BACKGROUND AND OBJECTIVES

A Behavioral Emergency Reponse Team (BERT) is a rapid response team for behavioral emergencies, which include clinical psychiatric emergencies, coping/stress reactions and physical or verbal conflicts.1 Use of BERT activations (“alerts”) in inpatient pediatric populations is understudied. The objective of this study is to determine any differences by race for individuals receiving BERT alerts at a single pediatric hospital.

METHODS

A cross-sectional retrospective review of all inpatient BERT alerts was conducted between January 1, 2018, and December 31, 2020. Primary outcome was presence or absence of a BERT alert during a single hospital admission (Financial Identification Number). A mixed-effects logistic regression model was conducted to test the association between race and BERT alert as well as physical restraint (PR) use, adjusting for age, gender, ethnicity, payor, and mental health diagnosis (MHD) and clustering at the patient level.

RESULTS

A total of 683 alerts occurred between the years 2018 and 2020. Admissions for Black patients had higher odds (adjusted odds ratio [aOR], 2.1; 95% CI, 1.6–2.8; P < .001) of having a BERT alert. Admissions with private insurance had lower odds (aOR, 0.4; 95% CI, 0.3–0.5; P < .001) of any BERT alert. Having an MHD was associated with higher rates of BERT alert (46.7% vs 15.0%; P < .001) and PR use (aOR, 16.7; 95% CI, 5.1–55.0; P < .001).

CONCLUSION

Patients with Black race, government insurance, and MHD had a disproportionate number of BERT. MHD and BERT alerts are associated with increased PR use.

Behavioral Emergency Response Teams (BERTs) are used in hospital settings to bring trained team members to the bedside to mitigate patient or visitor behavioral escalation and prevent potential safety issues for patients, families, and health care team members.1–5 BERTs were first formed as a modification of rapid response teams, which are used to identify and address early signs of medical decompensation.3 The structure of a BERT varies by institution. Members may include security officers, nursing staff, medical and psychiatric care providers, and providers with training in de-escalation such as social workers.2,3,6 For caregivers and visitor escalation, many institutions do not respond with BERTs but rather have a security-only response, although a qualitative study reports that some institutions do use BERTs in this context.7 Based on successes in reducing both staff injuries and patient restraint use, BERTs have been widely implemented, including in pediatric hospitals.2,4 

Although there is some information about BERT activation (“alert”) use in pediatric settings, little is known about the use of BERT responses in the general pediatric inpatient population. A possible unintended consequence of BERTs is disproportionate use for racially and ethnically minoritized patients. Data from security-only emergency responses at 1 adult hospital demonstrated that Black patients had higher odds of experiencing a security emergency response than white patients.8 There is also limited and inconsistent data around racial and ethnic disparities in pediatric patients around physical restraint (PR) and chemical restraint use, which are relevant as BERT interventions. Although PR has not been examined in the context of behavioral response teams, 1 study demonstrated higher odds of PR use for Black as compared to white pediatric patients in the emergency department.9 However, a study of PR use in pediatric patients on a medical-behavioral unit did not find differences in race,10,11 and another study looking at chemical restraint use in pediatric patients in the emergency department also found no increased use in Black patients.11 It is important to examine utilization of BERTs in the context of restraint use because there are no studies that have looked at racial and ethnic disparities for BERT activations in pediatric patients.

Our objective was to determine whether there are differences by race, payor, and presence of mental health diagnoses between the inpatient pediatric population and patients receiving BERT alerts. We hypothesized that admissions for Black patients, those on government-funded insurance, and those with a mental health diagnosis (MHD) have higher proportion of BERT alerts compared with admissions for white patients, those with private insurance, and those without documented MHD.

This study was conducted at a single-site pediatric hospital within a larger hospital system. We conducted a cross-sectional retrospective study of all inpatient BERT alerts at Riley Hospital for Children in Indianapolis, Indiana, at Indiana University Health between January 1, 2018, and December 31, 2020. We abstracted demographic characteristics for all inpatient admissions during the time period. The institutional review board at Indiana University approved this study and waived the need for informed consent. All analyses were performed with SAS version 9.4.

