Longer-Term Impacts of Mental Health Courts: Recidivism Two Years After Exit

The study MHC, which serves the District of Columbia, accepts arrestees with severe mental illness who are competent, are charged with misdemeanors, and have no pending charge or conviction in the prior five years of a dangerous or violent felony. The MHC monitors participants for treatment and behavior compliance at monthly status hearings. When participants demonstrate four to six consecutive months of substantial compliance with court mandates, they graduate and have their charges dismissed. Those who repeatedly fail to meet court mandates are returned to TCC for adjudication.

Unlike most MHCs, this court benefits from a federal pretrial services agency that screens all arrestees for severe mental illness the day after arrest and dedicates a specialized services unit (SSU) to provide individualized supervision and services (case management and linkage to mental health service agencies, plus drug testing and treatment for participants with co-occurring mental and substance use disorders) during pretrial release to persons who screen positive, whether their cases are diverted to the MHC or remain in TCC. On receiving an MHC-eligible defendant, an SSU case manager is expected to refer that person to MHC shortly thereafter.

The study MHC follows the basic procedures of most MHCs that distinguish them from TCCs (2,5). The judge engages defendants in ensuring that they understand the process, especially its voluntary nature and the treatment and behavior mandates, and in discussing their progress or lack thereof. The judge, as well as the MHC team, is supportive, encourages behavioral changes, and focuses on problem solving rather than on offenses and adjudication of guilt (9,12,21,24,26).

Controlling for possible confounders, we compared recidivism over two years between MHC participants after they left the MHC program and MHC-eligible defendants in TCC who were also under SSU supervision on pretrial release and who received individualized plans from the same package of services and supervision from the same pretrial services and community agencies. We hypothesized that MHC participants would have fewer arrests in the two years postexit than they did in the two years prior to their key arrest and that they would have lower recidivism than the comparison group. We also hypothesized that MHC completers (graduates who fulfilled their individualized mandates) would have lower recidivism than noncompleters (those ejected for noncompliance with court mandates or who chose to opt out) because completers get a “full dose” of MHC (2).

We used administrative data from the pretrial services agency for MHC participants (N=408) and MHC-eligible defendants (N=687). The comparison group was defined as all misdemeanor arrestees who screened positive for severe mental illness, who did not have a recent (current or past five years) violent or dangerous felony charge or current domestic violence charge, who were released pretrial to SSU during the first two years of the MHC’s operation (October 2007–November 2009), whose cases were not disposed of by another diversion court, and who had complete data. Because of data availability, our violence and dangerous criterion for the comparison group was broader than that for the MHC participants (charge rather than conviction), yielding a more conservative estimate of recidivism for the TCC sample. From the MHC sample, we excluded those sent back to TCC at their first MHC hearing (N=37); those who did not attend their first MHC hearing (N=3); and those who had administrative case closures because of death, sickness, or other reasons (N=6) (final sample N=408). Data included sociodemographic, court supervision, and arrest information (in the District of Columbia) from two years before the key arrest that brought each person into the sample through two years after disposition of the key arrest in the MHC or the TCC. MHC judges’ records on processing participants supplemented these data. The institutional review boards of the two universities ((North Carolina State University and Georgetown University) as well as the pretrial services agency and the MHC judges approved the research.

The dependent variable was postexit recidivism, measured over the two years after the key arrest disposition (for MHC completers, the key arrest disposition occurred at exit from the MHC). The four measures were any rearrest, number of rearrests, any felony arrests, and time to rearrest. We excluded from postexit recidivism any rearrests that occurred after the key arrest but before court exit or court disposition. These rearrests occurred before our follow-up period; in addition, the MHC in this study often forgives such rearrests as slip-ups. However, we employed these rearrests as predictors of postexit recidivism, because they indicated noncompliance with MHC mandates and continuation of old behavior. We did not count the key arrest among the prior arrests. To predict recidivism in multivariate models that controlled for possible confounders, we employed logistic regression. To predict time to rearrest, the Cox proportional hazards model was employed. Possible confounders included sociodemographic variables (age, gender, and race), illegal drug use (a positive screen for illegal drugs at the first check-in with the case manager), prior arrests (in the two years before the key arrest), and any predisposition arrest (that is, between the key arrest and court disposition).

The MHC and TCC groups were similar in age (MHC, 41.4±10.99; TCC, 40.7±11.63). The TCC group had a larger proportion of men, compared with the MHC group (N=435, 63%; and N=204, 50%; χ²=18.69, df=1, p=.001). Among MHC participants, 90% (N=367) were African American, and 10% (N=41) were white. Among TCC participants, 93% (N=636) were African American, 7% (N=48) were white, <1% (N=2) were Asian, <1% (N=1) was coded as other. The groups differed in prior criminal offending. A larger proportion of the TCC group tested positive for illegal drug use at first check-in (N=439, 64%, compared with N=220, 54%; χ²=10.64, df=1, p=.001). Also, as shown in Table 1, the TCC group had more arrests than the MHC group in the two years before the key arrest (2.32 versus 1.76; t=5.31, df=1,093, p=.001).

