Involuntary psychiatric hospitalization is accepted by many clinicians as a part of clinical practice that is sometimes necessary, yet it remains controversial (
1–
4). Arguments in favor cite the duty to protect patients and society from foreseeable harm and the potential for involuntary care to preserve the right to treatment for patients who lack insight (parens patriae) (
1,
2). Arguments against cite the imprecision of risk prediction, the unethical nature of linking detention to mental illness, the adverse impact on help-seeking behavior, the risk of iatrogenic traumatization, and the incompatibility with a recovery model (
3–
5).
Given extensive data showing that Black patients are more often subjected to coercive treatments, including restraint and seclusion, discussions of coercive practices in psychiatry must attend to race and racism (
6 –
12). Further, characterization of racial and ethnic inequities is particularly relevant at a time when there is significant energy to address structural racism in mental health care (
13).
Compared with studies examining inequities in restraint and seclusion, there are few such investigations related to civil commitment (
14). Most studies come from the United Kingdom, with very few tracking legal status from admission through discharge (
6,
15). Although the literature consistently shows ethnoracial inequities in involuntary hospitalization, it is unclear whether these findings are driven primarily by interpersonal bias, clinical differences between ethnoracial groups, or differential exposure to social determinants of health. A better understanding of the determinants of commitment would inform interventions to ensure that involuntary commitment is employed only as a last resort and in an equitable fashion.
The Bias In Acute Services project used data collected prospectively over a 6-year period on all admissions to a general inpatient psychiatric unit in a large general hospital in Boston. This data set was designed to investigate clinical, demographic, and treatment-related predictors of various outcomes, including restraint and involuntary hospitalization, to support quality improvement work. This project focused on patient ethnoracial identity and involuntary hospitalization. The inpatient unit serves a sociodemographically and clinically diverse patient population and is thus well suited to examine determinants of involuntary hospitalization. We hypothesized that patients of color (defined in this study as all non-White patients) would be overrepresented among those involuntarily held in the hospital at measured time points and that such inequities would not be fully explained by other demographic and clinical factors.
Methods
Data Collection
The study was conducted on a 24-bed adult psychiatry unit in a large academic medical center. The unit admits patients with psychiatric and substance use disorders and often manages patients with significant medical comorbid conditions. All admissions were longitudinally followed to the point of discharge between August 1, 2012, and December 31, 2018. For patients with multiple admissions during the study, only initial admissions were included. This resulted in a sample of 4,489 unique patients from a total of 5,832 admissions during the study period. All individuals who declined to provide a self-identified race were excluded, leaving a final sample of 4,393 patients.
Information regarding racial and ethnic identity was obtained by patient self-report at hospital registration. Patients could identify as White, Black, Hispanic or Latinx, Asian, or other (which included patients identifying as two or more races). Notably, prior to 2015, the hospital’s registration policies allowed patients to select Hispanic or Latinx as their primary race. After 2015, the hospital added a separate question about Hispanic or Latinx ethnicity to be consistent with the approach used by the U.S. Census Bureau. For the purposes of this study, after 2015, when a patient’s race was listed as other or unknown but ethnicity was reported as Hispanic or Latinx, that patient was grouped with patients who before 2015 identified their race as Hispanic or Latinx. During the same time frame, if patients chose to identify with one of the available options for race (White, Black, Asian, or other or multiracial) and selected Hispanic or Latinx as their ethnicity, they were grouped with the race they selected (White, Black, Asian, or other or multiracial).
Trained administrative staff used a standardized data collection tool to prospectively record demographic and clinical data via chart review. They supplemented their data collection as needed with conversations with the clinical team during daily multidisciplinary meetings. In preparation for the final data analysis, we used additional retrospective chart reviews to identify missing data. Less than 5% of data were missing for any variable in the final analysis.
Prior to analysis of the relationship between ethnoracial identity and legal outcomes, research staff a priori identified a list of potentially relevant confounders. The following sociodemographic variables were selected: age, gender, housing status, and insurance status. The following clinical variables were selected: admission day of week, referral source, treatment care team, and psychiatric diagnoses (both the primary billing diagnosis and secondary psychiatric diagnoses listed on the discharge summary).
This study was reviewed by the Partners Healthcare (now known as Mass General Brigham) Institutional Review Board and deemed to be a quality improvement project. It was approved by the departmental chief quality improvement officer.
Legal Procedures
In Massachusetts, patients can be involuntarily admitted to a psychiatric unit by a licensed clinician. At admission, the patient must be offered the opportunity to sign in to the hospital on a conditional voluntary legal status (hereafter referred to simply as “voluntary”). If the patient declines or lacks the capacity to sign in, then the physician must determine whether the patient meets commitment criteria on the basis of imminent danger to self or others or grave inability to care for self. The admitting physician can admit such patients for an involuntary hold for an observation period of 3 business days (equivalent to a 72-hour emergency hold in most states). During this 3-day period, patients with decisional capacity can choose to sign in and convert their legal status to voluntary. Alternatively, if a patient on a voluntary status requests discharge when the team believes that the patient may meet commitment criteria, the patient's status can be converted to a 3-day involuntary hold (hereafter referred to as a “vol-to-invol conversion”). The potential for this holding period is the reason that voluntary legal status is considered “conditional.”
