Clinical Outcomes of Specialized Treatment Units for Patients With Serious Mental Illness in the New York City Jail System

Jails and prisons in the United States are now responsible for caring for large numbers of individuals with serious mental illness. Estimated at 15%−20% of the incarcerated population (1, 2), the rate of serious mental illness is many times higher than that in community samples (3, 4). Several factors have contributed to this intersection of mental illness and criminal justice involvement. These include drug laws disproportionately affecting those with comorbid substance use and mental disorders, homelessness and shortages of affordable housing, insufficient funding for community treatment, and deinstitutionalization (the closing of public psychiatric hospitals that began in 1955 and accelerated in the late 1960s), paired with civil commitment statutes requiring a person to meet the criterion of dangerousness for involuntary or emergency psychiatric hospitalization. This combination of factors has led to reduced access to hospital beds (5–9).

U.S. jails receive over ten million admissions annually (10). Correctional health systems shoulder the enormous responsibility of initial identification of health and mental health needs of persons admitted to their care. In correctional settings, research on the prevalence of mental illness and related outcomes has found that those with mental illness are at elevated risk of violent victimization, self-injurious behavior, and clinical destabilization, resulting from psychiatric medication nonadherence in punitive environments, where psychiatric care is generally considered to be below community standards (11–16). Approximately 16% of the average daily New York City (NYC) jail population is diagnosed as having serious mental illness (unpublished data, NYC Correctional Health Services, March 2019), defined as having one or more diagnoses in the following DSM-5 categories: schizophrenia spectrum and other psychotic disorders, bipolar spectrum disorders, depressive disorders, or posttraumatic stress disorder (PTSD). Any DSM-5 disorder that causes significant clinical distress to the point of functional impairment may also qualify as a serious mental illness.

To minimize the considerable health and safety risks to this population, roughly half of the 11 jails in NYC’s jail system have maintained over 20 mental observation units—specialized housing areas for patients with serious mental illness—for the past 30 years. Early attempts at improving mental health care and decreasing physical harm on these units found that increased clinical programming reduced injuries related to custodial uses of force, self-injuries, and need for suicide watch (17). Despite these efforts, patients were frequently and unpredictably moved on and off units, and consistency in both health and custody staffing assignments remained challenging. Continuity of care was limited, and the physical space—loud and crowded, with little natural light—was not conducive to high-quality psychiatric care.

As part of larger reforms to jail-based health care delivery that began in 2014, NYC funded an interagency collaboration between its Department of Health and Mental Hygiene, then the oversight agency for the jail health care contract, and Department of Correction to reform care and treatment for patients with serious mental illness. In 2015, NYC moved the management of jail health services to a direct care model under a newly created division of Correctional Health Services within its public health care system, NYC Health + Hospitals. Existing jail mental observation units (single cell housing, as is all cell housing in the jail system) were renovated to become PACE (Program for Accelerating Clinical Effectiveness) units, designed to have large, open spaces, confidential interview rooms, adequate space for protected group activities, staff offices, and as much natural light as possible. The location of each PACE unit within a specific jail was based on multiple factors, including the experience of the jail’s correctional staff in working with individuals with serious mental illness; the jail’s ability to efficiently access treatment for comorbid conditions, such as substance use disorders and chronic medical conditions; and the ability to renovate the selected unit within budgetary constraints. Multidisciplinary mental health treatment teams, including a psychologist leader, psychiatric providers, nurses, counselors, treatment aides, and art therapists, were established. Correctional officers participating with health personnel in 40 hours of specialized mental health training were selected to staff these units and be part of the treatment teams. Mechanisms for clinical staff to communicate relevant information to security staff while respecting patient confidentiality were created.