Our hospital implemented its BERT in April 2011. The team is activated when concern for behavioral escalation occurs. This activation can be initiated by any member of the nursing or medical team, as well as the patient or family/visitor, and is projected via an overhead auditory page and electronic page. The mandatory team respondents include police officers and social workers with training in behavioral de-escalation, the patient’s primary medical team (ie, the team’s senior resident or attending), and bedside nursing. Chaplaincy and unit nursing leadership are invited but not required to attend each BERT alert. The team structure did not significantly change during the COVID-19 pandemic, except when social workers briefly worked remotely and would virtually call into each BERT alert. Records of the individual who activates the BERT alert are not kept. Measurement of PR use (as recorded in electronic medical records [EMRs]) was also obtained. Measurement of chemical restraint was not obtained.

Police records were linked using patient name and room number to a specific Financial Identification Number (FIN), which corresponds to a single patient admission. The FIN was then used to identify the patient’s EMR chart and linked to a corresponding BERT alert note left by social work on the date of the alert. BERT alerts that could not be validated with a specific FIN (due to incorrect or missing patient names, room numbers, or dates) resulted in the exclusion of that BERT alert in the data analysis. If there was a police record of a call but no corresponding BERT alert note by social work, this was deemed a “security-only” call. BERT alerts that occurred outside of general inpatient locations (ie, medical/surgical units and intensive care units [ICUs]), such as outpatient clinics, emergency departments, inpatient physical medicine and rehabilitation units, and inpatient psychiatric units, were excluded. Primary analyses were performed at the admission (FIN) level whereby only the first BERT alert within 1 FIN was counted. A sensitivity analysis was conducted on the first BERT alert called per admission to eliminate confounding information. The primary outcome was whether any type of BERT alert (parent, patient, or visitor) occurred during the admission. Secondary outcomes included for which individual (patient, parent) the alert was called.

The person for whom the alert was called (patient, parent, visitor) was obtained from the BERT alert notes. Variables obtained from the EMR included the patient’s gender (male, female), race (white, Black, other), ethnicity (Hispanic, not Hispanic), insurance (government funded, private, self-pay), MHD, and PR use. MHD was determined by using validated International Classification of Diseases codes from the Child and Adolescent Mental Health Disorder Classification System.12 PR use was documented as present if EMR order for PR occurred during the same admission, whether a BERT alert occurred or not. Correlation of timing of BERT alert and PR use was not performed.

De-identified information was gathered into a Research Electronic Data Capture database (REDCap). Data were flagged if there was a discrepancy between the patient, location, and time of the alert recorded in the police record and the EMR. The flagged data were audited and verified for accuracy by the study team members. EMR data were used as the ultimate source of accuracy.

Race was included as a covariate in this study to better understand if the process of calling a BERT alert was occurring disparately for Black patients and families. It is important to note that race is a social construct. It is the effects of persistent racism over time that have led to racial and ethnic health inequalities, not biological differences. Of note in our medical system, a patient’s gender, race, and ethnicity are typically self-reported upon admission to the hospital by either the patient or guardian. No information about the parent’s race was obtained, as this is not regularly gathered during registration processes and this study was done retrospectively. As such, demographic information during a BERT alert for a parent is reflective of the patient assigned to the FIN. Given that our system does not have high rates of race other than Black or white, we opted to only include the 3 categories in our analysis (see Table 1 footnote).

TABLE 1.

Bivariate Relationship Between Behavioral Emergency Response Team Alerts and Sociodemographic Characteristics

