Abstract
Introduction:
Black patients are more likely to receive physical restraints in both emergency department (ED) and inpatient settings, yet the mechanisms behind these disparities remain unclear. This study aims to assess whether hospital type—academic tertiary care center versus county-funded trauma hospital—drives the association between race and restraint use.
Methods:
We used a community-based participatory research approach to analyze electronic medical record data from patients seen for a mental health emergency in 2019 and 2020 at two San Francisco hospitals, University of California, San Francisco Parnassus Hospital (UCSF) and Zuckerberg San Francisco General Hospital (ZSFG). Descriptive statistics and chi-square tests assessed patient characteristics by restraint, followed by multivariable logistic regression to estimate associations between race and restraint, stratified by hospital.
Results:
We identified 6,631 unique encounters, with 1,051 (15.8%) involving physical restraints. At ZSFG ED, 19.2% of patients were restrained, and 12.6% at UCSF ED. In combined adjusted hospital analysis, Black patients had higher odds of restraint (adjusted odds ratio [aOR] 1.54, 95% confidence interval [CI] 1.16–2.05). Stratified analysis showed significantly higher odds of restraint for Black patients at UCSF ED (aOR 1.77, 95% CI 1.12–2.79), while ZSFG ED showed marginally higher odds of restraint for Black patients (aOR 1.44, 95% CI 0.99–2.09).
Discussion:
Our findings suggest that hospital type may not significantly influence disparate restraint use in EDs. Further research should explore other potential contributors, including prehospital interactions with emergency services.
Health Equity Implications:
These results highlight the need for interventions that address the broader impact of structural racism on patients in mental health crises.
Introduction
Black patients in the USA face inequities driven by systemic, institutional, cultural, interpersonal, and internalized racism. These disparities are pronounced in mental health care and may be exacerbated when receiving emergency services. Emergency Departments (EDs) serve as critical access points for patients with mental health concerns, with the proportion of ED visits related to mental health rising from 7% in 2007 to 11% in 2016. 1
Compared with White Americans, Black Americans are more likely to be diagnosed with psychotic disorders, 2 perhaps due to inappropriate overdiagnosis, 2 yet access mental health care far less frequently. 3 Despite higher prevalence of substance use and psychotic disorders, 4 Black Americans are less likely to receive appropriate treatment. 5 Disparities in mental health care extend to coercive interventions, with Black patients experiencing physical restraints 6 and/or court-ordered involuntary holds 7 at higher rates than White patients.
San Francisco (SF) is a city that faces significant racial and socioeconomic disparities in health care access, shaped by historical redlining, structural inequities, 8 and rising homelessness, 9 further compounding barriers to mental health services. Black individuals in SF are hospitalized at higher rates than other racial groups, 10 potentially reflecting inadequate outpatient treatment and a greater need for acute psychiatric care. However, no studies to date have examined how hospital characteristics influence the use of physical restraints in EDs across SF.
In this exploratory study, we evaluate differences in physical restraint application on ED mental health patients at two hospitals in SF—a publicly funded trauma center and an academic tertiary care center. While our data focus on SF, the structural issues driving these disparities are not unique to the city. Given the inclusion of both an academic and a public hospital, our findings may offer insights generalizable to the health systems both locally and nationally. 11
Methods
Study design and setting
This retrospective cross-sectional study examined adult patients who sought emergency mental health care at two EDs. For the University of California, San Francisco Parnassus Hospital (UCSF), patient encounters between January 1, 2019, and December 31, 2020, were included, while for the public hospital, Zuckerberg San Francisco General Hospital (ZSFG), data were collected from August 3, 2019, to December 31, 2020, following the transition to an electronic medical record (EMR) system. The UCSF Institutional Review Board approved this study, and reporting adheres to STROBE guidelines. 12
UCSF is an urban, academic center located in central SF, with an annual ED census of over 40,000 visits. All patients presenting with mental health-related concerns are evaluated by an emergency medicine provider, and, if needed, a psychiatry resident is available 24/7 for consultation under attending psychiatrist oversight.