Each PACE unit has a primary goal of maintaining or restoring, as best as possible in jail, clinical stability for specific subpopulations of individuals with serious mental illness. The first PACE unit, opened in January 2015, is a “hospital step-down” unit for males returning from inpatient psychiatric hospitalization and is intended to reduce additional hospitalizations. A second unit opened in February 2015 as an “acute care” unit for men at high risk of clinical decompensation in jail (e.g., through psychiatric medication nonadherence, social isolation, or poor maintenance of daily living activities) and is intended to prevent hospitalizations. The third opened in June 2015 as a “psychiatric assessment” unit for evaluation of patients with complicated or unclear diagnoses. After several years noting that patients admitted to this unit were frequently diagnosed as having previously unrecognized neurodevelopmental disorders, the unit transitioned into care specifically for that population. The fourth unit, opened in September 2016, focuses on competency maintenance, with the purpose of maintaining clinical stability for men returning from a state hospital after being found not competent to stand trial. Additional units have opened as of May 2019—one for women and another supporting community reentry for men serving misdemeanor sentences.

Core components of the PACE philosophy include patient-centered crisis deescalation, incentives programming emphasizing positive reinforcement over punishment, and the belief that patients should be active and responsible members of their treatment teams. In addition, because staff are consistent, they are able to develop strong therapeutic relationships and robust treatment plans with patients. Group activities occur throughout the day, including community meetings, creative arts therapy, and discussion groups. Medication management focuses on patient engagement rather than coercion. The jails do not utilize treatment over objection for nonemergency psychiatric purposes. Individuals who require that level of psychiatric intervention are referred for admission to one of two public NYC hospitals with inpatient psychiatric units for individuals in Department of Correction custody.

Referrals to PACE units are made exclusively by Correctional Health Services mental health or hospital clinical staff, and all referrals are then screened for admission criteria by a supervising psychologist. Variables that typically affect jail housing decisions (e.g., criminal charge, age, security classification, gang membership, and prior violence) are rarely compelling enough to override the clinical recommendation for PACE. In the absence of extremely dangerous security situations, Department of Correction allows mixed security classification housing for individuals with significant health needs. The primary reason for not admitting a patient to PACE is bed availability. At any time, there are waiting lists for the PACE units.

We conducted a retrospective, observational cohort study to evaluate the impact of the first four PACE units on clinical and safety outcomes for individuals with serious mental illness. We compared PACE patients to those with similar characteristics who, because of limited PACE bed availability, were placed in other mental observation units or other housing during the same period. Our hypotheses were that the care delivered to patients on the PACE units would result in decreased rates of self-injury and injuries due to violence, specifically fights, and increased psychiatric medication adherence, which has been shown to be associated with improved clinical stability (18), compared with patients in a control group. Impact on psychiatric hospitalization, an important outcome, will be described in a subsequent report, given the complexity of the analyses involved.

Methods

Inclusion Criteria

We considered jail admissions of male adults (ages ≥18) who were diagnosed as having serious mental illness and in the jail census for ≥14 days between January 1, 2016, and March 31, 2018. The time frame allowed for each new PACE unit to operate for ≥6 months before being assessed and for ≥1 year of assessment. We selected 14 days as the minimum time needed for treatment to have a potential effect on outcomes of interest. PACE patients included those housed in one or more of the four PACE units for ≥14 days during this period and whose qualifying PACE admission was the first during their incarceration. Of 655 patients admitted to PACE during the time period, 302 met the ≥14-day and first-admission criteria. Individuals for the control group were identified from 4,740 unique patients with serious mental illness whose first admission to jail during the study period was ≥14 days and who were not admitted to PACE because of limited bed availability. Most of these patients were housed in facilities that had PACE units or similar population and housing profiles, including mental observation units.

