Health Care Costs of Individuals With Chronic Psychotic Disorders Who Experience Incarceration in Ontario

Chronic psychotic disorders such as schizophrenia are severe and disabling mental disorders (1) and are associated with poor health, social, and economic outcomes (2). For example, individuals with chronic psychotic disorders are at increased risk for criminal justice involvement, such as arrests and incarceration, and are overrepresented in correctional settings (3). Moreover, they are more likely than individuals without chronic psychotic disorders to experience reincarceration after release from correctional facilities (3). Although individuals with schizophrenia have a higher risk for arrest and incarceration than those in the general population, many of their actions that lead to arrest result from being victimized by others (4).

Compared with persons with schizophrenia but without encounters with the criminal justice system, individuals with schizophrenia and criminal justice system encounters are more likely to engage in substance use and less likely to adhere to antipsychotic treatment (5), which may affect health care costs. Furthermore, persons with schizophrenia have poorer levels of mental health functioning and are more likely to be hospitalized and to use emergency services (5). Thus, although health care costs for all individuals with chronic psychotic disorders are substantial (6), costs are likely even higher for criminal justice–involved individuals. One U.S. study found that individuals with schizophrenia or bipolar disorder with criminal justice involvement had higher total mean costs (by 26%) for mental health and substance use services than those with schizophrenia or bipolar disorder without criminal justice encounters (7). However, the study did not examine costs of general medical health care, which can also be substantial for this population (6).

Our hypothesis was that individuals with chronic psychotic disorders who experience incarceration have higher total health care costs than individuals with comparable disorders who have not experienced incarceration. If this hypothesis is true, higher health care costs among individuals with chronic psychotic disorders may signal the need for interventions and policies that help such individuals avoid future criminal justice system involvement. Using administrative correctional and health care data, we estimated 1-year health care costs in the year before incarceration among individuals with chronic psychotic disorders who experienced incarceration and among comparable individuals who did not experience incarceration.

Methods

Setting and Study Design

The Ontario provincial correctional system is responsible for the detention of people on remand (i.e., those arrested and awaiting trial) and people with a custodial sentence of <2 years. This study used a matched case-control design, in which “cases” were defined as individuals experiencing incarceration and “controls” as those who did not. Cases included all individuals released from Ontario’s provincial correctional facilities (i.e., prisons) between January 1, 2010, and December 31, 2010 (8, 9). The index release date was defined as the date of first release in 2010; the index admission was defined as the incarceration event leading to the index release. The look-back period was defined as the 3 years before the date of the index admission (i.e., from 2007 to 2009) to ensure the same observation period for correctional and health care cost data for each case regardless of the duration of the index incarceration. Health care costs were examined in the year before the index admission to understand how these costs differed between cases and controls. Given Ontario’s public health care system, where access to care is not dependent on ability to pay, health care costs and experiencing incarceration can be assumed to be independent of each other. The study was approved by the Hamilton Integrated Research Ethics Board (study 4575).

Data

The Ontario Ministry of the Solicitor General provided data on individuals released from provincial correctional facilities in 2010. The data included information on sex, age, address on correctional entry, number of previous incarcerations, dates of admission to and of release from, and reasons for release from provincial correctional facilities.

Health services utilization data were obtained from ICES (formerly known as the Institute for Clinical Evaluative Sciences), an independent, nonprofit research institute in Toronto. Health care databases described below captured administrative, clinical, and demographic data on all health care encounters covered by the public health care system. Data on hospital-based care were captured in the Discharge Abstract Database (all acute medical hospitalizations and psychiatric hospitalizations in nonpsychiatric designated beds), the Ontario Mental Health Reporting System (all psychiatric hospitalizations in psychiatric designated beds), the Continuing Care Reporting System (continuing and long-term care), and the National Rehabilitation Reporting System (rehabilitation). Data on ambulatory care, such as emergency department visits, were recorded in the National Ambulatory Care Reporting System. The Ontario Health Insurance Plan claims database captured data on physician visits and laboratory and diagnostic claims. The Ontario Drug Benefit Program database included information on outpatient prescription drugs dispensed to individuals covered by the public provincial drug plan (i.e., individuals ages ≥65 years and individuals ages <65 years who were living in a long-term care home or a home for special care or a Community Home for Opportunity, receiving professional home and community care services, enrolled in the Trillium Drug Program, or on social assistance). The Home Care Database records all visits provided by home care professionals. (See table in the online supplement to this article.)

