminal justice records. The study was approved by a number of federally sanctioned institutional review boards at the local and state level in addition to the study's coordinating center institutional review board.

Table Graphic Jump LocationTable 1. Characteristics of Study Subjects

The study courts are similar across many aspects including the types of crimes and clinical diagnoses they admit. There are some court-specific differences. For example, we found differences in the length of court supervision at the 1-year mark; the percentage of participants still receiving court supervision ranged from 40% to 84%.8 Successful completion and termination rates vary as well; 7% to 41% had graduated by 12 months, and 3% to 39% had been terminated. Clearly, how each court interprets its eligibility criteria, guidelines for success and termination, and period of supervision may vary.

Using program data from the study sites, we find that 71% of MHCs and 38% of TAUs received community mental health treatment (includes outpatient treatment, case management, and medication management) in the 12-month follow-up period (odds ratio, 4.1). Of those who received treatment, those in the MHC group (median, 20.2 hours) received significantly more treatment than those in the TAU group (median, 8.6 hours; P < .001). These outcomes and treatment differences were taken into account in our analyses by using propensity scores where applicable.

VARIABLES

The variables described here include the public safety outcomes of number of new arrests, annualized arrest rates, and county jail and state prison incarceration days. Arrest data were obtained from the individual's Federal Bureau of Investigation report and include only new arrests, excluding warrants and violations. Annualized rates of arrest are number of new arrests for days not incarcerated in that county or state prison system. Incarceration days were acquired from the local jail records and the state departments of correction.

Rearrest, based on the Federal Bureau of Investigation reports, is measured as a binary variable and indicates (yes/no) whether the person was arrested in the post–18-month period. The rearrest rate is measured by the number of arrests in that period corrected for time in the community. Similar to other studies,17 we normalize arrest rates by constructing an annualized number of arrests variable, which is computed as the number of arrests for this 18-month period divided by the time in the community and multiplied by 365. A limitation of this calculation is that it does not include days in a psychiatric hospital, as they were not accessible from each site. The incarceration variable measures time (in days) spent in jail and prison during the 18-month period. The change scores are computed as the difference between the post–18-month period value of the variable and the pre–18-month period value. Jail records do not indicate why someone is booked into jail: for a new arrest, to serve a sentence, to be held for another jurisdiction, or on a warrant for a technical violation. Consequently, one limitation to these data are that we cannot ascertain what proportion of jail days are for MHC sanctions or for other reasons.

Explanatory variables include study group (MHC vs TAU); individual characteristics such as white (yes/no; self-report or official records), female (yes/no), age in years, and most severe diagnosis (schizophrenia, bipolar disorder, depression, or other) obtained from the MHC evaluation or jail treatment records; study site; drug and alcohol use to intoxication in the 30 days prior to MHC or jail involvement (yes/no); and prior number of arrests and incarceration days in the 18 months before entering the MHC or jail. Diagnosis was obtained from tracking data provided by each site and treatment records, when available. Behavioral health measures including recent drug and alcohol use were obtained through self-report at baseline. All other data, which were collected on all subjects regardless of participation in the follow-up interview, were obtained from official records.

