Background: Security emergency responses (SERs) are utilized by hospitals to ensure the safety of patients and staff but can cause unintended morbidity. The presence of racial inequities in SER utilization has not been clearly elucidated. We sought to determine whether Black and Latinx patients experience higher rates of SER and physical restraints than white patients in a non-psychiatric inpatient setting.

Methods: We conducted a retrospective cohort study. We included all patients discharged from September 2018 through December 2019 from the main building of Brigham and Women’s Hospital, a tertiary hospital in Boston, MA. The exposure was race, as reported by patients at time of registration. The primary outcome was whether an SER was called on a patient. The secondary outcome was the incidence of physical restraints among patients who experienced an SER. We conducted a Firth logit analysis for the primary outcome and incorporated age, sex, mental health and substance use diagnoses, and length of stay as covariates in our multivariable models. We conducted a Poisson regression with the same covariates for the secondary outcome.

Results: Among 24,212 patients, 18,394 (76.0%) patients identified as white, 2,346 (9.3%) Black, and 1,703 (7.5%) Latinx. 66 (2.8%) Black patients had an SER activated during their first admission, compared to 37 (2.2%) Latinx patients and 288 (1.6%) white patients. In a Firth logit multivariable model, Black race was associated with higher adjusted odds of an SER than white race (adjusted odds ratio (aOR) 1.38 [95% Confidence Interval: 1.03,1.84], p = 0.028), and Latinx ethnicity was not (aOR 1.11 [0.78, 1.60], p = 0.56). In a Poisson multivariable model among patients who had an SER called, Black race was associated with lower physical restraint incidence than white race (adjusted incidence rate ratio 0.49 [0.30, 0.79], p = 0.004).

Conclusions: Black race was associated with higher odds of an SER and lower incidence of physical restraints compared to white race. Staff racial bias may impact the decision to call security. Future efforts should focus on identifying effective interventions to address this inequity.

IMAGE 1: Patient demographics

IMAGE 2: Univariate and multivariable models