Outcomes

Self-injury was defined as a documented incident of deliberately causing or attempting to cause harm to oneself (e.g., cutting, hanging, foreign body ingestion, and head banging). Self-injury data, including date and time, location, and nature of the injury, were obtained from Correctional Health Services’ database of clinician-verified incidents. Injuries due to violence were defined as any physical injury caused by an act of physical violence between two or more patients and documented in the electronic medical record (EMR). Data regarding adherence to psychiatric medications were extracted from the EMR and separate pharmacy and nursing records. Adherence was measured by dividing total number of doses of psychiatric medication received by total number of doses prescribed for each patient. For PACE patients, only outcomes that occurred while the patients were housed on PACE units were considered.

Analysis

We used propensity score matching (PSM) (19) to select the final control cohort. Demographic, health, and incarceration-level covariates with largely complete information were extracted from the EMR and used to match patients in the PACE group with potential control patients on these variables. Psychiatric diagnoses were categorized on the basis of either DSM-IV (20) or DSM-5 (21) classification, depending on when the patients were incarcerated. The criminal charge of highest severity (e.g., New York State Penal Law–defined violent felony, nonviolent felony, misdemeanor [22]) on the incarceration constituted the “charge” variable.

We used the PSMATCH procedure in SAS to calculate the propensity score, the predicted probability of each patient’s being placed in the PACE units based on the selected covariates. One-to-one matching of all PACE patients to control patients on the scores was completed by using an optimal full-model approach. The underlying assumption of PSM is that matching based on observed covariates removes any systematic differences between nonrandomized treatment and control groups.

Balance diagnostics were assessed on the absolute standardized mean differences for each covariate, and characteristics of the matched groups were compared by using bivariate statistics. The observation period for the PACE cohort began at admission to PACE. The observation period for each control patient began at the point in his jail admission when his matched control entered PACE, thereby ensuring equal jail exposure prior to the observation period. For a small number of matched sets where this was not feasible, the date of the first comprehensive mental health evaluation was used as the starting point for control patients.

We conducted descriptive analyses using frequencies with percentages for categorical variables and medians with interquartile range (IQR) for nonnormally distributed continuous variables. PACE length of stay (LOS) represented the consecutive number of days spent on a PACE unit for the patient’s qualifying admission. Chi-square and Wilcoxon signed-rank tests were employed to determine group differences.

Rates of self-injury and injury due to fights were calculated in 100 person-days, using the observation period length. Incident rate ratios with 95% confidence intervals (CIs) comparing PACE patients with control patients were calculated by using Poisson regression. Analyses of self-injury and injuries due to violence were conducted over 30- and 60-day intervals to account for shorter and varying incarceration LOS. Differences in psychiatric medication adherence over a maximum of 30 days, a routine Correctional Health Services metric, were assessed with t tests. Rates for patients discharged from PACE (for PACE patients) or from jail (for control patients) before the 30- and 60-day endpoints were presumed consistent with what they would have been had they had the requisite follow-up time to endpoints. All patients had at least 14 days of follow-up. Analyses were conducted with SAS and SPSS, version 24. Two-sided tests were significant at p<0.05. The study was approved by the BRANY Institutional Review Board with waiver of informed consent.

Results

Patient Characteristics

A total of 302 patients, all male, constituted the PACE cohort (Table 1). The median age was 36 years; those on the competency maintenance unit were significantly older (median=43), compared with patients on other units (range=28–37; p<0.001). Most patients were non-Hispanic black or Hispanic (82%). The most common psychiatric diagnostic category was schizophrenia spectrum and other psychotic disorders (81%); however, the psychiatric assessment unit, designed for individuals with neurodevelopmental disorders, had a significantly lower prevalence (44%) than other units (range=88%–92%; p<0.001). There were also unit differences in the prevalence of depressive disorders; neurodevelopmental disorders; and PTSD, trauma, and other stress-related disorders (p<0.001), although absolute numbers of such diagnoses were small. Sixty-seven percent of patients were diagnosed as having substance use disorders, with no appreciable differences by PACE unit. The most commonly prescribed psychiatric medications across PACE units were long-acting injectable antipsychotic medications (33%), risperidone (38%), and valproic acid (35%).