STATISTICS
Sample Selection Bias

The participants in this study were not randomly assigned to the 2 study groups, although efforts were made to match the 2 samples as described above. To address possible sample selection bias, we used a modification of the propensity score approach proposed by Rosenbaum and Rubin.18,19 We constructed a logistic regression model with a binary dependent variable indicating 1 = MHC and 0 = TAU. First, we entered in the model the basic variables of age, race, sex, and site. Second, we included a pool of all available potential explanatory variables: personal characteristics including ever married, education, lived with biological father until 15 years of age, father ever arrested, and father used illegal drugs; mental health history including age the individual first saw a mental health professional, age at first mental hospitalization, and ever having psychiatric hospitalization; current mental health factors including mutually exclusive diagnostic category, Insight and Treatment Attitudes Questionnaire, Colorado Symptom Index scores, mental health treatment in past 6 months, self-reported compliance with treatment and medication, and other types of medical treatment; child physical and sexual abuse and baseline violence; substance use and treatment such as alcohol and illegal drug use in past 30 days and received substance abuse treatment in past 6 months; criminal justice variables including age at first arrest, number of arrests since 15 years of age, number of pre–18-month incarceration days and arrests; annualized pre–18-month arrests; and target arrest, charge level, and most serious offense. Variables selected for the model by the stepwise procedure were marital status, Colorado Symptom Index, days using illegal drugs in last 30 days, diagnosis of depression, ever been hospitalized, received treatment for medical problem, violence at baseline interview, age at first arrest, and target arrest charge level (warrant, violation, misdemeanor, felony). These variables, along with those entered on the first stage (age, race, sex, and site), constitute the variables included in the propensity score model. The model has good characteristics, with a pseudo R2 (Nagelkerke) of 0.244 and an area under the curve of 0.750. We used this model to generate the propensity scores, and the resulting propensity score is included in all models comparing MHC and TAU samples, thus adjusting for selection bias.

0-Inflated Models

Some of the outcome variables have many 0s. For example, about 46% of the people had no arrests in the post–18-month period. Because so many 0s cannot be handled by ordinary least squares regression, we address this problem by implementing the 0-inflated Poisson (ZIP) models that are specifically designed to handle counts of rate variables with many 0s. The Poisson regression model is a type of generalized linear model and is also called a log-linear model. The ZIP model20,21 is a special Poisson mixture model with 2 classes, 1 of which has a fixed value of 0, and the other different from 0. The model defines unobserved heterogeneity with the purpose of distinguishing between the subjects who were not arrested at all from those who were. After the ZIP models were estimated, we performed the Vuong test22 in which the ZIP model is compared with the standard Poisson model. In all cases described in this article, the Vuong test has very large positive values, favoring the ZIP model.

Quantile Regression Model

Some of our outcome variables have outliers and large variance. For example, the change score of the annualized number of arrests exhibits overdispersion, particularly for the TAU group, in which the standard deviation was 8 times larger than the mean. Usually the overdispersion problem can be addressed with standard Poisson or negative binominal regression but the models are not statistically significant in this case. Therefore, we used the quantile regression model.23,24 The main characteristic of the model is that, instead of using the deviation from the mean (as in OLS), it can use quantile (or percentile) for this purpose. Most used are the median and the other quartiles and the interquartile range. The quantile regression is robust to outliers, and it can handle the problem with unequal variation for variables and samples. In addition, we bootstrapped the standard errors of the coefficients for the quantile regression models. The interpretation of the models' coefficients is similar to the regular OLS regression except that the model is based not on the mean (OLS) but on a particular quantile (Q1, median, Q3) or the interquartile range.

There was no stepwise selection procedure available for either ZIP models or quantile regression. We first ran a standard multiple regression model with all available variables under a stepwise procedure. Then we included the selected variables in the quantile regression model, together with the propensity score variable to adjust for sample selection bias. In the presence of overdispersion, we bootstrapped the standard errors of the coefficients for the quantile regression models and the logistic regression model for rearrest.

The 2 major outcomes that have become the public policy criterion standard as to whether MHCs work without compromising public safety are arrests and jail days.

ARRESTS

Arrests are examined 18 months before and after MHC enrollment for the experimental group (MHC group) and for 18 months before and after the target jail admission for the jail sample (TAU group). Because MHCs are a postbooking diversion, by definition subjects in both groups have arrest histories in the 18-month pre-entry period. When excluding the target arrest from the data, the 2 groups remain similar in the pre–18-month period, with 93% of the MHC and 95% of the TAU sample having at least 1 additional prearrest. In the post–18-month period, however, the MHC sample (49%) is significantly less likely than the TAU sample (58%) to be arrested (P = .006).