ZSFG is the only level one trauma center in SF, serving a predominantly underserved, urban population. The ED sees over 60,000 annual visits. 13 At ZSFG, medically stable patients presenting solely for mental health concerns may be directly transported to the Psychiatric Emergency Services unit by Emergency Medical Services (EMS) and bypass the main ED. This practice was suspended during the COVID-19 pandemic to ensure all patients were screened for COVID-19 before receiving care.
Community-based participatory research
We partnered with the Rafiki Coalition for Health and Wellness, a community organization with strong roots in SF’s Black community. In partnership with their executive director, we identified three additional individuals to serve on our Community Advisory Board (CAB). CAB members were selected based on their lived experience, interest in improving mental health care in the community, and availability to participate. CAB meetings were conducted throughout the study, with members shaping the conceptual framework, variable selection, and result interpretation.
Selection of participants
This study included all adults who presented to the ED who met one of the following criteria: (1) receipt of a psychiatric consultation in the ED, (2) medical admission following a suicide attempt, or (3) placement on an ED hold. An ED hold is a temporary measure placed by ED staff on patients deemed a potential risk to themselves or others, or to be unable to care for themselves due to severe mental illness. This designation remains in place while patients await assessment to determine the need for involuntary psychiatric detention.
Measures
All measures were extracted from the EMR. The primary outcome was a binary designation of physical restraint application, determined by the presence of an EMR order for physical restraints. All patients placed in restraints must have a corresponding EMR order placed at the time of restraint or immediately after.
The primary exposure variable was patient race, as documented in the EMR. At both hospitals, it is standard for patients to self-select race, ethnicity, and other demographic descriptors for inclusion in the EMR when possible. If a patient has been seen previously, these descriptors are carried forward to subsequent encounters. EMR categories of race included American Indian or Alaskan Native, Asian, Black, declined to answer, Native Hawaiian or Pacific Islander, Other, Other Pacific Islander, Unknown, and White. For analysis, racial groups were consolidated into four categories: Asian, Black, White, and Other, due to limited sample sizes in some groups. Ethnicity was documented as Hispanic or non-Hispanic. Alternative methods of reporting race and ethnicity were explored, including not combining groups and incorporating race and ethnicity together (e.g., non-Hispanic Black, Hispanic White).
Models were adjusted for demographic, socioeconomic, and health risk factors likely to influence risk of restraint: age, sex, insurance type, housing status, time of ED arrival, method of ED arrival, and documented history of mental health or substance use-related diagnoses in the EMR. Hospital type accounted for system-level factors.
Data analysis
Descriptive analyses consisted of frequency distributions of study sample characteristics using chi-square test of homogeneity to assess differences between patients who were and were not restrained. Primary analyses used multivariable logistic regression to estimate the association between patient race and restraint, adjusted for covariates using clustered standard errors to account for repeated observations of patients with multiple encounters. Stepwise and nested regressions assessed if covariates were independently associated with the outcome (p < 0.05). Because patient sex and site location may modify the primary association between race and risk of restraint, we conducted two secondary analyses: (1) stratified models by sex and (2) stratified models by hospital. All data analysis was completed using Stata 17.0 (StataCorp, version 17, 2021. College Station, TX).
Results
Patient demographics
We identified 6,631 encounters across both sites during the study period: 3,244 (48.9%) at ZSFG and 3,387 (51.1%) at UCSF, of which 1,051 (15.8%) had an order for physical restraints. Given no significant difference in results combining race and ethnicity (Supplementary Table S1), we report estimates separately. White race was most common (n = 2,864, 43.2%), followed by Other (n = 1,609, 24.3%), Black (n = 1,394, 21.0%), and Asian (n = 764, 11.5%). Within the ‘Other’ category less than 1% identified as American Indian or Alaskan Native, Native Hawaiian or Pacific Islander, or Other Pacific Islander, and the remainder selected ‘Other’. Most individuals identified as not Hispanic/Latino/a (n = 5,439, 81.9%) (Table 1). Differences in patient characteristics by hospital can be reviewed in Supplementary Table S2.
Descriptive Characteristics of Cohort by Restraint Order
One or more prior International Classification of Diseases (ICD) codes present in chart prior to presentation.
