Affordability of Forensic Assertive Community Treatment Programs: A Return-on-Investment Analysis

Adults with serious mental illness are overrepresented throughout the criminal legal system, including in jail and prison settings (1–4). They are held longer when incarcerated, have more disciplinary problems, and have higher rates of victimization and suicide compared with other inmates (1, 3). They are also at increased risk for injury or death after encounters with police (5, 6). These issues, along with recent calls for downsizing of correctional facilities and defunding of police (7, 8), have highlighted the need for effective community-based interventions for individuals with serious mental illness who are involved in the criminal legal system.

Forensic assertive community treatment (FACT) has recently emerged as a service delivery model for persons with serious mental illness who have histories of repeated arrest and incarceration (9). FACT is an adaptation of assertive community treatment (ACT), an evidence-based model for delivery of comprehensive mental health, addiction treatment, and housing support services via mobile multidisciplinary teams in community settings (10). Although ACT effectively prevents psychiatric hospitalization, studies examining ACT’s impact on criminal legal outcomes have failed to demonstrate effectiveness in preventing arrest and incarceration among patients with serious mental illness (11–14). To address this need, FACT was developed. Its key components include targeting criminogenic risk factors, collaboration with criminal legal professionals, and emphasis on problem solving rather than punishment (9, 15).

Reviews of FACT effectiveness (16–19) have indicated that FACT has shown promise in reducing arrest and incarceration rates, including in a recent study in Belgium (18). Conclusions about FACT effectiveness have been limited, however, by the absence of a standardized FACT model and a paucity of randomized controlled trials. To address these issues, investigators at the University of Rochester Medical Center (URMC) Department of Psychiatry obtained funding from the National Institute of Mental Health to develop and test a standardized FACT model. On the basis of a theoretical framework for preventing criminal recidivism (20), experience with a prototype model (21), and focus group input from FACT service providers and recipients (22), the Rochester FACT model was developed and operationalized into an experimental 13-item FACT fidelity scale in preparation for controlled testing (23). Patients randomly assigned to receive FACT subsequently had significantly fewer convictions, fewer days in jail, and fewer days in the hospital, along with better engagement in outpatient treatment, compared with patients in usual care (23).

Despite growing evidence of FACT’s effectiveness, however, cost has emerged as a potential barrier to FACT implementation (24–26). Only one study has examined FACT cost outcomes and found that increased FACT costs were partially, but not completely, offset by decreased jail and inpatient costs (27). Evidence from nonforensic ACT programs suggests that reductions in inpatient care utilization are a potential source of cost savings (28–31). More direct evidence supporting FACT’s positive economic impact is therefore needed to demonstrate its financial viability and sustainability.

In this article, we present results of a retrospective secondary data analysis from URMC’s federally funded Rochester FACT study to further examine the cost-effectiveness of FACT. Specifically, we tested the hypothesis that cost savings would outweigh additional costs of FACT program implementation, potentially providing an economic justification for support from a government payer’s perspective.

Methods

Study Design and Participants

URMC Institutional Review Board approval was obtained to conduct a retrospective secondary analysis of deidentified data previously collected as part of the Rochester FACT study. The study was a randomized controlled trial conducted between February 2011 and May 2014 to examine the effectiveness of the Rochester FACT model at preventing jail and hospital use. Details of the Rochester FACT study were previously published (23) and are summarized below.

Seventy adults with DSM-IV diagnoses of psychotic disorders who had been arrested on misdemeanor charges were recruited through the public defender’s office in Monroe County, New York. All participants entered the study under a conditional discharge status whereby their sentences had been suspended on the conditions of avoiding further criminal activity and accepting mental health treatment. The study was presented as a treatment option after participants’ cases were adjudicated to minimize risk for coercion, and all participants provided informed consent.