Simply being rearrested or not, however, is a blunt measure of recidivism. It does not take into account time at risk of rearrest. Therefore, as shown in Table 2, we calculated the annualized rearrest rates of the MHC and TAU samples for the time they were known to be in the community. Both samples show a decline in annual arrest rate from 2.1 to 1.3 per year in the MHC group and from 2.6 to 2.0 per year in the TAU group. However, the 0.8 per year reduction in the MHC group is significantly different (P < .001) than the 0.6 per year reduction in the TAU. With the exception of the Minnesota site, the percentage of reduction in arrests per year was greater for MHC than for TAU, being as much as 5 times as much in San Francisco and 2½ times as much in Indianapolis.

Table Graphic Jump LocationTable 2. Annualized Arrests by Sample and Site

One final lens on rearrest is to examine how the MHC subjects do during court supervision and once supervision has ended. At 12 months, we identified the court status of the MHC sample across all 4 sites: 60% were still receiving MHC supervision, 20% had graduated, and 20% have been terminated. The annualized arrest rate while receiving court supervision for the MHC subjects is 1.04 arrests, including those still receiving supervision, graduated, or terminated. For subjects who are either terminated or graduated before 12 months, their postsupervision annualized arrest rate is 0.79 up to the time of their termination or graduation. A new arrest does not necessarily preclude graduation or result in termination. This postsupervision rearrest rate is a bit deceptive, however, in that the postsupervision rearrest rate is 1.33 for subjects in the MHC group who are terminated and only 0.07 for those who graduated. Clearly, there is a longer-term effect of supervision that continues after court supervision ends. That the rate for persons whose participation was terminated is much higher after supervision ends is somewhat tautological because one reason that MHC enrollees are excluded is that they have new arrest charges.

INCARCERATION DAYS

The second major measure of recidivism analyzed is the number of postentry jail and prison days. Table 3 shows that, for the MHC sample, there is a small increase in the number of incarceration days from the pre–18-month period (73 days) to the post–18-month period (82 days). For the TAU sample, however, there is a 105% increase in incarceration days (from 74 to 152 days). The difference in the post–18-month period between the MHC and TAU is significant (P < .001). Likewise, the difference is statistically significant for all 4 sites. In addition, the magnitude of change in incarceration days (9 vs 78 days) of the 2 samples is statistically significant (P < .001) and consistent across all 4 sites.

Table Graphic Jump LocationTable 3. Average Incarceration Days Before and After 18 Months
COMPARISON IN THE MHC

The final analyses of public safety outcomes focuses on what type of defendants do better or worse in MHCs using a ZIP regression model to examine 2 outcomes: annualized number of postdiversion arrests and number of jail and prison days for 18 months after diversion. We entered the explanatory variables indicated in the “Methods” section into the 2 ZIP regression models: annualized pre–18-month arrests (Table 4) and annualized pre–18-month incarceration days (Table 5). For both models, the criminogenic factors are the most consistently significant. Annualized rearrests in the post–18-month follow-up period is more likely for those who have more pre–18-month annualized arrests and more pre–18-month incarceration days. In addition, those who received no mental health treatment in the 6 months prior to entering the MHC at baseline are also more likely to be arrested in the 18-month follow-up.

Table Graphic Jump LocationTable 4. Factors Related to Annualized Number of Post–18-Month Arrests: 0-Inflated Poisson Model
Table Graphic Jump LocationTable 5. Factors Related to Number of Post–18-Month Days of Jail Reincarceration: 0-Inflated Poisson Model

Similarly, factors associated with more incarceration days during follow-up include the criminogenic factors of annualized pre–18-month arrests and number of pre–18-month incarceration days. Also, a number of clinical factors emerge. As with annualized postarrests, the absence of treatment at baseline is highly associated with more incarceration days during follow-up. Further, having a diagnosis of schizophrenia or depression rather than bipolar disorder and having used illegal drugs in the past 30 days are significantly associated with more incarceration days during the follow-up.