Overall (N = 24 332)Any AlertPatient AlertParent Alert
No (n = 24 088)Yes (n = 244)P ValueaNo (n = 24 236)Yes (n = 96)P ValueaNo (n = 24 183)Yes (n = 149)P Valuea
Race    <.001   <.001   <.001 
 Black 4448 (18.9) 4359 (18.7) 89 (37.6)  4410 (18.8) 38 (40.4)  4396 (18.8) 52 (36.1)  
 Otherb 698 (3.0) 697 (3.0) 1 (0.4)  697 (3.0) 1 (1.1)  698 (3.0) 0 (0.0)  
 White 18  384 (78.1) 18  237 (78.3) 147 (62.0)  18  329 (78.2) 55 (58.5)  18  292 (78.2) 92 (63.9)  
Mental mealth diagnosis    <.001   <.001   .003 
 No 20  617 (84.7) 20  487 (85.0) 130 (53.3)  20  595 (85.0) 22 (22.9)  20  511 (84.8) 106 (71.1)  
 Yes 3715 (15.3) 3601 (15.0) 114 (46.7)  3641 (15.0) 74 (77.1)  3672 (15.2) 43 (23.9)  
Gender    .80   .57   .98 
 Female 11  074 (45.5) 10  961 (45.5) 113 (46.3)  11  028 (45.5) 46 (47.9)  11  006 (45.5) 68 (45.6)  
 Male 13  252 (54.5) 13  121 (54.5) 131 (53.7)  13  202 (54.5) 50 (52.1)  13  171 (54.5) 81 (54.4)  
Ethnicity    .03   .16   .03 
 Hispanic 2545 (10.6) 2531 (10.7) 14 (5.9)  2540 (10.6) 5 (5.3)  2538 (10.6) 7 (4.8)  
 Not Hispanic 21  448 (89.4) 21  223 (89.3) 225 (94.1)  21  358 (89.4) 90 (94.7)  21  310 (89.4) 138 (95.2)  
Insurance    <.001   .002   <.001 
 Government funded 14  752 (60.7) 14  550 (60.4) 202 (82.8)  14  674 (60.6) 78 (81.3)  14  628 (60.5) 124 (83.2)  
 Other 208 (0.9) 207 (0.9) 1 (0.4)  208 (0.9) 0 (0.0)  207 (0.9) 1 (0.7)  
 Private 8902 (36.6) 8862 (36.8) 40 (16.4)  8885 (36.7) 17 (17.7)  8878 (36.7) 24 (16.1)  
 Self-pay 459 (1.9) 458 (1.9) 1 (0.4)  458 (1.9) 1 (1.0)  459 (1.9) 0 (0.0)  
Age, y    .01   <.001   .11 
 0–3 11  009 (45.2) 10  921 (45.3) 88 (36.1)  11  006 (45.4) 3 (3.1)  10  926 (45.2) 83 (55.7)  
 4–6 2680 (11.0) 2652 (11.0) 28 (11.5)  2671 (11.0) 9 (9.4)  2662 (11.0) 18 (12.1)  
 7–12 4603 (18.9) 4553 (18.9) 50 (20.5)  4573 (18.9) 30 (31.2)  4581 (18.9) 22 (14.8)  
 13–18 5423 (22.3) 5347 (22.2) 76 (31.1)  5370 (22.2) 53 (55.2)  5398 (22.3) 25 (16.8)  
 19+ 617 (2.5) 615 (2.6) 2 (0.8)  616 (2.5) 1 (1.0)  616 (2.6) 1 (0.7)  
Overall (N = 24 332)Any AlertPatient AlertParent Alert
No (n = 24 088)Yes (n = 244)P ValueaNo (n = 24 236)Yes (n = 96)P ValueaNo (n = 24 183)Yes (n = 149)P Valuea
Race    <.001   <.001   <.001 
 Black 4448 (18.9) 4359 (18.7) 89 (37.6)  4410 (18.8) 38 (40.4)  4396 (18.8) 52 (36.1)  
 Otherb 698 (3.0) 697 (3.0) 1 (0.4)  697 (3.0) 1 (1.1)  698 (3.0) 0 (0.0)  
 White 18  384 (78.1) 18  237 (78.3) 147 (62.0)  18  329 (78.2) 55 (58.5)  18  292 (78.2) 92 (63.9)  
Mental mealth diagnosis    <.001   <.001   .003 
 No 20  617 (84.7) 20  487 (85.0) 130 (53.3)  20  595 (85.0) 22 (22.9)  20  511 (84.8) 106 (71.1)  
 Yes 3715 (15.3) 3601 (15.0) 114 (46.7)  3641 (15.0) 74 (77.1)  3672 (15.2) 43 (23.9)  
Gender    .80   .57   .98 
 Female 11  074 (45.5) 10  961 (45.5) 113 (46.3)  11  028 (45.5) 46 (47.9)  11  006 (45.5) 68 (45.6)  
 Male 13  252 (54.5) 13  121 (54.5) 131 (53.7)  13  202 (54.5) 50 (52.1)  13  171 (54.5) 81 (54.4)  
Ethnicity    .03   .16   .03 
 Hispanic 2545 (10.6) 2531 (10.7) 14 (5.9)  2540 (10.6) 5 (5.3)  2538 (10.6) 7 (4.8)  
 Not Hispanic 21  448 (89.4) 21  223 (89.3) 225 (94.1)  21  358 (89.4) 90 (94.7)  21  310 (89.4) 138 (95.2)  
Insurance    <.001   .002   <.001 
 Government funded 14  752 (60.7) 14  550 (60.4) 202 (82.8)  14  674 (60.6) 78 (81.3)  14  628 (60.5) 124 (83.2)  
 Other 208 (0.9) 207 (0.9) 1 (0.4)  208 (0.9) 0 (0.0)  207 (0.9) 1 (0.7)  
 Private 8902 (36.6) 8862 (36.8) 40 (16.4)  8885 (36.7) 17 (17.7)  8878 (36.7) 24 (16.1)  
 Self-pay 459 (1.9) 458 (1.9) 1 (0.4)  458 (1.9) 1 (1.0)  459 (1.9) 0 (0.0)  
Age, y    .01   <.001   .11 
 0–3 11  009 (45.2) 10  921 (45.3) 88 (36.1)  11  006 (45.4) 3 (3.1)  10  926 (45.2) 83 (55.7)  
 4–6 2680 (11.0) 2652 (11.0) 28 (11.5)  2671 (11.0) 9 (9.4)  2662 (11.0) 18 (12.1)  
 7–12 4603 (18.9) 4553 (18.9) 50 (20.5)  4573 (18.9) 30 (31.2)  4581 (18.9) 22 (14.8)  
 13–18 5423 (22.3) 5347 (22.2) 76 (31.1)  5370 (22.2) 53 (55.2)  5398 (22.3) 25 (16.8)  
 19+ 617 (2.5) 615 (2.6) 2 (0.8)  616 (2.5) 1 (1.0)  616 (2.6) 1 (0.7)  
a