UCSF, University of California, San Francisco Parnassus Hospital; ZSFG, Zuckerberg San Francisco General Hospital.
Physical restraint use
At the ZSFG ED, 19.2% (n = 623) of patients were restrained, whereas 12.6% (n = 428) were restrained at the UCSF ED. Among the full study cohort, 19.3% (n = 269) of Black patients, 19.9% (n = 352) of patients aged 40–49, 22.2% (n = 119) of patients who were Self-Pay, 19.5% (n = 387) of people experiencing homelessness, and 21.2% (n = 695) of those arriving by ambulance were restrained.
Regression analyses
Asian patients had the lowest prevalence of restraints and served as the reference group in multivariable regression analyses. Black patients had the highest odds of being restrained (OR 1.77, 95% confidence interval [CI] 1.37–2.28) (Table 2), an association that attenuated but persisted after controlling for ethnicity, sex, age, time of arrival, insurance type, housing status, preexisting mental health or substance use disorder, arrival method, and hospital (adjusted odds ratio [aOR] 1.54, 95% CI 1.16–2.05) (Table 2). When stratified by hospital, Black patients had significantly higher odds of restraint at the UCSF ED (aOR 1.77, 95% CI 1.12–2.79) (Table 3). At the ZSFG ED, Black patients had higher odds of restraint (aOR 1.44, 95% CI 0.99–2.09), though not statistically significant, while Other Race had higher odds (aOR 1.53, 95% CI 1.02–2.30) of restraint (Table 3).
Univariate and Multivariable Logistic Regression Model Predicting Restraint Use, Combined Hospital Analysis a
Bolded p values significant at 95% confidence level.
The adjusted multivariate model included race, ethnicity, sex, age, time of arrival, insurance type, housing status, mental health disorder, substance use disorder, arrival method and hospital.
CI, confidence interval; UCSF, University of California, San Francisco Parnassus Hospital; ZSFG, Zuckerberg San Francisco General Hospital.
Multivariable Logistic Regression Model of Restraint Use by Hospital a
Bolded p values significant at 95% confidence level.
The adjusted multivariate model included race, ethnicity, sex, age, time of arrival, insurance type, housing status, mental health disorder, substance use disorder, arrival method and hospital.
CI, confidence interval; UCSF, University of California, San Francisco Parnassus Hospital; ZSFG, Zuckerberg San Francisco General Hospital.
Other covariates of interest
Arrival method
Many patients arrived by EMS (n = 3,280, 49.5%) (Table 1). In univariable analyses, those arriving by EMS had 2.55 (95% CI 2.18–2.98) the odds of restraint compared with those who self-presented (Supplementary Table S3). Arriving by ambulance was the single strongest predictor of restraint use in combined hospital multivariable analysis (aOR 2.35, 95% CI 1.99–2.78) (Table 2) and stratified analyses (UCSF aOR 5.83, 95% CI 4.25–8.00 and ZSFG aOR 1.40, 95% CI 1.14–1.73) (Table 3).
Comorbidity
Overall, 26.7% (n = 1,772) of individuals had a diagnosed comorbid mental health disorder, and 8.2% (n = 541) overall had a diagnosed substance use disorder. In multivariable combined-hospital analyses, presence of a mental health disorder was associated with lower odds of restraint (aOR 0.69, 95% CI 0.57–0.83, Table 2). Presence of a substance use disorder diagnosis was not associated with restraint in either univariable or multivariable analyses.
Sex
In combined hospital multivariable regression analysis stratified by sex, we found that the association between Black race and restraint use held for both male (aOR 1.46, 95% CI 1.02–2.11) and female (aOR 1.74, 95% CI 1.10–2.75) patients. Males less than 29 years of age had lower odds of restraint compared with those 60 or older (aOR 0.67, 95% CI 0.48–0.95), and those 40–49 had higher odds of restraint (aOR 1.37, 95% CI 1.05–1.79), but these associations were not significant among female patients (Supplementary Table S3).