Sixty-eight percent of PACE patients were charged with a violent felony. Charge profile differed significantly across units (p=0.025), with the psychiatric assessment unit and the competency maintenance unit having greater proportions of patients with a violent felony charge (71% and 83%, respectively). Six percent of patients had a parole violation. Overall median LOS in PACE was 58.5 days; patients in the acute care unit had significantly shorter LOS (38.5 days; p<0.001).

Prior to PSM, PACE and potential control patients (N=4,740) differed significantly on most variables (Table 2). After PSM, PACE patients and matched control patients (N=302 pairs) differed only with respect to clozapine prescription (p<0.001). Patients prescribed clozapine were more likely to be housed on the competency maintenance unit or acute care unit to optimize treatment monitoring. We chose to retain clozapine in the medication adherence analysis, assuming, given the low rates of clozapine prescription, that any contribution of clozapine adherence would not significantly affect results.

Self-Injury

The overall 30- and 60-day rates of self-injury events for study patients were both 0.10 per 100 person-days (Table 3). PACE patients had lower rates of self-injury, compared with control patients, at both 30 and 60 days (PACE patients, 0.08 and 0.08 per 100 person-days, respectively; control patients, 0.11 and 0.13 per 100 person-days, respectively), although the differences did not reach statistical significance.

Injury Due to Violence (Fights)

The overall rate of injuries due to violence was 0.21 per 100 person-days over 30 days and 0.18 per 100 person-days over 60 days (Table 3). PACE patients had between 65% and 70% fewer injuries due to fights at 30 and 60 days, compared with control patients (p<0.001).

Psychiatric Medication Adherence

The overall mean psychiatric medication adherence over 30 days was 66.5% (Table 4). Mean psychiatric medication adherence was 76.7% for all PACE patients and 55.3% for control patients (p<0.001).

Discussion

The PACE program, designed to optimize safety and treatment for patients with serious mental illness in NYC jails, demonstrated significant improvements in psychiatric medication adherence and injuries due to violence. These changes have not only been reflected in a higher quality of mental health care, but they have also contributed to a major cultural shift in the approach to mental health in the jail system. On the basis of anecdotal reports, reductions on these units in injuries due to violence—although not representative of the universe of possible violence (e.g., fighting not resulting in injury and violence involving nonpatients)—have led both health care and correctional staff to believe that interventions aimed at treatment and support are more effective at encouraging prosocial behavior than authoritarian, punitive approaches. We posit that the significant reduction in violent injuries after 60 days indicates that the more patients adapt to a therapeutic, community-oriented approach, the less likely they are to engage in interpersonal violence.

Health care and correctional staff appear to also have shifted their attitudes related to forced medication in the years since the PACE program began. The NYC jail system does not allow nonemergency psychiatric treatment over the individual’s objection. Patients with serious mental illness who are at risk of harm to self or others and who consistently refuse psychiatric medications are referred to the hospital for a petition to the court for an order for forced medication. This order is effective only while the patient is hospitalized. Prior to PACE, when patients returned from the hospital and quickly decompensated because the court order for medication was no longer in effect, robust, although ultimately unsuccessful, administrative efforts were made to implement forced medication in the jails. After the PACE units demonstrated that psychiatric medication adherence could substantially increase through more active engagement with patients about their treatment and more flexible approaches to medication administration (e.g., holding medication if patients initially refuse or are asleep or involving the treatment team in medication administration), efforts to implement forced medication have subsided.

We were surprised to find that observed reductions in the rates of self-injury over 30 and 60 days associated with PACE were not statistically significant, although we found the results encouraging, nonetheless. One explanation could be the long-standing rigorous attention to and surveillance of self-injury, as well as policy revisions and staff training related to that surveillance that were implemented in 2016–2017. These factors may have lessened the difference between the PACE units and other units with respect to self-injury. Further analysis is needed to better understand these findings, especially given the complexities of self-injury in a jail setting, where hurting oneself may be the only behavior that draws attention to the real needs of the patient (23, 24).