Among MHC participants, completers (N=238) and noncompleters (N=170) were similar in sex and age, but noncompleters differed significantly by race. Among noncompleters, the proportion of whites was lower than among completers (N=7, 4%, compared with N=34, 14%; χ²=11.34, df=1, p=.001). In addition, a larger proportion of noncompleters tested positive for illegal drug use at first check-in (N=131, 77%, compared with N=89, 37%; χ²=62.79, df=1, p=.001). As shown in Table 1, noncompleters also had more arrests than completers in the two years before the key arrest (1.92 versus l.64, t=2.19, df=406, p=.033). A larger proportion of noncompleters than completers had a drug offense as their key arrest charge (N=79, 47%, compared with N=85, 36%; χ²=4.77, df=1, p=.029).

Table 1 presents a pre- and postarrest comparison for the TCC group and MHC completers and noncompleters. In each group, most participants were not rearrested. However, MHC participants were significantly less likely than TCC participants to be rearrested in the follow-up period (38% versus 48%; χ²=10.99, df=1, p=.001). The mean number of rearrests was significantly lower in the MHC group compared with the TCC group (.75 versus 1.16; t=4.27, df=1,093, p=.001). Most rearrests were misdemeanors; less than 15% of either group was arrested for a felony, and the difference between the groups was nonsignificant (not shown).

In all groups, a significant decline in the average number of arrests was noted—from over two in the two years before the key arrest to less than one in the two years after court exit or court disposition. At the two-year follow-up, the proportion rearrested was smallest among MHC completers (25%), and this group also averaged the lowest number of rearrests (.42), whereas noncompleters had the largest proportion of rearrested participants (55%) and the highest mean number of rearrests (1.21). These findings indicate that the completers were responsible for MHC participants’ having a significantly smaller mean number of arrests than the comparison group.

Because of differences between samples, we controlled for possible confounding variables with logistic regression to determine whether MHC participants had lower two-year recidivism rates than the TCC group. We first modeled the effect of MHC participation on odds of rearrest and then sequentially added three sets of possible confounders: sociodemographic variables of age, gender, and race; criminogenic factors of illegal drug use and number of prior arrests; and the process variable of any predisposition arrest. Models comparing the full MHC and TCC samples (not shown) indicated that when the models did not control for confounding variables, the MHC participants were less likely (odds ratio [OR]=.66, p<.001) than the TCC sample to be rearrested in the two years after court exit or court disposition. When the models controlled for sociodemographic variables, the MHC participants continued to have lower odds of rearrest (OR=.69, p<.004). However, MHC participation lost significance as a factor in rearrest in subsequent models that added illegal drug use, number of prior arrests, and any arrest between key arrest and court disposition.

Table 2 presents the same logistic regression analyses but with MHC participants separated into completers and noncompleters (the reference group is the TCC group). The first model shows that MHC completion made an even larger difference than MHC participation in lowering the odds of arrest after court exit. Compared with the TCC sample, completers were approximately two-thirds less likely (OR=.37) to be rearrested, whereas rearrest among noncompleters was not significantly different compared with TCC defendants. In each subsequent model, MHC completion maintained its significant effect. By the final model with all confounders included, completers were significantly less likely (OR=.52) than the TCC sample to be rearrested. Older participants and female participants were significantly less likely than younger and male participants to be rearrested, and each arrest in the year before the key arrest and any predisposition arrest increased the likelihood of rearrest. Illegal drug use was not a significant predictor of rearrest. Noncompletion of the MHC program became a significant predictor of rearrest in the third and fourth models, increasing the likelihood of arrest over that of the TCC group.

We found similar results explaining differences in time to rearrest after court exit or court disposition. Controlling for predictor variables from the final logistic regression models, Cox regression analysis indicated that MHC participation alone was not significant but that completion was significant (hazard ratio=.58, p=.001). Figure 1 shows the estimated cumulative probability of rearrest at specified intervals for MHC completers and noncompleters and the TCC group. MHC completers had the longest time to rearrest, whereas the TCC group had a significantly shorter time and MHC noncompleters had the shortest. The distance between curves is the estimated postdisposition positive effect of MHC completion at any time point; thus, at 18 months, for example, 20% of completers had been arrested, compared with 47% of noncompleters and 40% of the TCC group.