The appropriate question for MHCs is not, “do they work?” but, “for whom, and under what circumstances, do they work?” Nonetheless, public policy debates about these courts demand some global assessments. As we have seen here, across 4 diverse MHCs, MHC participants have significantly better outcomes on arrests and number of incarceration days than the TAU jail comparison group. On 5 key public safety outcome measures (subsequent arrest rates, number of subsequent arrests, reduction in pre- to post-MHC arrests, number of subsequent incarceration days, and change in pre- to post-MHC subsequent incarceration days) the overall MHC group is significantly lower than the TAU group.

Looking at the 4 sites individually, the pattern of MHC participants being lower than the TAU participants in number of arrests and both number of days of incarceration variables holds across all 4 sites. On arrest rates and pre/post-MHC number of arrests, the Minnesota site is different from the other 3 in that no significant effect for being in the MHC is found. One possible explanation for this inconsistency is found in a July, 2009, Minnesota in-house article.25 Looking at 2007 and 2008 data from a 225-person MHC sample, researchers found no significant difference before and after MHC in average number of arrests, just as we did. However, when they examined a subgroup with longer exposure to the program (n = 25) and one that specifically received housing via the program (n = 10), both had statistically significant improvement after compared with before court involvement. Our sample is comparable with their total group, in which they too found no differences, suggesting that the same factors (ie, length in program and access to housing) may be instrumental in achieving these public safety outcomes. The average number of jail days increased for both samples. However, the small increase of 9 days for the MHC is not statistically significant and is unlikely to have practical implications. At first glance, data suggest that, because incarceration days increased for the MHC, the goal of reduced incarceration was not met. However, when compared with the 78-day increase for TAUs, the MHCs did much better than the TAUs in the follow-up (F1 = 76.98; P < .001).

This first multisite, prospective study of MHCs offers encouragement that they can achieve the public safety outcomes that funders and the public want. Our data do not comprehensively address the key questions of who the courts are most effective for or what mechanisms produce positive outcomes. These important questions await further data from this and other studies. Until then, it appears that MHCs are diversion programs for justice-involved persons with mental illness and, usually, co-occurring substance abuse disorders that warrant public policy support.

Correspondence: Henry J. Steadman, PhD, Policy Research Associates Inc, 345 Delaware Ave, Delmar, NY 12054 (hsteadman@prainc.com).

Submitted for Publication: January 22, 2010; final revision received July 26, 2010; accepted August 2, 2010.

Published Online: October 4, 2010. doi:10.1001/archgenpsychiatry.2010.134

Financial Disclosure: None reported.

Funding/Support: This study was supported by the Research Network on Mandated Community Treatment of the John D. and Catherine T. MacArthur Foundation.

Additional Contributions: The authors would like to thank John Monahan, PhD, and other Research Network members for their insightful input into the study; Asil Ozdogru, MA, and Karli Keator, BA, for their contributions to ongoing data collection, management, and analysis; and Kathleen Bolling, MA, Pam Stenhjem, MA, and the numerous on-site research assistants and others who facilitated data collection.