The Registered Persons Database, a population-based registry maintained by the Ontario Ministry of Health, was used to obtain population and demographic data on residents who use the health care system, such as date of birth, sex, address, date of death, Ontario Health Insurance Plan eligibility and status changes, and postal code of residence, which was used to obtain data on neighborhood-level income quintile and rurality of residence (when such data were missing in this database, Ministry of the Solicitor General data were used).

The corrections data were linked to ICES data through deterministic and probabilistic linkage analysis by using health card numbers (resulting in a linkage rate of 97%) (9). Data were analyzed at ICES; under Ontario’s health information privacy law, ICES is allowed to collect and analyze demographic and health care data, without consent, for health system evaluation and improvement. The use of health care data in this study was authorized under section 45 of Ontario’s Personal Health Information Protection Act.

Population

All deidentified records for individuals released from provincial correctional facilities between January 1, 2010, and December 31, 2010, were selected. Data on individuals ever diagnosed as having a chronic psychotic disorder before incarceration (i.e., cases) were selected with a validated algorithm (10). Individuals whose data could not be linked to a valid identifier in the data, with missing sex or age, of ages <18 or >105 years at the time of first correctional facility release date in 2010, who died before correctional release, with missing postal code, or not residing in Ontario during the 3 years before the admission date of the index incarceration were excluded.

Individuals in the case group were matched 1:2 on sex (male or female) and age (±6 months from birth date) to individuals in the control group, as is commonly done in case-control studies, in order to understand how both groups differed on variables of interest, such as neighborhood-level income and illness duration. Data from individuals in the matched control group were obtained from the Registered Persons Database from a pool of individuals diagnosed as having a chronic psychotic disorder by using the same inclusion and exclusion criteria applied to the case group, and their 2007–2009 data were extracted for analysis.

The following sociodemographic and clinical characteristics were used for each group: sex, age at index admission to prison, neighborhood-level income quintile, urban or rural residence, and duration of chronic psychotic disorder. For the case group, the latter was the period between the date of first diagnosis of chronic psychotic disorder and the admission date for the index release date from prison in the administrative data; for the control group, psychotic disorder duration was assessed as the period between the date of first diagnosis of chronic psychotic disorder and January 1, 2010 (i.e., the index date for controls because they did not have an incarceration release date), in the administrative data. These descriptive summaries were replicated for each group by sex. Between-group differences were examined as standardized mean differences (SMDs), where an SMD of >0.10 was considered large (11). Data on average length of incarceration, number of previous incarcerations, and reason for release (released on bail, released at court, satisfied sentence, release to hospital, or other) were also examined for those in the case group.

Data Analysis

A cost algorithm was used to estimate all direct patient-level health care costs to third-party payers (i.e., the Ontario Ministries of Health and Long-Term Care) (12). The costing method as defined in the algorithm consisted of a bottom-up and microcosting approach to cost services at the individual patient level, which identified individual episodes of care or utilization in the health care system and attached prices or costs paid for each care episode or encounter. Given Ontario’s public health insurance system, prices are rarely set by providers in a private marketplace; therefore, costs and amounts paid by the Ministry of Health were used. In cases where individual unit costs were not available (e.g., long-term care), a top-down approach was used, which allocated corporate aggregate costs to individual visits or cases or care episodes. Further details on the costing methodology can be found elsewhere (12). Costs captured by this algorithm account for >90% of all government-paid health care services (13). Total mean 1-year costs for the year before the admission date of the index incarceration in 2010 were estimated for cases and for the same period for controls. Costs by sex and health service for both groups and by mental health– and nonmental health– related care (using a previously defined algorithm [14]) were also estimated. Costs were expressed in 2018 Canadian dollars.