Mixed-effects logistic regression with a fixed effect for the demographic variable of interest and a random effect for patient to account for the correlation of repeated admissions within patient.

b

Other category: Asian, 523; multiracial, 114; Native American, 36; Pacific Islander, 25.

We used mixed-effects logistic regression for bivariate comparisons of demographics (race, ethnicity, gender, age, and payor source) and MHD with BERT alerts. The model included a fixed effect for the demographic variable of interest and a random effect for patient to control for repeated admissions by the same patient.

We conducted a multivariate mixed-effects logistic regression model to examine the association between patient race and any BERT alert adjusting for MHD and sociodemographic factors. Similar analyses were conducted to examine association of race with a patient BERT alert and parent BERT alert, as separate outcomes. Nonparent visitor alerts were not included in the analyses outside of reporting frequency, due to the low volume. If an admission had both patient and parent alerts, that admission would be assessed as yes to any BERT alert, yes to patient BERT alert, and yes to the parent alert in the various outcomes.

To assess the association between patient race and restraint use, we conducted a multivariate mixed-effects logistic regression model with PR (yes/no) as the outcome adjusting for sociodemographics and mental health. Recognizing that young children under 6 years are rarely placed in restraints, we limited this analysis to patients 7 years and older.

We performed sensitivity analyses using data at the patient level. For the first analysis, we analyzed data patients with multiple admissions using only their first admission. For the second analysis, data for patients with multiple admissions were aggregated to create 1 record per patient. We conducted a multivariate logistic regression with these 2 data sets in which any BERT alert (patient or parent) was the outcome to examine the association of race with BERT alerts adjusting for sociodemographic characteristics. Similar analyses were conducted in these 2 sensitivity data sets using patient BERT alert and parent BERT alert as the outcome, separately.

During our study period, there were 24 332 admissions at the children’s hospital. We identified a total of 683 alerts reported between the years 2018 and 2020. We excluded 135 of these records due to a lack of corresponding EMR records and 179 for not occurring on the inpatient unit. We included 369 unique alerts linked to 244 unique admissions from 233 separate patients (Figure 1). Demographic information of our population stratified by alert status (yes/no) is presented in Table 1. Overall, 54.5% were male, 78.1% were white, 89.4% were non-Hispanic, and 60.7% had government-funded insurance.

FIGURE 1.

Inpatient Behavioral Emergency Response Team (BERT) alert identification.

Abbreviations: EMR, electronic medical record; FIN, Financial Identification Number; MRN, medical record number; SW, social work.
FIGURE 1.