Given the health risks associated with incarceration, the PACE units were designed to improve clinical outcomes for patients during incarceration, rather than as a direct attempt to reduce recidivism upon patients return to the community. Although an important goal for the nation is to minimize criminal justice involvement of individuals with serious mental illness, reincarceration is mostly related to issues beyond the control of jail health services, including police practices, criminal justice and community policies, and socioeconomic and criminogenic risk factors. NYC has seen marked declines in jail admissions and the average daily jail population in recent years, which has been attributed to falling rates of crimes, arrests, and prosecutions; increases in alternatives-to-incarceration programming; and bail reform legislation. These changes may affect arrest and incarceration differently depending on the individuals’ defining characteristics. We, as the correctional health entity, have limited access to legal outcomes or criminal case histories that would allow us to assess and control for factors that may drive recidivism.

The financial investment in each PACE unit is considerable and includes both health care and custodial personnel costs, as well as capital investment in environmental improvements. We suspect that the reductions in injuries are associated with longitudinal cost savings for medical care (and custodial costs, such as transportation to hospitals) and clear health benefits to patients. Improvements in psychiatric medication adherence can reasonably be expected to be associated with improved clinical stability and thus with reduced need for costly psychiatric hospitalization. We anticipate that a forthcoming report related to the impact of PACE on psychiatric hospitalization will provide additional evidence for cost savings. Finally, and perhaps more important given the health risks of jail, improved clinical stability related to psychiatric medication adherence for patients with serious mental illness may reduce LOS in jail by reducing delays associated with competence restoration and additional charges filed for jail violence (25). Further research is needed to validate this assumption.

We believe that the results show that the PACE program demonstrates health benefits for patients with serious mental illness in jail. This assessment did have some limitations. First, the specificity of the patient sample and structure of the NYC jail system limit the generalizability of the findings to other jail settings. However, there are individuals with serious mental illness in most, if not all, jails in this country. Diagnoses, including the determination of serious mental illness, were based on clinical interviews, available collateral information, and observable behavior; structured diagnostic tools were not utilized. Regardless, the concepts and practices of PACE related to persons with serious mental illness, however defined, are almost certainly translatable elsewhere.

Next, although PSM is a robust procedure that matches cases to controls on a substantial number of observed variables, we could not control for variables that were not measured. For example, we could not systematically assess acute psychiatric symptoms or decompensation, which often precipitate a PACE referral. It is possible that control patients were not as symptomatic as PACE patients during the observation period, thus contributing to residual selection bias. Nevertheless, if this was true, our findings are likely conservative and underestimate the treatment effect of PACE. Another limitation is that outcome analyses did not include jail facility or housing unit assignment by using multilevel modeling, because assignments are frequently variable and would have been difficult to assess for the control cohort.

In addition, other important health outcomes, such as symptomatic improvements (or the opposite), were recorded qualitatively in clinical notes and could not be measured. Injuries due to fights evaluated in this study may underestimate the scope of violence, verbal and physical, and violence involving nonpatients. However, fights among patients are the most specific indicator of patient-related behavior (as opposed to violence perpetrated or instigated by nonpatients) and so likely the most sensitive to clinical interventions.

Finally, we did not report on between-unit differences in the primary outcome variables, because this was beyond the scope of this first report about the overall PACE program. Differences between units that warrant further exploration were noted.

Conclusions

Creating a program of intensive, specialized, multidisciplinary treatment units in a jail setting yielded significant benefits related to patient psychiatric medication adherence and incidents of violence, as well as more modest reductions in self-injury. This first report on PACE findings highlights the value of a rehabilitative approach to jail incarceration and, we hope, will lead to more robust studies and program development in other jail settings to reduce the impact of incarceration on individuals with serious mental illness.