Boothroyd  RAPoythress  NGMcGaha  APetrila  J The Broward Mental Health Court: process, outcomes, and service utilization. Int J Law Psychiatry 2003;26 (1) 55- 71
PubMed
Steadman  HJDavidson  SBrown  C Law & psychiatry: mental health courts: their promise and unanswered questions. Psychiatr Serv 2001;52 (4) 457- 458
PubMed
 GAINS Web site. http://www.gainscenter.samhsa.gov/html/programs/jd_map.aspAccessed October 29, 2009
Redlich  ADSteadman  HJMonahan  JRobbins  PCPetrila  J Patterns of practice in mental health courts: a national survey. Law Hum Behav 2006;30 (3) 347- 362
PubMed
Steadman  HJNaples  M Assessing the effectiveness of jail diversion programs for persons with serious mental illness and co -occurring substance use disorders. Behav Sci Law 2005;23 (2) 163- 170
PubMed
Steadman  HJMorris  SMDennis  DL The diversion of mentally ill persons from jails to community-based services: a profile of programs. Am J Public Health 1995;85 (12) 1630- 1635
PubMed
Steadman  HJRedlich  ADGriffin  PPetrila  JMonahan  J From referral to disposition: case processing in seven mental health courts. Behav Sci Law 2005;23 (2) 215- 226
PubMed
Redlich  ADSteadman  HJCallahan  LRobbins  PCVessilinov  ROzdoğru  AA The use of mental health court appearances in supervision [published online July 17, 2010]. Int J Law Psychiatry 2010;
PubMed10.1016/j.ijlp.2010.06.010
Griffin  PASteadman  HJPetrila  J The use of criminal charges and sanctions in mental health courts. Psychiatr Serv 2002;53 (10) 1285- 1289
PubMed
Christy  ACPoythress  NGBoothroyd  RAPetrila  JMehra  S Evaluating the efficiency and community safety goals of the Broward County Mental Health Court. Behav Sci Law 2005;23 (2) 227- 243
PubMed
Cosden  MEllens  JSchnell  JYamini-Diouf  Y Executive summary: Evaluation of the Santa Barbara County mental health treatment court with intensive case management.  Santa Barbara, CA Counseling/Clinical/School Psychology Program, Gevirtz Graduate School of Education, University of California, Santa Barbara2004;http://consensusproject.org/downloads/exec.summary.santa.barbara.evaluation.pdfAccessed November 25, 2009
Moore  MEHiday  VA Mental health court outcomes: a comparison of re-arrest and re-arrest severity between mental health court and traditional court participants. Law Hum Behav 2006;30 (6) 659- 674
PubMed
McNiel  DEBinder  RL Effectiveness of a mental health court in reducing criminal recidivism and violence. Am J Psychiatry 2007;164 (9) 1395- 1403
PubMed
Herinckx  HASwart  SCAma  SMDolezal  CDKing  S Rearrest and linkage to mental health services among clients of the Clark County mental health court program. Psychiatr Serv 2005;56 (7) 853- 857
PubMed
Ferguson  AHornby  HZeller  DAlaska Mental Health Trust Authority, Outcomes From the Last Frontier: an Evaluation of the Anchorage Mental Health Court.  South Portland, ME Hornby Zeller Associates2008;
Wolff  NPogorzelski  W Measuring the effectiveness of mental health courts: challenges and recommendations. Psychol Public Policy Law 2005;11 (4) 539- 56910.1037/1076-8971.11.4.539
Loughran  TAMulvey  EPSchubert  CAFagan  JPiquero  ARLosoya  SH Estimating a dose-response relationship between length of stay and future recidivism in serious juvenile offenders. Criminology 2009;47 (3) 699- 740
PubMed
Rosenbaum  PRRubin  DB The central role of the propensity score in observational studies for causal effects. Biometrika 1983;70 (1) 41- 5510.1093/biomet/70.1.41
Rosenbaum  PR Observational Studies. 2nd New York, NY Springer-Verlag2002;
Lambert  D Zero-inflated Poisson regression models with an application to defects in manufacturing. Technometrics 1992;34 (1) 1- 1410.2307/1269547
Long  JS Regression Models for Categorical and Limited Dependent Variables.  Thousand Oaks, CA Sage Publications1997;
Vuong  QH Likelihood Ratio Tests for model selection and non-nested hypotheses. Econometrica 1989;57 (2) 307- 33310.2307/1912557
Koenker  RHallock  K Quantile regression. J Econ Perspect 2001;15 (4) 143- 156
Koenker  R Quantile Regression.  New York, NY Cambridge University Press2005;
Baiocchi  AWhetstone  S Midterm evaluation: FUSE. http://documents.csh.org/documents/ResourceCenter/HotTopicsSH/2010-FrequentUsers/MidtermReport_FUSE.PDFAccessed August 23, 2010