Inpatient Behavioral Emergency Response Team (BERT) alert identification.

Abbreviations: EMR, electronic medical record; FIN, Financial Identification Number; MRN, medical record number; SW, social work.
Close modal

Of the 369 inpatient BERT alerts, 186 were for the patient, 173 were for a parent, and 10 were for a nonparent visitor. Approximately 1.0% (244/24 332) of all general inpatient admissions during the time of the study had any BERT alert, with a range of 1 and 4 BERT alerts per admission. Six admissions had only visitor-related alerts, 89 had only patient-related alerts, and 142 had only parent-related alerts. Seven had both patient- and parent-related alerts.

In bivariate analyses, multiple factors are associated with BERT alerts (Table 1). Specifically, BERT alerts were activated at a higher rate in patients who identify as Black race, or non-Hispanic ethnicity. Insurance differed signficantly (P < .001) by presence or absence of BERT alerts. Admissions with BERT alerts had a higher percentage of having government-funded insurance and a lower percentage of private-funded insurance than admissions without BERT alerts. Admissions with BERT alerts had a significantly higher percentage of MHD than admissions without BERT alerts (46.7% vs 15.0%; P < .001). The association of race, insurance status, and MHD with BERT alerts remained significant when patient alerts and parent alerts were separated. A separate analysis of visitor-related alerts was not done due to low numbers.

In multivariable analyses, for all types of alerts, we found patients with Black race had statistically significant higher odds of a BERT alert compared with white patients, even while accounting for ethnicity, insurance status, and presence of an MHD. For any BERT alert, Black patients had higher odds of BERT alerts (adjusted odds ratio [aOR], 2.1; 95% CI, 1.6–2.8; P < .001; Table 2). For patient and parent BERT alerts considered separately, Black patients had higher odds of BERT alert (patient BERT aOR, 2.8; 95% CI, 1.7–4.6; parent BERT aOR, 1.8; 95% CI, 1.3–2.6). Additionally, admissions with MHD had higher odds of BERT alerts in all adjusted models compared to admission without.

TABLE 2.

Multivariate Relationship Between Behavioral Emergency Response Team Alerts and Sociodemographic Characteristics

Any AlertPatient AlertParent Alert
OR (95% CI)P ValueaOR (95% CI)P ValueaOR (95% CI)P Valuea
Race  <.001  <.001  .001 
 Black 2.1 (1.6–2.8) <.001 2.8 (1.7–4.6) <.001 1.8 (1.3–2.6) .001 
 Other 0.2 (0.0–1.5) .12 0.9 (0.1–7.0) .92 – – 
 White (reference) –  –  –  
Mental health diagnosis 4.8 (3.6–6.4) <.001 13.3 (8.0–22.1) <.001 2.6 (1.8–3.9) <.001 
Gender 
 Female 1.0 (0.8–1.4) .84 1.1 (0.7–1.8) .59 1.0 (0.7–1.4) .95 
 Male (reference) –  –  –  
Ethnicity 
 Hispanic (reference) –  –  –  
 Not Hispanic 1.6 (0.9–2.9) .09 1.4 (0.5–3.5) .53 2.2 (1.0–4.9) .04 
Insurance  <.001  .008  <.001 
 Private 0.4 (0.3–0.6) <.001 0.4 (0.2–0.7) .002 0.4 (0.3–0.6) <.001 
 Other/self-pay 0.3 (0.1–1.2) .09 0.5 (0.1–3.5) .45 0.2 (0.0–1.6) .14 
 Government funded (reference) –  –    
Age, y  .35  <.001  .04 
 0–3 0.9 (0.6–1.5) .71 0.1 (0.0–0.5) .003 1.2 (0.7–2.1) .52 
 4–6 (reference) –  –  – – 
 7–12 1.0 (0.6–1.7) .93 1.9 (0.9–4.4) .11 0.7 (0.4–1.4) .29 
 13–18 1.1 (0.7–1.8) .66 2.1 (1.0–4.7) .06 0.7 (0.3–1.2) .19 
 19+ 0.3 (0.1–1.2) .09 0.4 (0.0–3.3) .39 0.2 (0.0–1.8) .16 
Any AlertPatient AlertParent Alert
OR (95% CI)P ValueaOR (95% CI)P ValueaOR (95% CI)P Valuea
Race  <.001  <.001  .001 
 Black 2.1 (1.6–2.8) <.001 2.8 (1.7–4.6) <.001 1.8 (1.3–2.6) .001 
 Other 0.2 (0.0–1.5) .12 0.9 (0.1–7.0) .92 – – 
 White (reference) –  –  –  
Mental health diagnosis 4.8 (3.6–6.4) <.001 13.3 (8.0–22.1) <.001 2.6 (1.8–3.9) <.001 
Gender 
 Female 1.0 (0.8–1.4) .84 1.1 (0.7–1.8) .59 1.0 (0.7–1.4) .95 
 Male (reference) –  –  –  
Ethnicity 
 Hispanic (reference) –  –  –  
 Not Hispanic 1.6 (0.9–2.9) .09 1.4 (0.5–3.5) .53 2.2 (1.0–4.9) .04 
Insurance  <.001  .008  <.001 
 Private 0.4 (0.3–0.6) <.001 0.4 (0.2–0.7) .002 0.4 (0.3–0.6) <.001 
 Other/self-pay 0.3 (0.1–1.2) .09 0.5 (0.1–3.5) .45 0.2 (0.0–1.6) .14 
 Government funded (reference) –  –    
Age, y  .35  <.001  .04 
 0–3 0.9 (0.6–1.5) .71 0.1 (0.0–0.5) .003 1.2 (0.7–2.1) .52 
 4–6 (reference) –  –  – – 
 7–12 1.0 (0.6–1.7) .93 1.9 (0.9–4.4) .11 0.7 (0.4–1.4) .29 
 13–18 1.1 (0.7–1.8) .66 2.1 (1.0–4.7) .06 0.7 (0.3–1.2) .19 
 19+ 0.3 (0.1–1.2) .09 0.4 (0.0–3.3) .39 0.2 (0.0–1.8) .16 

