A Randomized Controlled Trial of the Rochester Forensic Assertive Community Treatment Model

The study took place at two primary sites, an academic medical center and a criminal court, in Monroe County, New York. All clinical interventions were based at Strong Ties Clinic, an outpatient program of the University of Rochester Medical Center (URMC) Department of Psychiatry. The clinic offers mental health services, primary medical care, and pharmacy services to all patients. Study recruitment was conducted at the Monroe County Courthouse, a public facility that houses Rochester City Court and attorney offices. Administrative approval was received from the New York State Office of Mental Health in November 2010 to begin initiating the research program at Strong Ties. The study was approved by the URMC Research Subject Review Board.

A parallel-group randomized controlled design was used to assess service utilization outcomes over a one-year study period. Study inclusion criteria were ages 18 and older, presence of a DSM-IV-TR psychotic disorder, ability to speak English, adequate capacity to consent to research, recent conviction on a misdemeanor charge, and eligibility for a conditional discharge. All participants entered the study under a conditional discharge status, whereby their pre-enrollment sentences were suspended pending successful compliance with legal stipulations that included accepting mental health treatment and avoiding further criminal activity. Individuals involved in probation, parole, mental health court, assisted outpatient treatment, or other forms of legal leverage at the time of recruitment were excluded to enable comparison of leveraged (FACT) and nonleveraged treatment groups. Individuals facing felony charges were also excluded, but those with prior felony convictions were eligible for enrollment.

A three-step process was utilized whereby potential study participants were first identified with assistance from the Public Defender’s office. Those who expressed interest in the study after pleading guilty and accepting a conditional discharge subsequently met with a research team member to provide informed consent in the presence of their defense attorneys. All participants consented to both the treatment intervention and the data collection arms of the study (26). Consenting individuals were then randomly assigned to a treatment group by using computer-generated assignment cards within a courtroom setting. [Participant identification, recruitment, and randomization are described in detail in an online supplement to this article.]

Following randomization, participants were assessed at baseline. All baseline data were collected by the study health project coordinator, a research psychologist with ACT team experience. Baseline demographic data were obtained through face-to-face interviews and review of health records, and psychiatric diagnosis was confirmed by using the Structured Clinical Interview for DSM-IV-TR (27). Baseline clinical assessments included the Brief Psychiatric Rating Scale (BPRS) to assess symptoms (28), the Insight and Treatment Attitudes Questionnaire (ITAQ) to assess insight into need for treatment (29), the Medication Adherence Rating Scale (MARS) to assess medication adherence (30), the Addiction Severity Index (ASI) to assess severity of addiction (31), the Treatment Self-Regulation Questionnaire (TSRQ) to assess motivation for treatment (32), and the MacArthur Community Violence Inventory (MCVI) to assess involvement in violence as a victim or a perpetrator (33). To control for baseline recidivism risk, participants were also assessed with the Level of Service Inventory–Revised (LSI-R), an assessment tool shown to predict risk of reoffending (34).

The New York State Division of Criminal Justice Services (NYS DCJS) provided statewide criminal justice service utilization data for all participants, including arrest, conviction, and sentencing data (37). Incarceration data on all participants was obtained from the Monroe County Sherriff’s Office, which oversees the 1,475-bed Monroe County jail (38). Mental health service utilization data were obtained from the Monroe County Mental Health Service (MCMHS) database (39). MCMHS is a countywide system that spans all publicly funded mental health agencies in Monroe County. Data provided by MCMHS contained psychiatric hospitalization, emergency room, and outpatient service use data for all participants. All data sets were provided in deidentified electronic formats with common participant identification numbers that enabled linking for data analysis.

A Poisson regression model was used to detect changes in jail time, the primary outcome measure, between the two treatment groups. The study sample size allowed detection of an 11% or greater change in the average number of jail days during a one-year follow-up period with 80% power and a type 1 error rate of 5% for a two-sided hypothesis test. The observed effect size substantially exceeded 11%, indicating sufficient power to detect such an outcome. Baseline categorical variables are presented as summary counts and percentages, and differences between the baseline variables of the treatment groups were analyzed by either Fisher’s exact test for 2×2 matrices or Pearson’s chi-square test with Yates’ continuity correction. For continuous variables, the means, standard deviations, and p values resulting from Wilcoxon’s rank sum test with a continuity correction are presented. In cases where variables contained one or more missing values, the number of participants with complete data is presented. Analyses examining the association between baseline variables and failed conditional discharge status are summarized. Participants with a failed conditional discharge were removed from the intervention arm of the study protocol for failing to comply with the judge’s legal conditions but were continued in the data collection arm. These univariate analyses were conducted by using logistic regression analyses, with the reported p values coming from a likelihood ratio test of whether the baseline variable had an effect (versus no effect) on the odds of failed conditional discharge status. These analyses did not account for possible differences in duration of conditional discharge status, only whether or not a failed conditional discharge occurred during the study period.