Individuals who were receiving mandated treatment as stipulated by probation, parole, mental health court, or assisted outpatient treatment programs were excluded from study enrollment to enable comparison of leveraged and nonleveraged treatment groups. This step was completed because use of legal leverage was a key component of the Rochester FACT intervention that distinguished it from the control group (20). By excluding those already on mandated treatment, the study team sought to minimize the confounding effects of legal leverage in the control group. Individuals with current felony charges were also excluded, but those with previous felony convictions were eligible for enrollment.

In preparation for study initiation, staff who delivered FACT services received 6 hours of instruction on mental health and criminal legal collaboration. This group included all clinicians on the FACT treatment team, the presiding judge who would provide judicial oversight of FACT treatment, and attorneys for FACT study participants. In addition, the FACT team psychiatrist provided informal and ongoing instruction and support to FACT team clinicians during regular treatment team meetings.

After baseline assessment, study patients were randomly assigned to receive either FACT or enhanced treatment as usual (ETAU) for a 1-year period. Patients in the FACT group were treated by a mobile multidisciplinary team with high fidelity to the ACT model as evidenced by a baseline score of 4.75 out of a possible 5.00 on the Dartmouth Assertive Community Treatment Scale (32). In addition, all FACT group patients received judicial monitoring by the presiding judge through regular court appearances. Study participants assigned to ETAU received expedited access to outpatient mental health treatment and intensive case management at a case manager–to-client ratio of 1:15. Case managers worked to connect ETAU patients to transportation, housing, finances, and other community resources as needed. ETAU patients did not have active judicial monitoring and were not scheduled to appear before the presiding judge until 9 months after study enrollment.

Data Collection and Analysis

Outcome results of the trial consisted of data on criminal legal service utilization (e.g., arrests, convictions, and incarcerations) and mental health service utilization (e.g., outpatient mental health contacts, psychiatric emergency department [ED] visits, and psychiatric hospitalizations). To the extent that reductions in incarceration and hospitalization represented cost savings from the government payer’s perspective, these reductions captured the total return from the FACT intervention. This return, however, represented gross rather than net savings because it did not reflect the costs associated with implementing the FACT intervention (i.e., the investment cost). In this study, therefore, the overall economic value of the FACT intervention was expressed as return on investment (ROI), calculated by dividing the return by the corresponding investment cost.

The investment portion of the ROI quotient was categorized into two types: variable costs (i.e., costs that changed depending on the number of patients treated by FACT) and fixed costs that did not change depending on the number of patients treated. The primary fixed costs for the FACT intervention group included cost of training as well as follow-up and monitoring of cases by the presiding judge during weekly 2-hour court hearings. The amount of time and effort devoted to court hearings did not vary by the number of patients served and therefore was considered a fixed cost for this study. Moreover, because no direct fee was associated with the training, the training cost was represented by the total hourly wages incurred by the psychiatrists who provided the training and by the presiding FACT court judge who received the training. For the control group, the fixed cost was zero because no training was provided and the judge performed substantially less follow-up and monitoring of patients.

Variable cost included per-patient-per-encounter direct costs of clinical treatments. For the FACT intervention group, this cost was represented by the New York State Medicaid reimbursement rate for ACT, specifically, the adult ACT 68 Upstate rate (33). If a patient had incurred at least 72 FACT or ACT visits during the year (or at least six visits per month, which was the threshold for determining full vs. partial reimbursement rate), the per-visit rate was calculated by multiplying $1,684 by 12 (i.e., the full per-month rate multiplied by 12 months in a year), which was then divided by the total visit count incurred during the year. If the patient had incurred <72 FACT or ACT visits, $842 (the partial per-month rate) was multiplied by 12 and then divided by the total visit count.

For participants in the ETAU group who had outpatient clinic visits and received intensive case management in lieu of FACT, the variable cost included per-visit costs of outpatient clinic visits, which were estimated on the basis of 2019 New York State certified community behavioral health clinic (CCBHC) rates (34). The CCBHC rate was a per capita reimbursement rate designed to cover costs of a wide range of outpatient mental services that a patient could reasonably expect to receive, such as outpatient clinic treatment and case management. Therefore, although patients in the control group received intensive case management support, the cost of case management services was not separately included as investment because the per capita CCBHC rates already included reimbursements for case management costs.