Discussion
To our knowledge this is the first study to compare variations in the use of physical restraints among adults specifically with mental health-related ED visits in a public vs. private medical institution within the same city. Our findings indicate a disproportionate burden of restraint use among Black-identifying patients presenting to the ED with mental health concerns. This effect remained significant at the private, academic hospital and was just slightly below the significance threshold at the public hospital. We acknowledge the findings for Black-identifying patients were not significant at the public hospital; however, given that the pooled estimate as well as the estimate in the private hospital were significant, and the lower bound of the CI was 0.99 with an upper greater than 2, it is likely the nonsignificant finding in the public hospital was due to chance. Further work to explore this relationship with a larger sample size at the public hospital may be needed to clarify the association.
In the combined hospital analysis, patients aged 40–49, insured by Medicare, or transported to the hospital via ambulance had increased odds of being restrained. Notably, ethnicity was not a significant factor in restraint use across any analysis. The overall similarity in findings between the two hospital settings suggests that hospital type is not a primary driver of racial disparities in restraint use.
Our results align with a recent systematic review of restraint use among all comers in EDs, which found higher odds of restraint among Black individuals, whereas Hispanic patients had a lower prevalence. 6 However, a recent study specifically examining ED restraints use reported significantly higher odds of restraint use among non-Hispanic Black and Hispanic-identifying patients than non-Hispanic White individuals. 14 Consistent with prior research, we found that patients with non-private insurance—including Medicaid, Self-Pay, public insurance, or “lack of private insurance”—had higher odds of restraint use. 15
Although numerous studies have demonstrated a consistent association between mental health diagnoses and increased restraint use,14–16 we observed a decrease in the odds of restraint use among patients with a documented mental health diagnosis. One possible explanation is that ED providers, when treating patients with known mental illness and access to their EMR history, may be better able to anticipate patient behavior and thus less inclined to use restraints during crises. Conversely, patients with undiagnosed mental illness may be perceived as more unpredictable or threatening, increasing the likelihood of restraint use. In addition, patients in our cohort who have previously accessed mental health care services within our health system may be less likely to require ED ‘crisis care’ than someone without regular mental health care. Future research should explore differences in restraint use based on whether a patient has a known versus unknown mental health diagnosis at the time of ED presentation.
The mechanisms leading to the differential application of restraints on Black patients in EDs have yet to be definitively described, though the associations described above are hypothesis generating. Patients with mental health crises may present with acute agitation, yet standardized protocols for management are inconsistently applied. 17 Prior studies indicate that ED staff harbor implicit racial biases,18–21 which can influence critical decision-making, particularly in high-risk situations. When managing agitated patients, these biases may contribute to differential treatment. 22
Institutional policies related to staffing may also play a role. Critical care nurses have described how staffing ratios and physical and psychological staff exhaustion may contribute to restraint use, 23 and EDs face similar staffing shortages. 24 Future studies by this group include qualitative work with community members and ED staff to better elucidate these mechanisms.
Patients arriving by EMS had higher odds of being restrained compared with those who self-presented to the ED. This may reflect a greater severity of illness among patients arriving via EMS. Alternatively, prehospital interactions with bystanders, police, EMS, and other first responders may exacerbate agitation, leading to increased restraint use upon ED arrival. Other research has shown that police transport to the ED mediates the relationship between Black race and physical restraint use. 25 Future research should investigate whether prehospital restraint placement influences subsequent ED restraint use. Additional research is needed to understand demographic patterns in EMS utilization for mental health emergencies and to identify potential disparities in prehospital care that may contribute to differential restraint use in the ED.
Limitations
Our primary exposure variable was race, not experiences of racism, and these two constructs are not interchangeable.26,27 Although racism is a fundamental driver of racial disparities in ED restraint use, its measurement within health systems remains challenging, limiting our ability to directly quantify its impact. As a result, our findings are based on available data and require inferential interpretation. Because health-promoting opportunities are influenced by both race and socioeconomic status (SES), 28 we examined the independent role of race while controlling for SES by incorporating variables like insurance and housing status into our models. However, residual confounding likely remains. In forthcoming work, we aim to address this by incorporating qualitative data to better understand the underlying mechanisms behind the identified differences. We hypothesize that interpersonal, institutional, and structural racism may play a significant role.