Abbreviation: OR, odds ratio.

a

Mixed-effects logistic regression with a fixed effect for the demographic variable of interest and a random effect for patient to account for the correlation of repeated admissions within patient.

Of all pediatric admissions, only 41 (0.2%) admissions involved use of PRs. PR were used in 7.0% (17/244) admissions with BERT alerts compared to 0.1% (24/24 088) of admissions without BERT alerts. In multivariate analysis of patients older than the age of 6 years (Table 3), Black patients had higher odds of PR use (aOR, 3.9; 95% CI, 1.1–13.4; P = .03). Additionally, MHD was associated with increased odds of PR use (aOR, 14.9; 95% CI, 4.6–48.1; P < .001).

TABLE 3.

Multivariate Association of Physical Restraint Use With Sociodemographics, Mental Health Diagnosis, and Behavioral Emergency Response Team Alert Excluding Patients 0–6 Years of Age

Physical Restraint
OR (95% CI)P Valuea
Race 
 Black 3.9 (1.1–13.4) .03 
 White (reference) —  
Mental health diagnosis 14.9 (4.6–48.1) <.001 
Female 1.0 (0.4–2.7) .98 
Ethnicity 
 Hispanic (reference) —  
 Not Hispanic 0.5 (0.1–2.6) .42 
Insurance 
 Not government funded 0.3 (0.1–0.9) .04 
 Government funded (reference) –  
Age, y 
 7–12 (reference) –  
 13+ 4.8 (1.3–18.2) .02 
Physical Restraint
OR (95% CI)P Valuea
Race 
 Black 3.9 (1.1–13.4) .03 
 White (reference) —  
Mental health diagnosis 14.9 (4.6–48.1) <.001 
Female 1.0 (0.4–2.7) .98 
Ethnicity 
 Hispanic (reference) —  
 Not Hispanic 0.5 (0.1–2.6) .42 
Insurance 
 Not government funded 0.3 (0.1–0.9) .04 
 Government funded (reference) –  
Age, y 
 7–12 (reference) –  
 13+ 4.8 (1.3–18.2) .02 

Abbreviation: OR, odds ratio.

a

Mixed-effects logistic regression with a fixed effect for the demographic variable of interest and a random effect for patient to account for the correlation of repeated admissions within patient.