For each study patient, counts of jail days, psychiatric ED visits, and days in hospitals (i.e., components of the return portion of ROI) as well as counts of FACT visits and outpatient clinic visits (i.e., components of the investment portion of ROI) were multiplied by the corresponding unit cost values. (A table showing detailed descriptions of the unit cost values used in the analysis is available in an online supplement to this article.) The sum of the dollar values across the categories was then calculated for each patient to estimate the total cost of care and the total investment cost associated with the patient. Because study patients varied in their lengths of study participation, patients’ total cost of care and total investment costs were divided by their corresponding time in the treatment arm in years to obtain dollar values on a per-patient-per-year (PPPY) basis. To obtain 95% CIs around the resultant ROI point estimates, we used a percentile bootstrap method with 1,000 replications (35).

Results

As reported in the previously published study (23), the baseline data of the 70 study participants showed that the study sample had a mean±SD age of 37.5±11.6 years; most participants were male (N=43, 61%), belonged to racial-ethnic minority groups (African American or Hispanic; N=57, 81%), were unemployed (N=66, 94%), and had never been married (N=59, 84%).

Table 1 summarizes the differences between the FACT intervention and ETAU groups. The FACT intervention group had shorter psychiatric hospital stays (statistically significant at p<0.05) and fewer days in jail (statistically significant at p<0.10). Patients in the FACT intervention group had significantly higher rates of outpatient treatment use mainly because of FACT clinical visits; patients in the ETAU group had significantly more outpatient clinic visits compared with those in the FACT group, although at an overall lower rate compared with the FACT visits.

Table 2 summarizes the details of the resource utilization cost reductions (i.e., returns) and the implementation cost increases (i.e., investment) attributable to FACT. Although none of the four categories of resource utilization was individually associated with a statistically significant cost reduction attributable to FACT, the total sum of the reductions of $27,588 PPPY was statistically significant (p<0.05). Taking the ratio between the total utilization cost reductions and the investment cost (Table 3), we found that the ROI for FACT was 1.46 with the fixed cost and 1.50 without it. However, these ROI estimates did not reach statistical significance, as indicated by the 95% CIs that included 1, indicating the break-even point.

Figure 1 illustrates the sensitivity of the ROI estimate to a 10% increase or decrease in the payment rate of each utilization or investment category, while holding other payment rates constant. The ROI estimate was most sensitive to changes in FACT reimbursement rate and per diem psychiatric inpatient reimbursement rate. Specifically, a 10% increase in FACT payment rate (i.e., a higher investment cost) reduced ROI from 1.50 to 1.34, whereas a 10% decrease in the payment increased ROI to 1.69. However, a 10% increase in the psychiatric inpatient reimbursement rate (i.e., a higher return) increased ROI from 1.50 to 1.61, whereas a 10% decrease reduced ROI from 1.50 to 1.38. Table 4 shows reimbursement rates for ACT programs, psychiatric hospitalization, and jail across four locations in the United States, illustrating that ROI estimates are subject to context-dependent factors such as regional variations in reimbursement rates and costs of care.

Discussion

In this study, we quantified the economic value of FACT in dollar terms from the government payer perspective. The results indicate that the Rochester FACT model is associated with a positive ROI and that the main sources of the return were reductions in inpatient care and jail days. Our findings also suggest that the positive ROI did not come at the expense of public safety because results from a previous study (23) indicate that the FACT intervention group had fewer total arrests, although this decrease was not statistically significant.

