Gilbert and colleagues (2010) found that the odds of arrest in any given month for participants who were currently receiving Assisted Outpatient Treatment (AOT) were significantly lower than the odds for participants in the pre–AOT and prevoluntary agreement group (the reference group). The odds of arrest were nearly two thirds lower for participants currently receiving AOT, compared with the odds of arrest for the reference group.
However, there were no statistically significant differences in the odds of arrest between those with a current voluntary agreement and those in the reference group. The adjusted predicted probabilities of arrest in any given month were 3.7 percent for the reference group, 2.8 percent for individuals currently under a voluntary agreement and 1.9 percent for individuals currently on AOT.
Within-group analyses conducted by Link and colleagues (2011) showed that the risk of arrest was significantly higher for individuals during the period before Assisted Outpatient Treatment (AOT) than during the period of AOT. Though the risk of arrest went up slightly in the period after AOT was discontinued, this difference was not significant. For an individual who had ever received AOT, the risk of any arrest was 2.66 times greater before AOT as it was while receiving AOT.
The between-group results showed that the risk of arrest among individuals in the comparison group who were never assigned to AOT was significantly higher than the risk of arrest for the AOT group while they were assigned to AOT. Compared with individuals during and shortly after the period of assignment to AOT, the comparison group who never received AOT had nearly double the odds of arrest.
Arrests for Violent Offenses
Within-group analyses found that individuals receiving AOT were at significantly lower risk of arrest for a violent offense than they were before AOT. The risk of arrest for a violent offense was 8.61 times greater before AOT as it was while receiving AOT. However, because arrests for violent offenses were relatively rare, between-group analyses found there were no significant differences between the AOT and comparison groups on the odds of arrest for violent offenses.
An initial analysis performed by Swanson and colleagues (2000) found there was no significant difference in the rate of violence between the group randomly assigned to involuntary outpatient commitment (OPC) and the control group (32.3 percent in the OPC group versus 36.8 percent in the control group).
However, multivariate analysis showed that controlling for baseline history of violence and substance misuse, extended OPC was associated with significantly lower odds of any violent behavior during the year of the study. Treatment group members who received more than 180 days of OPC were only about one third as likely to commit a violent act during the year, compared with their control group counterparts. However, treatment group members receiving fewer than 180 days of OPC did not differ from the control group with respect to risk of violence.
Extended Outpatient Commitment
This initial analysis did not include the seriously violent group, nor was the length of exposure considered. When study participants with a history of serious violence were included in the analysis and the OPC intervention was defined as having received at least 6 months of court-ordered treatment, the treatment group had a significantly lower rate of violence during the year, compared with the control group (26.7 percent, versus 41.6 percent). This result should be viewed with caution, because the analysis included participants with a history of serious violent behavior who were not randomized to treatment.
Extended Outpatient Commitment Combined With Regular Community-Based Services
Additional analysis looked at whether OPC interacts with the provision of outpatient services to reduce the risk of violent behavior. An initial analysis found that OPC alone did not significantly reduce the risk of violent behavior. Similarly, receiving frequent outpatient services alone was not associated with less violence. However, a combination of both variables (at least 6 months of OPC with an average of three or more outpatient visits per month in the community) did significantly reduce the risk of violence. The predicted probability of any violent behavior was cut in half, from 48 percent to 24 percent, attributable to extended OPC and regular outpatient services. Again, this result should be viewed with caution, because the amount of time on OPC was neither random nor controlled for experimentally.
An evaluation of New York State’s Assisted Outpatient Treatment (AOT) by Gilbert and colleagues (2010) examined whether individuals had lower arrest rates when receiving AOT or voluntary enhanced services than before initiating either one. Study participants were sampled from AOT program rosters of mental health service recipients in six New York counties. Based on structured interviews, 181 individuals who had either received court-mandated AOT (n=139) or had signed a voluntary service agreement (n=42) at some point during the study period were identified for the study. The sample was 60 percent male, with an average age of about 34. The sample was approximately 46 percent white, 35 percent African American, 11 percent Hispanic, and 7 percent Asian/Pacific Islander or other. There were no significant differences between the groups on age, gender, race or ethnicity, and education. Individuals with voluntary agreements were more likely to have a primary diagnosis of major depressive disorder (17 percent), compared with individuals who received AOT (4 percent), and were also more likely to reside in regions outside New York City (98 percent of the voluntary agreement group, compared with 60 percent of the AOT group).
The primary outcome of interest was arrest rates. Arrest records for the 181 participants from Nov. 1, 1999, to Feb. 28, 2008, were obtained from the New York State Division of Criminal Justice Services. Demographic information was collected during the interviews, and diagnostic data was extracted from matched Medicaid claims files. The primary diagnosis was classified into four categories: schizophrenia, bipolar disorder, major depressive disorder, and other.
A dummy variable was created to indicate whether the study participants had been arrested during any given month. Analytic comparison groups of person-month observations were constructed to reflect status in regard to AOT or a voluntary service agreement. A single category of observations (pre–AOT and prevoluntary service agreement) was used as the reference group for comparison. This category included all person-months before an individual either received AOT or signed a voluntary agreement. This created five groups of observations that were compared: 1) pre–AOT and prevoluntary agreement, 2) current AOT, 3) current voluntary agreement, 4) post–AOT, and 5) postvoluntary agreement.