Using only data from each patient’s first admission, Black patients had a higher odds of any BERT alert (adjusted OR, 2.7; 95% CI, 1.4–5.4; P = .004)) in multivariable analyses. For patient and parent alerts analyzed separately, patient Black race had higher odds of a parent BERT alert but not of patient BERT alert (Supplemental Table 1). Using data aggregated to each patient, (Supplemental Table 2), Black race was associated with BERT alert activation (any, patient only, and parent only).

Although BERT alerts occur in a small percentage of pediatric patients admitted to our hospital, BERT alerts are more likely to occur during admissions for Black patients and their families even after controlling for insurance status and other covariates, including MHD. To our knowledge, this disproportionate use in BERT alerts has not been previously described in the literature for pediatric inpatients.

Our study adds to existing literature by uncovering disproportional utilization of BERT alerts in the inpatient pediatric setting. It is unique in that the intervention being studied is intended to provide behavioral and emotional support to patients, families, and staff as compared to PRs and security-only responses. Despite BERTs intending to be supportive, inequitable deployment may suggest a bias in the perception of which patients and families require a behavioral intervention.

One objective of BERTs is to reduce restraint use.4 Overall, admissions for Black patients had higher PR use. Prior literature characterizes demographics of PR use in pediatric inpatient populations and demonstrates disproportional security-only responses in adult inpatient populations. This is similar to a prior study demonstrating racial disparities in PR use for pediatric patients in the emergency department but contrasts with a study demonstrating no difference in the race of patients who had PR use in a medical-behavioral unit.10,11 Further studies are needed to understand how BERTs can offset disparate PR and chemical restraint use in Black patients.9,13,14 

Differing from other systems that activate only for patients, our BERT system is activated for both patients and parental behavioral escalations. We found significant differences in BERT alerts, not only for Black patients but also for parents of Black patients. Our EMR collects race/ethnicity of patients but does not elaborate on the race/ethnicity of parents, which is similar to a majority of pediatric hospitals.15 Many challenges surround collection and documentation of pediatric patient demographic information,16 which extends beyond the scope of our study. Because of this limitation, we are unable to comment on whether the race of the parent in BERT alerts accounts for any differences.

Our study also reveals a greater use of BERT alerts when a patient had an MHD. Previous studies have identified as many as 80% of pediatric patients receiving BERT alerts have psychiatric comorbidity.17 We considered that disproportionately high rates of MHD in Black patients may contribute to higher rates of BERT alerts. However, in our population, MHD were lower for admissions of Black patients than of white patients. This is consistent with prior reports.18 Hospitalization is a stressful time for any patient/family, regardless of the presence of absence of MHD. Children hospitalized with acute mental illness receive tailored supports to help them with their illness and specifically attuned to trying to prevent behavioral escalations, whereas these supports might not be readily available for patients not identified with MHD. As mental health care needs expand, hospitals and inpatient teams must consider how to support patients with mental health diagnoses on their medical and ICU teams.

Done well, BERTs can be one way of addressing the growing mental health crisis in children and young adults, while potentially decreasing interventions with harmful effects (ie, PRs). From a patient or caregiver’s perspective, BERTs can be viewed as therapeutic or punitive, depending upon how the BERT is framed and implemented. BERTs may be viewed as a traumatic experience specifically when restraints are used. It is important that systems recognize that enforcement of safety in the United States takes the form of policing of Black individuals more often than white individuals. Therefore, careful consideration should be given to the involvement of security personnel in hospital BERT responses, as Black patients can have increased fear and higher mental health concerns after negative interactions with police.19 

Our study has significant implications for health care systems aiming to address disparities in health care outcomes and provides many opportunities for further investigation. It is important that when systems identify existing disproportionality and disparity in clinical processes, leaders formulate and conduct a concerted response to address and mitigate bias. Other authors have suggested creating and using consensus guidelines or standardized checklists to reduce disparities in PR use.13 Hospitals vary widely in how they organize their BERTs.20 Standardized processes and preventative measures implemented with an equity lens are needed to help address racial, diagnosis-specific, and socioeconomic disparities.21 Preventative measures may include needs-based screening, identification of high-risk psychiatric or medical conditions, or environmental modifications for high-risk patients. Although studies on patient-provider racial concordance have shown mixed findings,22 increasing diversity among health care staff may be another intervention that mitigates racial disparities. Addressing disparities within organizations requires a multitiered approach (as outlined in Supplemental Table 3). Potential next steps to expand this research may include quality improvement investigation of screening tools or environmental modifications. Qualitative investigation of the patient and parent experience of BERTs and BERT alerts can provide important stakeholder insight. Thoughtful use of equitable interventions can hopefully decrease disparities while keeping health care setting safe.