Figures

Tables

Table Graphic Jump LocationTable 1. Characteristics of Study Subjects
Table Graphic Jump LocationTable 2. Annualized Arrests by Sample and Site
Table Graphic Jump LocationTable 3. Average Incarceration Days Before and After 18 Months
Table Graphic Jump LocationTable 4. Factors Related to Annualized Number of Post–18-Month Arrests: 0-Inflated Poisson Model
Table Graphic Jump LocationTable 5. Factors Related to Number of Post–18-Month Days of Jail Reincarceration: 0-Inflated Poisson Model

References

Boothroyd  RAPoythress  NGMcGaha  APetrila  J The Broward Mental Health Court: process, outcomes, and service utilization. Int J Law Psychiatry 2003;26 (1) 55- 71
PubMed
Steadman  HJDavidson  SBrown  C Law & psychiatry: mental health courts: their promise and unanswered questions. Psychiatr Serv 2001;52 (4) 457- 458
PubMed
 GAINS Web site. http://www.gainscenter.samhsa.gov/html/programs/jd_map.aspAccessed October 29, 2009
Redlich  ADSteadman  HJMonahan  JRobbins  PCPetrila  J Patterns of practice in mental health courts: a national survey. Law Hum Behav 2006;30 (3) 347- 362
PubMed
Steadman  HJNaples  M Assessing the effectiveness of jail diversion programs for persons with serious mental illness and co -occurring substance use disorders. Behav Sci Law 2005;23 (2) 163- 170
PubMed
Steadman  HJMorris  SMDennis  DL The diversion of mentally ill persons from jails to community-based services: a profile of programs. Am J Public Health 1995;85 (12) 1630- 1635
PubMed
Steadman  HJRedlich  ADGriffin  PPetrila  JMonahan  J From referral to disposition: case processing in seven mental health courts. Behav Sci Law 2005;23 (2) 215- 226
PubMed
Redlich  ADSteadman  HJCallahan  LRobbins  PCVessilinov  ROzdoğru  AA The use of mental health court appearances in supervision [published online July 17, 2010]. Int J Law Psychiatry 2010;
PubMed10.1016/j.ijlp.2010.06.010
Griffin  PASteadman  HJPetrila  J The use of criminal charges and sanctions in mental health courts. Psychiatr Serv 2002;53 (10) 1285- 1289
PubMed
Christy  ACPoythress  NGBoothroyd  RAPetrila  JMehra  S Evaluating the efficiency and community safety goals of the Broward County Mental Health Court. Behav Sci Law 2005;23 (2) 227- 243
PubMed
Cosden  MEllens  JSchnell  JYamini-Diouf  Y Executive summary: Evaluation of the Santa Barbara County mental health treatment court with intensive case management.  Santa Barbara, CA Counseling/Clinical/School Psychology Program, Gevirtz Graduate School of Education, University of California, Santa Barbara2004;http://consensusproject.org/downloads/exec.summary.santa.barbara.evaluation.pdfAccessed November 25, 2009
Moore  MEHiday  VA Mental health court outcomes: a comparison of re-arrest and re-arrest severity between mental health court and traditional court participants. Law Hum Behav 2006;30 (6) 659- 674
PubMed
McNiel  DEBinder  RL Effectiveness of a mental health court in reducing criminal recidivism and violence. Am J Psychiatry 2007;164 (9) 1395- 1403
PubMed
Herinckx  HASwart  SCAma  SMDolezal  CDKing  S Rearrest and linkage to mental health services among clients of the Clark County mental health court program. Psychiatr Serv 2005;56 (7) 853- 857
PubMed
Ferguson  AHornby  HZeller  DAlaska Mental Health Trust Authority, Outcomes From the Last Frontier: an Evaluation of the Anchorage Mental Health Court.  South Portland, ME Hornby Zeller Associates2008;
Wolff  NPogorzelski  W Measuring the effectiveness of mental health courts: challenges and recommendations. Psychol Public Policy Law 2005;11 (4) 539- 56910.1037/1076-8971.11.4.539
Loughran  TAMulvey  EPSchubert  CAFagan  JPiquero  ARLosoya  SH Estimating a dose-response relationship between length of stay and future recidivism in serious juvenile offenders. Criminology 2009;47 (3) 699- 740
PubMed
Rosenbaum  PRRubin  DB The central role of the propensity score in observational studies for causal effects. Biometrika 1983;70 (1) 41- 5510.1093/biomet/70.1.41
Rosenbaum  PR Observational Studies. 2nd New York, NY Springer-Verlag2002;
Lambert  D Zero-inflated Poisson regression models with an application to defects in manufacturing. Technometrics 1992;34 (1) 1- 1410.2307/1269547
Long  JS Regression Models for Categorical and Limited Dependent Variables.  Thousand Oaks, CA Sage Publications1997;
Vuong  QH Likelihood Ratio Tests for model selection and non-nested hypotheses. Econometrica 1989;57 (2) 307- 33310.2307/1912557
Koenker  RHallock  K Quantile regression. J Econ Perspect 2001;15 (4) 143- 156
Koenker  R Quantile Regression.  New York, NY Cambridge University Press2005;
Baiocchi  AWhetstone  S Midterm evaluation: FUSE. http://documents.csh.org/documents/ResourceCenter/HotTopicsSH/2010-FrequentUsers/MidtermReport_FUSE.PDFAccessed August 23, 2010