A generalized linear model with a gamma distribution and a log link was estimated to model total costs (15). Variables identified in previous research to have an impact on health care use and costs were included in the model (16). The main independent variable was incarceration (no vs. yes); control variables included sex (female vs. male), age group (18–29 vs. 30–39, 40–49, 50–59, 60–69, and ≥70 years), neighborhood income quintile (5, high, vs. 1, low; 2, medium low; 3, medium; and 4, medium high), rural residence (no vs. yes), and duration of chronic psychotic disorder (≤1 year vs. 1–2 years, 3 years, 4–5 years, 6–10 years, and >10 years). Data from individuals with missing data on neighborhood income quintile and rurality (N=216) were excluded from this analysis. Model output was presented as marginal effects for ease of interpretation; robust standard errors were estimated. All analyses were done with SAS, version 9.4.

Results

At the date of release (i.e., in 2010), individuals with a chronic psychotic disorder who experienced incarceration (i.e., cases) (N=3,197) had a mean age of 37 years, and most were men (84%) and lived mainly in urban settings (92%) (Table 1). All individuals in the case group, except one, were matched 1:2 to control group individuals (N=6,393). Persons from low-income neighborhoods were overrepresented in the case group (42% vs. 32%, SMD=0.20) (Table 1). Cases also had a shorter duration of a chronic psychotic disorder (2,801 days vs. 3,236 days, SMD=0.21).

In the case group, women were slightly older than men (mean age 38.5 vs. 36.8 years, SMD=0.16), but both sexes were generally similar in terms of neighborhood-level income quintile, urban residence, and duration of chronic psychotic disorder (Table 2). Women had a shorter mean length of incarceration (33.4 vs. 71.1 days, SMD=0.32) and fewer previous incarcerations (mean 4.04 vs. 4.76, SMD=0.15). The reasons for release also varied by sex: for women, the most common reason was released at court (65%) and for men, it was released at court (35%) and satisfied sentence (33%).

Individuals in the case group had total mean 1-year costs of $15,728 before incarceration; costs of psychiatric hospitalizations ($8,797) and physician services ($2,302) accounted for 56% and 15%, respectively, of the total costs (Table 3). Control group individuals had total mean 1-year costs of $11,588, with costs of psychiatric hospitalizations ($6,032) accounting for 52%, and physician services ($1,547) and outpatient prescription drugs ($1,557) accounting for roughly 13% each (Table 3). Among those who experienced incarceration, 68% (N=2,174) had public drug coverage, compared with 54% (N=3,452) of those who did not experience incarceration. Median and maximum costs also differed between the two groups ($2,407 and $652,519 for cases vs. $4,474 and $382,696 for controls, respectively), indicating differences in cost distributions. Most costs were due to mental health– related care (76%, [$11,993] and 72% [$8,320] for cases and controls, respectively). (Results broken down by mental health care and general medical care are available from the authors on request.) The between-group 1-year cost difference was $4,140, mostly due to psychiatric hospitalizations, emergency department visits, and physician services.

Costs for women and men in the case group were $20,648 and $14,763, respectively (a difference of $5,885) (Table 4); furthermore, the female case group had almost double the costs of the female control group ($11,338). The difference between male case and control groups was smaller ($14,763 vs. $11,637). The cost breakdown was qualitatively the same across subgroups compared with the full sample.

In the regression analysis (Table 5), incarceration in the following year, being female, and living in an urban area were associated with higher costs. Individuals in the 30–39 and 40–49 age groups had lower costs compared with those ages 18–29; those with duration of a chronic psychotic disorder of 1–2 years had higher costs than those with duration of ≤1 year. All other variables were not statistically significant. The main cost drivers included incarceration in the following year, being between 18 and 29 years old at index incarceration admission date, and duration of chronic of chronic psychotic disorder of 1–2 years. In particular, individuals in the case group had $4,827 higher costs than those in the control group (p<0.001), and individuals ages 30–39 and 40–49 years had $4,448 and $4,218 lower costs than those of ages 18–29, respectively (p<0.001). Individuals with duration of a chronic psychotic disorder of 1–2 years had $6,812 higher annual costs than those with a duration of ≤1 year (p=0.004).

Discussion

As we had hypothesized, individuals with chronic psychotic disorders who experienced incarceration had higher 1-year health care costs than those who did not. The between-group mean cost difference was $4,140, mostly driven by higher costs arising from psychiatric hospitalizations and, to a lesser extent, costs of emergency department visits and physician services. Among those who experienced incarceration, women had costs that were almost $6,000 higher than the costs for men; furthermore, women who experienced incarceration had substantially higher costs than women who did not experience incarceration, by just over $9,000. The regression analysis confirmed these findings.