At the end of the 3-day involuntary holding period (either 3 days after the involuntary admission or the end of the 3-day hold after a vol-to-invol conversion), the physician can either discharge the patient or petition the court for involuntary civil commitment of up to 6 months. The patient is assigned legal counsel, and a court date is scheduled within 5 business days, although it is often postponed. If at any point the patient agrees to sign in voluntarily, the court hearing can be canceled. While awaiting the court date, if the patient is no longer felt to meet commitment criteria, the physician may discharge the patient. If the judge does not grant longer-term commitment at the hearing, then the patient is immediately discharged.
The primary outcomes of interest were involuntary legal status at admission, filing of a formal civil commitment petition, and proceeding with a court hearing. An additional primary outcome was “discharge from a 3-day hold” (either after an involuntary admission or after a vol-to-invol conversion), because this represents a particular group of patients who declined to stay in the hospital voluntarily but were ultimately found not to meet commitment criteria. Finally, a secondary outcome of invol-to-vol conversion was derived from the primary-outcomes patients in this group who were admitted involuntarily but later signed into the hospital voluntarily.
Analysis
Statistical analyses were carried out in R. The distributions of categorical variables between groups of interest were compared by using chi-square tests of association, and continuous variables were compared by using one-way analysis of variance. Univariate analysis was performed to assess the rates of legal status–related outcomes of interest between racial groups; it was also performed on a set of a priori–defined covariates (as described above). Multivariate models for each legal outcome were derived by using backward stepwise multivariate regression, beginning with all covariates significant in univariate analysis below a threshold of p<0.2 and removing covariates until all that remained met a significance threshold of p<0.1. Sensitivity analyses were run with the full list of identified covariates, and the results were similar to those obtained from backward stepwise regression. Adjusted odds ratios (aORs) and 95% confidence intervals (CIs) were determined for each variable included in the model. A significance threshold for primary and secondary outcomes of p<0.05 was selected a priori, and no adjustments were made for multiple comparisons. When calculating ORs for likelihood of the secondary outcome of invol-to-vol conversion, only patients who began the admission involuntarily were included.
Results
A flow diagram in the
online supplement summarizes the legal pathways that all 4,393 patients followed from admission to discharge. At admission, 28% (N=1,240 patients) of the sample was on an involuntary legal status. Many of these patients either converted to voluntary admission and were discharged or were discharged after a 3-day hold. Therefore, the sample of patients on the commitment pathway gradually decreased: initial involuntary 3-day hold, 28% (N=1,240); petition filed for a court hearing, 7% (N=315); court hearing held, 3% (N=127); and formal civil commitment granted, 3% (N=122). The vast majority of patients admitted voluntarily left on the same legal status (89%, N=2,813 of 3,153), whereas many patients admitted involuntarily converted to voluntary status by discharge (47%, N=583 of 1,240).
Table 1 shows the demographic, clinical, and legal characteristics for the full sample in aggregate and by ethnoracial category. Overall, 52% of the sample was male, and the mean age of the sample was 43.4 years. On average, patients of color (defined in this study as all non-White patients) were younger, compared with White patients. The gender distributions for White patients and patients of color were similar. Homelessness was more common among Black patients (33% versus 18% for the overall sample). Both lack of insurance and accessing care through the emergency department were also more common among patients of color, compared with White patients. Diagnostically, the most striking differences were in the incidence of psychotic disorders, which were markedly more common among patients of color (25% for White patients versus 41%–58% for the other groups).
Table 2 shows the univariate logistic regression analysis of legal outcomes by the covariates used in our analysis, many of which were significantly associated with legal status. Patients who were experiencing homelessness were more likely than those who were housed to be involuntarily admitted. Patients who were commercially insured were less likely than those without commercial insurance to be involuntarily admitted. Patients with a primary diagnosis of bipolar disorder, schizophrenia spectrum disorder, or unspecified mood disorder were more likely than patients with depression to be on an involuntary status. Patients with substance use–related diagnoses were less likely than those without such diagnoses to be involuntarily admitted. In general, many of the same covariates that predicted involuntary admission were predictive of the outcomes discharged from 3-day hold and filed for commitment (
Table 2).