Our study was limited by a shorter data collection period at the public hospital compared with the academic hospital, and primarily during the fall and winter months. Seasonal variations in mental health symptoms 29 and in the use of restraint/seclusion 30 have been documented, and these temporal factors may have influenced our results. Furthermore, data were included during the time of the COVID-19 emergency stay-at-home measures, a period marked by significant increase in negative mental health symptoms. The potential impact of this rise in societal mental unwellness on our findings is unclear and may limit the generalizability of our conclusions.
We did not assess the appropriateness of restraint use in this study because the American Association of Emergency Psychiatry published Project Best Practices in the Evaluation and Treatment of Agitation (BETA), 31 which strongly recommends against restraint use whenever possible. According to these guidelines, any deviation (i.e., restraint use) should result in an equally distributed consequence or differences explained due to chance, which was not the case for Black patients in our study.
Health equity implications
A lack of evidence illustrating a broader impact on other patients of color suggests that a general racial equity framework may be insufficient to explain our findings. Instead, we propose that the racially differentiated use of physical restraints is a direct consequence of anti-Black racism. Anti-Black racism refers to dehumanization of Black individuals grounded in colonialism, enslavement, and ongoing subjugation of Black communities. Previous literature underscores the importance of explicitly naming this type of racism. 32 Engaging in discussions about anti-Black racism is essential for advancing health equity. Disparities in psychiatric care and other fields of medicine disproportionately affect Black individuals due to the historical and enduring impacts of anti-Black racism. For example, Black individuals are more likely to be placed on involuntary legal holds 33 and to receive inpatient psychiatric consultations, 7 but are less likely to receive outpatient psychiatric care, 34 even following acute psychiatric hospitalization.35,36 Restraint use is linked to a reduced likelihood of successful mental health follow-up. 37 Insufficient engagement in timely preventative and therapeutic care may increase the likelihood of acute presentations due to the deterioration of underlying mental and physical health conditions. 7 In this context, disproportionate use of physical restraints exacerbates the mental health equity crisis for Black patients.
Health systems magnify broader societal dynamics, and a greater focus on both the mechanisms that perpetuate bias and those that foster equity can provide valuable insights into how to reduce disparities beyond the medical context. Our study highlights the critical need for targeted mental health needs assessments, paired with interventions that are informed by an understanding of the broader impact of structural racism. There is a critical need in health to account for the historical trauma experienced by Black communities, both within and beyond medical care, and to avoid causing new trauma through coercive action. Through community partnerships, health systems must improve care delivery so that Black patients feel safe, welcome, and empowered to seek mental health services. These goals for health systems align with medicine’s objectives to alleviate suffering and guide healing and will require a commitment from health care leaders to fight anti-Black racism. This study underscores the need to implement reforms in the emergency mental health care system for more racially equitable and patient-centered care.
Authors’ Contributions
D.J.: Data curation, formal analysis, software, writing original draft, and writing—review and editing. L.A.: Data curation, formal analysis, and writing original draft. A.H.: Formal analysis, supervision, writing original draft, and writing-review and editing. K.W.: Formal analysis, supervision, writing original draft, and writing-review and editing. M.L.: Conceptualization, funding acquisition, methodology, project administration, resources, writing—review and editing, and supervision. A.B.: Conceptualization, project administration, and resources. V.J.: Conceptualization, methodology, and writing—review and editing. T.H.: Conceptualization, methodology, and writing—review and editing. L.A.: Project administration, and writing—review and editing. S.W.: Project administration, and writing—review and editing. R.D.: Conceptualization, funding acquisition, methodology, and writing—review and editing. A.R.H.: Conceptualization, funding acquisition, methodology, writing original draft, writing—review and editing, and supervision. W.M.: Conceptualization, funding acquisition, methodology, writing—review and editing, and supervision. M.D.T.: Methodology, data curation, formal analysis, and software. V.E.: Conceptualization, funding acquisition, methodology, project administration, writing original draft, writing—review and editing, and supervision.
Footnotes
Author Disclosure Statement
The authors have no interests to disclose.
Funding Information
Financial support was provided by the Resource Allocation Program at the University of California, San Francisco.
Supplemental Material
Abbreviations
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