Correspondence

CME
Meets CME requirements for:
Browse CME for all U.S. States
Accreditation Information
The American Medical Association is accredited by the Accreditation Council for Continuing Medical Education to provide continuing medical education for physicians. The AMA designates this journal-based CME activity for a maximum of 1 AMA PRA Category 1 CreditTM per course. Physicians should claim only the credit commensurate with the extent of their participation in the activity. Physicians who complete the CME course and score at least 80% correct on the quiz are eligible for AMA PRA Category 1 CreditTM.
Note: You must get at least of the answers correct to pass this quiz.
You have not filled in all the answers to complete this quiz
The following questions were not answered:
Sorry, you have unsuccessfully completed this CME quiz with a score of
The following questions were not answered correctly:
Commitment to Change (optional):
Indicate what change(s) you will implement in your practice, if any, based on this CME course.
Your quiz results:
The filled radio buttons indicate your responses. The preferred responses are highlighted
For CME Course: A Proposed Model for Initial Assessment and Management of Acute Heart Failure Syndromes
Indicate what changes(s) you will implement in your practice, if any, based on this CME course.
Submit a Comment
A common misapplication of propensity scores
Posted on March 1, 2011
Sue M. Marcus, PhD
New York State Psychiatric Institute,
Conflict of Interest: None Declared
Steadman et al1 evaluate the effect of mental health courts (MHCs) versus treatment as usual (TAU) for those people in the justice system with serious mental illness using observational data. Because treatment is not assigned randomly, selection bias is an issue, in that those in the MHC group may differ systematically at baseline from those in the TAU group.
Propensity score matching (PSM)2 is a popular statistical technique for addressing selection bias in observational studies. The Steadman et al paper1 generates propensity scores and ‘the resulting propensity score is included in all models comparing MHC and TAU samples, thus adjusting for selection bias.’ A common misconception is that adding the propensity score as a covariate in regression analyses will adjust for selection bias due to these covariates; however there is no theoretical guarantee that this approach adequately addresses selection bias.3,4 Rather, PSM should be used to provide a basis for matching MHC subjects to TAU subjects.2
Furthermore, the Steadman et al paper claims to use propensity scores to adjust for differences in received community mental health treatment during the 12-month follow-up period. However, adjustment for post- treatment covariates often leads to under- or over-estimation of the true treatment effect.5
Rather, Steadman et al could have used propensity score matching to address whether those who were and were not in the mental health courts differed with respect to important confounders. Consider, for example, an extreme hypothetical example. Suppose that those in the MHC group were all men and those in TAU were all women. In this case, it would be impossible to know the effect of MHCs for women and men together. Without understanding whether those in MHC and TAU have overlapping covariate distributions, it is impossible to know whether MHC is superior.