Univariate analysis of legal status outcomes by ethnoracial identity was notable for increased rates of involuntary admission among patients of color (
Table 3). This finding was especially prominent for patients who identified either as Black or as other or multiracial. After adjustment for covariates, these differences were attenuated, although they remained significant for the Black patient group (aOR=1.57) and for the other or multiracial group (aOR=2.12). Black patients were also more likely than White patients to have commitment petitions filed (OR=2.90) and to be discharged from a 3-day hold (OR=1.32); however, these findings did not reach statistical significance in multivariate analysis. Similarly, Asian patients were more likely than White patients to have commitment petitions filed (OR=2.10), but this finding did not reach significance in multivariate analysis. Finally, in univariate analysis only, Black and other or multiracial patients were found to be less likely than White patients to convert from an involuntary to a voluntary legal status during their admission.
Of the 127 patients who went to court, 122 (96%) were court committed to continued inpatient treatment. Given the small number of patients in this group who were released by the court (N=5), these patients were excluded from further analysis, and comparisons were conducted between those who were released before their court date (N=188) and those who were court committed (N=122). The number of patients from each ethnoracial group who reached this point in the commitment pathway was small. Therefore,
Table 4 summarizes clinical and demographic features from this final comparison in aggregate form. Overall, few differences reached statistical significance, but patients released before their court date were younger, compared with patients who were ultimately court committed.
Of note, over the 6-year study, no clear temporal effect on group-level differences was noted. (A table in the
online supplement shows rates of involuntary hospitalization by ethnoracial group for each year in the study.)
Discussion
In this study, patients of color were more likely than White patients to be involuntarily admitted. After adjustment for confounding variables, this pattern remained significant for Black patients and for other or multiracial patients. Evidence of ethnoracial inequity was found for other legal outcomes measured, particularly for Black patients; however, some findings did not remain significant in multivariate analysis.
In describing these findings, the term “inequity” was intentionally chosen in lieu of “disparity” or “difference.” This choice was made to call attention to the “systemic, avoidable, and unjust social and economic policies” that produce differential exposure to the social determinants of mental health, access to care, and other factors that we propose are upstream of our findings (
16).
To our knowledge, this is the largest U.S. study designed to investigate ethnoracial inequities in civil commitment. A similar U.S. study did not find race to be independently predictive of involuntary hospitalization, but that study’s power was limited by a smaller sample (227 patients) and one that was more clinically and racially homogeneous (84.9% Black) (
15). Another U.S. study from 2015 had a sample size similar to our own but included fewer details on legal outcomes (
17 ).
Europe has a more extensive literature on the determinants of involuntary hospitalization, particularly in the United Kingdom. Consistent with our data, a large meta-analysis of these studies showed that patients of color, and especially Black patients, were more likely than White patients to be civilly committed (
6). However, this meta-analysis did not adjust for any variables that might confound or mediate the relationship between ethnoracial identity and civil commitment.
Although few studies have been designed to investigate the causes of ethnoracial inequities in commitment, many explanations have been proposed. These include interpersonal racism, implicit bias, and differences between racial groups with respect to illness severity, access to and engagement with mental health care, and distribution of upstream social determinants of health (
6,
18–
23). Although we cannot address all these hypotheses, our findings add to the literature regarding the impact of illness expression and social determinants of health on commitment inequities. Prior work has shown that if diagnostic factors (specifically, increased prevalence of psychosis among patients of color) and social factors are controlled for, ethnoracial inequities in commitment are no longer observed (
24,
25). Our work contradicts these findings, showing that Black and other or multiracial identities predicted involuntary admission independent of diagnostic and socioeconomic covariates. Future work should interrogate the role of patient-provider–level interpersonal racism, differences in symptom severity, and prehospitalization treatment history in mediating inequities (
26,
27).
The fact that some of our findings did not remain significant in multivariate analysis is notable, but it does not necessarily imply that race was unrelated to these outcomes. First, although our sample size was large, some outcomes (filed for commitment and discharged from 3-day hold) were relatively low-frequency events. Given our findings at admission, future work investigating these outcomes in larger samples is warranted. Second, because of the pervasive influence of structural racism, many of the covariates used were likely not truly independent of race. Through this lens, rather than considering all covariates as “confounders,” many of the covariates may in fact mediate the impact of structural racism on outcomes (
28).
This study had several limitations. The single-site design is relevant, given variability in commitment laws and local practice patterns. We were unable to investigate information related to differences in the commitment criteria (harm to self, harm to others, or grave disability) used across groups. Further, there may have been variable classification of patients identifying as Hispanic or Latinx before and after 2015 because of institutional changes in coding of racial and ethnic identity during the study. Overall, however, any such risk was likely limited, given that the proportion of patients identifying as Hispanic or Latinx was consistent before and after 2015 and fairly closely in accord with the proportion of Hispanic or Latinx individuals both admitted to our general hospital and in our institution’s home neighborhood (data not shown). Finally, although we concluded on the basis of available data that race predicted involuntary admission independent of demographic and clinical covariates, additional data would help exclude other potential explanations. These include but are not limited to more detailed sociodemographic data, treatment history, information regarding the provider decision-making process, and more detailed data on clinical presentation and illness severity.