Analyses were also performed to test the study hypotheses that FACT would have a greater effect than enhanced treatment as usual in preventing jail and psychiatric hospital use and in promoting outpatient mental health service use. The effect of FACT upon service utilization outcomes is summarized through regression coefficients, p values, and confidence intervals. Because each outcome measure could be viewed as a count, a negative binomial regression model was used to conduct these analyses. Analyses examined full one-year outcomes for all participants, including those who exited the study intervention arm early, using a constant offset equal to the natural logarithm of 365 days (“intent-to-treat” analyses). Additional analyses were conducted to examine outcomes only while participants were enrolled in the intervention arm of the study protocol by using an offset equal to the natural logarithm of days in protocol (“per-protocol” analyses). The p values for both intent-to-treat and per-protocol analyses are presented. Also, an exploratory analysis was conducted by using stepwise backward elimination to identify baseline variables most predictive of study outcomes (40) [see supplement for details of this analysis].

All recruitment and study intervention activities were conducted between February 10, 2011, and May 14, 2014. Seventy participants were randomly assigned to receive either FACT (N=35) or enhanced treatment as usual (N=35) for one year. Forty-nine participants (70%) remained in the study intervention arm for one full year. Nineteen participants (27%), including nine from the FACT group and ten from the control group, were removed from the intervention arm by the judge for failing to comply with the conditions of discharge (that is, for demonstrating continuous treatment nonadherence combined with continuous criminal activity). One additional participant was removed from the control group by the county mental health director after the participant assaulted a nurse and a security guard while hospitalized. Individuals who failed to meet the terms of the conditional discharge were required to serve their original sentences, had no further judicial oversight, and became eligible for treatments other than the study treatment as determined by their treating clinicians. One FACT group participant died of medical causes unrelated to study participation. [A CONSORT diagram illustrating the flow of participants through the study is available in the online supplement.]

For all 70 study participants, the mean±SD number of days in the intervention arm of the study protocol was 329.7±67.0. The mean number of days in the intervention arm of the protocol was 326.8±78.6 for the FACT group and 332.5±53.9 for the control group, a nonsignificant difference.

Descriptive analyses of study variables, both overall and by intervention group, are shown in Table 1. Study participants had considerable impairment and a significant degree of criminal justice system involvement at baseline. Participants were predominantly male, African American, never married, and unemployed, and nearly half had not graduated from high school. The most common diagnosis was schizophrenia, and 70% of participants self-reported having a co-occurring substance use disorder. On average, participants spent over two months in jail during the year before enrollment and had over 16 lifetime arrests. Also, 30% of those interviewed with the MCVI (N=17 of 56) reported perpetrating violence toward others that resulted in injury during the six months before enrollment. Mean total BPRS scores indicated that participants had a moderate level of symptomatology, and mean total ITAQ scores suggested that they had only partial insight into their illnesses and the need for treatment. Mean total MARS scores showed that participants had low levels of medication adherence at baseline. Mean ASI scores suggested that overall addiction severity was in the low range. LSI-R scores indicated that participants had a moderate risk of criminal recidivism. Mean TSRQ scores showed that after randomization, there was a trend for participants in the control group to feel greater autonomy in deciding whether to take antipsychotic medications compared with participants in the FACT group. There were no significant differences between study groups in any baseline measure.

Baseline characteristics of participants who received a failed conditional discharges and those who did not were compared. As shown in Table 2, individuals with failed conditional discharges had higher LSI-R scores and more severe psychiatric symptoms at baseline compared with individuals with successful conditional discharges. There was also a trend among the participants with failed conditional discharges to have been sentenced to a higher number of lifetime jail days at baseline than participants with successful conditional discharges (p=.060).

State and county databases enabled both intent-to-treat and per-protocol analyses of service utilization outcomes. As shown in Table 3, these analyses produced very similar results. FACT group participants had fewer criminal convictions and spent fewer days in jail than control group participants. FACT group participants also spent more time in outpatient treatment and had more outpatient service contacts compared with control group recipients. In addition, FACT participants had less use of inpatient psychiatric services compared with control group participants. No significant differences were observed between the groups in numbers of emergency room visits, arrests, or incarcerations. However, approximately one-third of all FACT group incarcerations were due to judicial sanctions for noncompliance rather than new crimes. Analyses of incarceration data with and without sanction incarcerations are included in Table 3. Also, one FACT group participant had 50 emergency room visits, more than all other FACT participants combined. With this outlier removed, FACT group participants had significantly fewer emergency room visits compared with control group participants (.8±1.6 and 1.9±2.7, respectively; coefficient=–.93, standard error=.43, p=.030).