In addition, Steadman et al say that their data ‘do not comprehensively address the key questions of who the courts are most effective for’; however, they could have used propensity score matching to answer this question. Propensity score matching is particularly useful for examining who benefits most from the treatment, with respect to a large set of potential treatment-moderating characteristics.
Sue M. Marcus, Ph.D.1,2 Robert D. Gibbons, Ph.D.2
Author Affiliation:
1. Division of Biostatistics, Columbia University New York State Psychiatric Institute;
2. Departments of Medicine, Health Studies and Center for Health Statistics, University of Chicago. Correspondence: Sue Marcus PhD, New York State Psychiatric Institute, 1051 Riverside Drive Unit 48, New York NY 10032. Financial Disclosure: None reported.
References:
1. Steadman HJ, Redlich A, Callahan L, Robbins PC and Vesselinov R. Effect of mental health courts on arrests and jail days. Archives of General Psychiatry. 2010 online first.
2. Rosenbaum PR and Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983; 70:41 -55.
3. Rosenbaum PR. Observational Studies. New York: Springer-Verlag, 2002
4. Marcus S: Estimating the long-term benefits of Head Start. In: Oden S, Schweinhart LJ and Weikart DP with Marcus SM and Xie y: Into Adulthood: A Study of the Effects of Head Start 1999. Ypsilanti MI: High Scope Press.
5. Rosenbaum PR. The consequences of adjustment for a concomitant variable that has been affected by treatment. Journal of the Royal Statistical Society, Series A. 1984; 147,656-666.

Conflict of Interest: None declared
Authors' Reply to Critique
Posted on March 10, 2011
Henry J Steadman, PhD
Policy Research Associates, Inc,,
Conflict of Interest: None Declared
The propensity score matching (PSM) is a technique that we considered for this case. However, we decided to use a more direct, simple, and straightforward approach. First, we compute the propensity scores and then include them as an independent variables in the new regression model. In doing this we are using the classical interpretation of the partial regression coefficients or ceteris paribus in the multiple regression model. In other words, we interpret each regression coefficient as "all other things being equal or held constant." The variables “held constant” include the propensity score variable. In our interpretation of this standard approach, this means keeping the propensity score for the control/treatment group constant and thus adjusting for selection bias. This classical interpretation of the partial regression coefficients is still valid in this case. Of course, this is an observational study and a simple adjustment or even PSM does not solve the selection bias problem in its entirety. It is also true that the success of the approach is influenced by the covariate distributions and extreme cases such as the hypothetical example presented in the critique. To prevent this situation, we tried different approaches to investigate how robust our results are, including comparing results for low, medium, and high propensity scores and using nonparametric estimation methods. We also performed a thorough investigation of the distributions on major important variables, and we are confident that there are no extreme cases as suggested in the hypothetical example in the critique.

Conflict of Interest: None declared
Submit a Comment

Multimedia

Some tools below are only available to our subscribers or users with an online account.

Web of Science® Times Cited: 29

Related Content

Customize your page view by dragging & repositioning the boxes below.

Articles Related By Topic
Related Collections
JAMAevidence.com

Care at the Close of Life EDUCATION GUIDES
Palliative Care for Prison Inmates