Few studies have examined the costs of individuals diagnosed as having chronic psychotic disorders who experience incarceration (5, 7). Swanson et al. (7) examined costs of criminal justice involvement among adults with severe mental illness who received services across public agencies within Connecticut, where criminal justice involvement was defined as having an arrest that resulted in a criminal conviction, any period of incarceration, time spent on probation or parole, participation in a jail diversion program, or forensic mental health involvement (7). The authors found that the justice-involved group was significantly younger, more likely to be male, and more likely to be African American than their counterparts not involved with the justice system. They also reported that justice-involved individuals had total mean costs of mental health and substance use services of $31,196 (2007 US$), $6,468 higher than the costs of those without justice involvement ($24,728). In our study, individuals with chronic psychotic disorders who experienced incarceration were younger (mean age of 37 vs. 49 years) and more likely to be male (84% vs. 53%), compared with a prevalence sample of Ontarians with chronic psychotic disorders (6). Furthermore, individuals who experienced incarceration had health care costs that were just over $4,000 higher than costs for those without incarceration. The Swanson et al. study (7) also examined criminal justice system costs (except forensic care). When these costs (US$17,784) were taken into account, the justice-involved group incurred costs that were approximately double the costs for the group without justice involvement, that is, $48,980 versus $24,728 per person. Although we did not directly examine the total mean cost of incarceration per individual, we estimate that it would be $15,145 (2018 CAN$) (but likely higher for individuals with chronic psychotic disorders), for a combined cost of $30,873 (2018 CAN$), on the basis of the average length of incarceration of 65 days in the sample, and an overall average incarceration cost of $233 per day (17).

Related work has also shown that people with schizophrenia have an increased reincarceration risk, in part because of poor service utilization (3). In a study of individuals on remand with psychotic illness, reincarceration risk was associated with poor service and housing availability in the community upon release (18). Although we did not examine these risk factors, these findings, combined with the current results, suggest that the combination of high costs of care and subsequent criminal justice involvement could be seen as a marker of disease severity and complexity, highlighting the need for enhanced clinical services for people with psychotic illness. Moreover, women who experienced incarceration had substantially higher costs than men with and women without incarceration, suggesting the need for sex-tailored approaches for individuals with chronic psychotic disorders who experience incarceration.

This study used a unique data linkage between the Ministry of the Solicitor General and ICES, which enabled the identification of a population-based sample of individuals who experienced incarceration. Our study accounted for most costs covered under a universal health care system and adds to the literature by estimating cost differences by sex. However, we note some limitations. The diagnosis of chronic psychotic disorders relied on clinical codes in the administrative data rather than on a comprehensive clinical assessment; thus, some individuals may have been misclassified as having a chronic psychotic disorder when in fact they did not have such disorder and vice versa. Nonetheless, the algorithm we used has been validated with population-based administrative databases in Ontario. Our study may have captured data on individuals who experienced incarceration in the 3 years before the index incarceration admission date; the inclusion of data on these individuals had likely biased costs downward. Furthermore, the difference in drug coverage between the two groups likely contributed to higher average costs for individuals who experienced incarceration. This study used data from Ontario’s provincial correctional facilities but did not examine data on federal correctional involvement, including health care utilization in federal prisons, which is paid by the federal government and not included in provincial data. However, only a small percentage of individuals included in this study would have experienced federal incarceration during the study period. Finally, this study could not account for costs of specialized community-based drug and alcohol services, because these data are currently unavailable for research purposes at ICES.

Conclusions

The costs of health care for individuals with chronic psychotic disorders who experience incarceration are higher than care expenditures for comparable individuals with such disorders who do not experience incarceration. Women had substantially higher costs than men, and women who experienced incarceration had higher costs than women who did not experience incarceration. Higher health care costs may signal the need for interventions or policies that can help individuals with psychotic disorders avoid criminal justice system involvement. As such, future research should seek to determine cost-effective interventions for justice-involved individuals with chronic psychotic disorders. Moreover, future work should estimate the health care costs of individuals who experience other aspects of criminal justice encounters such as arrests or forensic hospitalizations.