The data was structured as one unit of observation per person, per month, over the 100-month study period. There were 9,229 person-months available for analysis: pre–AOT and prevoluntary agreement (n=7,097), current AOT (n=1,094), current voluntary agreement (n=468), post–AOT (n=361), and postvoluntary agreement (n=209).
Data was examined using estimated repeated-measures, multivariable logistic regression models to predict the odds of arrest by AOT and voluntary agreement status, with the models controlling for time and nonindependence of observations. The multivariate analysis also controlled for region, race and ethnicity, age, and sex.
Limitations of the study include a small sample size and nonrandomized study groups.
The 2011 evaluation by Link and colleagues examined whether Assisted Outpatient Treatment (AOT) under New York’s “Kendra’s Law” was associated with reduced arrests for violent and nonviolent offenses. The study employed a quasi-experimental design that compared the arrest rates of 183 individuals—86 who were assigned to AOT at some point in their lives and a comparison group of 97 who were never assigned to AOT. Participants were recruited from outpatient clinics in the New York City boroughs of the Bronx and Queens. After a complete description of the current study, written informed consent was obtained from each participant to conduct searches of arrest records.
There were no significant differences between the groups, except that there were significantly more men in the AOT group than in the comparison group (67 percent versus 54 percent). The AOT group was 57 percent African American, 27 percent Hispanic, and 16 percent white or other, with an average age of 36½. The comparison group that never received AOT was 51 percent African American, 31 percent Hispanic, and 19 percent white or other, with an average age of 37½.
The outcome of interest was arrest rates. Official arrest data, including the date of each offense and the associated charge, was collected from the New York State Department of Criminal Justice Services for each participant from the date he or she turned 18 until Jan. 1, 2007. A data file was constructed indicating whether a study participant was arrested in each month of observation. In addition to analyzing arrests for any offense, arrests were categorized according to whether the charge was for a violent offense (including murder, nonnegligent manslaughter, forcible rape, robbery, and aggravated assault).
In addition, New York State Office of Mental Health records were used to accurately identify periods in which individuals were assigned to AOT. For individuals ever assigned to AOT, three time periods were constructed: before AOT, during and 6 months after AOT, and more than 6 months after AOT ended. Logistic regression estimated by generalized estimating equations (GEE) was used to examine the odds of arrest. Fixed-effects logistic regression analyses were also conducted, but only for all arrests combined because arrests for violent offenses were too rare to produce stable estimates. The study included a comparison of the estimates of the effect of AOT on arrests from the fixed-effects and GEE analyses to determine whether the conclusion about the effect of AOT was the same in these complementary analytic approaches.
Swanson and colleagues (2000) employed a 1-year randomized trial to examine the effectiveness of involuntary outpatient commitment (OPC) combined with case management to reduce the incidence of violence among people with severe mental illness in North Carolina. Study participants were screened from a population of involuntarily hospitalized patients who had been court-ordered to undergo a period of OPC upon discharge.
Upon discharge from the hospital, 262 patients were randomly assigned to release or court-ordered treatment. Study participants in the experimental group received an initial period of OPC no longer than 90 days. After that, commitment orders could be renewed for up to 180 days if a psychiatrist and the court determined that the patient continued to meet the legal criteria for OPC. Participants in the control group received immunity from any OPC during the year of the study. All study participants received case management and other outpatient treatment at mental health programs in the area. An exception to the randomization procedures was made in cases where patients had a history of serious assault involving weapon use or physical injury to another person within the past year. This group (the seriously violent group) was required to undergo at least the initial period of OPC as ordered, but they were not randomly assigned to the group.
Of the 262 patients, 102 were randomly assigned to the OPC group, 114 were randomly assigned to the control group, and 46 were assigned to the seriously violent group. A little more than half of the study participants (53.4 percent) were male and 18–39 years old (52.7 percent). The majority of participants (66 percent) were African American, while the rest were non-Hispanic white (33 percent) or other (1 percent). Most of the sample had diagnoses of psychotic disorder (68 percent), 28 percent had bipolar disorder, and 4 percent had recurrent major depression. There were no significant differences between the groups on demographics and other characteristics.
Data was collected from structured interviews with study participants as well as family members, others who knew the participant well, and case managers. Data was also obtained from hospital records and outpatient service records. The outcome measures of interest included violence, psychiatric symptoms, functional impairment, insight, social support, substance use and misuse, medication adherence, and outpatient services utilization. Violent incidences included whether study participants had been picked up by police or arrested for physical assault on another person, had been in fights involving physical contact, or had threatened someone with a weapon.
Logistic regression was used to examine the relative impact of sustained OPC and outpatient services utilization on the incidence of violent behavior, controlling for baseline violence history and relevant covariates.
There were some limitations to the study. The study design deviated from a strict randomized controlled trial in two ways. First, the sample included a subgroup of study participants with a recent history of seriously violent behavior who could not be randomly assigned to the control group. Second, the amount of time on OPC was neither random nor controlled experimentally. The amount of time a study participant remained in OPC following the required 90 days of treatment varied depending on how the clinicians applied the legal criteria for renewal of OPC orders.