Although the positive ROI estimates were not statistically significant because of the large variance in the ROI component estimates and the relatively small sample size, our results are consistent with the expectation that FACT programs may pay for themselves from the government payer’s perspective. Moreover, we note that a positive ROI was observed even with a disproportionately high proportion of racial-ethnic minority patients in the sample (81%). Black and Hispanic individuals typically have multiple societal disadvantages, including poverty and less access to health care services compared with nonminority individuals (36). Because such individuals are highly overrepresented throughout the criminal legal system (37), the results of this study have implications for FACT as a potentially cost-effective strategy to address racial-ethnic disparities in mental health and criminal legal services.

Several limitations should be noted. This study considered only the returns directly related to psychiatric treatment and days in jail. Individuals with serious mental illness are known to have high medical costs (31, 32), but the potential impacts of FACT on costs associated with medical ED visits, medical-surgical inpatient admissions, and specialist visits were not included. This study also did not account for indirect benefits associated with FACT, such as reductions in costs associated with police involvement and crime-related damages as well as intangible costs of human suffering. Therefore, the ROI estimate reported in this study may be interpreted as a conservative estimate. Furthermore, a potential investment cost not included in our study is the cost incurred by the legal system in collaborating with the FACT team beyond attending the training sessions. Although the study protocol required meetings with FACT team clinicians, the study judge has indicated that collaboration by court staff did not require additional time or resources beyond their usual duties.

Another limitation of this study and of the earlier Rochester FACT study (23) was that neither study was designed to examine potential mechanisms underlying FACT effectiveness. For example, the effect of legal leverage might have been controlled for by comparing FACT with a control group that was equally leveraged through judicial monitoring. However, doing so would have fundamentally changed our original research question from determining whether the Rochester FACT model works to determining how it works. This latter question remains a topic for future research, and it also implies that ROI estimates can vary depending on how the control group treatment is defined. For instance, ROI estimates would likely be higher if patients in the control group had poor access to outpatient treatment and used ED and inpatient care instead.

An additional study limitation relates to the use of secondary data from a study conducted nearly a decade ago. This concern about the age of the data was mitigated by the interval validity of the Rochester FACT study as a randomized controlled trial and by the use of more recent reimbursement rates that were current as of 2019. Moreover, because ROI is a ratio of dollar values captured at a given time point, the ROI point estimates are not affected by annual inflation. Also, we note that the Rochester FACT study was designed to examine the model’s clinical effectiveness rather than its cost-effectiveness. Consequently, it is likely that the study was underpowered for the purposes of this article, which might explain why the ROI estimates were not statistically significant.

We also note that the generalizability of this study may have been limited by the lack of a standardized FACT model. Given significant variability in how current FACT programs are structured, how they operate, and what populations they target (15), programs deviating from the Rochester model may not achieve the same results. Because of the significant FACT investment cost that can have a disproportionately large impact on its ROI (as illustrated by Figure 1), government payers may be incentivized to reduce FACT and ACT reimbursement rates. Doing so, however, may compromise fidelity to the FACT model and lead to a lower ROI.

ROI estimates are highly sensitive to context-dependent factors such as regional variations in reimbursement rates and costs of care, as indicated by Figure 1. Table 4 further illustrates this point by showing comparisons of reimbursement rates for ACT programs (which are often the same rates as used for FACT programs) (38–40), psychiatric hospitalization (41–43), and jail costs (44–46) across four localities in the United States: Rochester, New York; Los Angeles; Massachusetts; and Rhode Island. Localities with higher reimbursement rates for ACT programs but lower rates for psychiatric inpatient stays (i.e., Los Angeles and Massachusetts) are generally expected to have a lower ROI than what was found in this study. Likewise, localities such as Rhode Island that have lower reimbursement rates for ACT but higher rates for psychiatric hospitalizations can generally be expected to have a higher ROI. Furthermore, depending on which government agencies or departments are responsible for jails, prisons, mental and general medical health care services, and other services, ROI may need to be recalculated to generate stakeholder-specific estimates.

Conclusions

The results of this study suggest that the FACT intervention has an approximately $1.50 return for every $1 spent on its implementation, even without considering potential returns from other direct and indirect sources. We conclude that FACT may be financially sustainable from the government payer’s perspective.