This study has several limitations. Each hospital’s BERT system differs in its parameters for calling an alert, team structure, team training, and scope of interventions. In addition, the demographics of patient populations and staff members caring for them may differ between institutions. These factors impact the generalizability of our findings to other institutions. We aim to address this limitation with future studies that examine institutional variations in BERT alerts. We relied on demographic information that is self-reported upon registration and recorded in our EMR and could not determine demographics of the parents, as these are not routinely collected upon hospital registration. Disconcordance between EMR- and self-reported race and ethnicity is more likely for minority races and ethnicities,23,24 although there may be higher concordance in pediatrics.23 During the time of this study, our EMR only allowed for binary responses (male, female) for gender. As there was a low total number of PRs documented, our findings may be difficult to generalize. Future studies should include physical and chemical restraint usage in larger populations across multiple institutions.

BERT alerts in the inpatient setting at one pediatric hospital occurred more often during admissions for Black patients and parents of Black patients, patients with government-funded insurance, and patients with mental health diagnoses. These findings may indicate that unconscious bias and/or structural factors leading to bias exist in the use of BERT alerts for admissions to the general inpatient pediatric setting. Further work is needed to understand and address disproportional use of this resource and work to address disparities in the care of patients with BERT alerts.

Drs Peterson, Kim, Krause, and Amaniampong participated in the concept and design of this study, analysis and interpretation of data, and drafting of the original manuscript. Drs Fauntleroy, LaMotte, Johnson, and Todd contributed to the analysis of data, as well as the drafting of the manuscript. Mr Perkins contributed to analysis and interpretation of data, as well as revising of the manuscript. Drs Cox and Tucker-Edmonds contributed to revision of the manuscript. All authors were responsible for the revising of the final manuscript and have approved the manuscript as submitted.

CONFLICT OF INTEREST DISCLOSURES: All authors report no conflicts of interest to disclose.

FUNDING: No funding was secured for this study.

COMPANION PAPER: A companion to this article can be found online at www.hosppeds.org/cgi/doi/10.1542/hpeds.2024-007775.

The authors would like to thank the integrated case management team, the nursing staff, and the security teams at Riley Hospital for Children at Indiana University Health for their contributions to collecting data and reviewing the current processes for behavioral alerts. We would also like to thank Shawnette Bellamy for her work in advancing diversity, equity, and inclusion at Riley. We additionally would like to thank Dr Katherine Auger for her contributions to providing edits to the final manuscript.

BERT

behavioral emergency response team

PR

physical restraint

aOR

adjusted odds ratio

MHD

mental health diagnosis

ICU

intensive care unit

FIN

Financial Identification Number

EMR

electronic medical record

1
Parker
CB
,
Calhoun
A
,
Wong
AH
,
Davidson
L
,
Dike
C
.
A call for behavioral emergency response teams in inpatient hospital settings
.
AMA J Ethics.
2020
;
22
(
11
):
E956
E964
. PubMed doi: 10.1001/amajethics.2020.956
2
Derscheid
DJ
,
Arnetz
JE
.
Patient and family member violent situations in a pediatric hospital: a descriptive study
.
J Pediatr Nurs.
2020
;
55
:
241
249
. PubMed doi: 10.1016/j.pedn.2020.07.014
3
Loucks
J
,
Rutledge
DN
,
Hatch
B
,
Morrison
V
.
Rapid response team for behavioral emergencies
.
J Am Psychiatr Nurses Assoc.
2010
;
16
(
2
):
93
100
. PubMed doi: 10.1177/1078390310363023
4
Zicko
CJM
,
Schroeder
LRA
,
Byers
CWS
,
Taylor
LAM
,
Spence
CDL
.
Behavioral emergency response team: implementation improves patient safety, staff safety, and staff collaboration
.
Worldviews Evid Based Nurs.
2017
;
14
(
5
):
377
384
. PubMed doi: 10.1111/wvn.12225
5
Nicome
R
,
Lo
HY
,
Gupta
S
, et al
.
Utilizing a behavioral health bundle to improve patient and clinician safety for hospitalized children
.
Pediatr Qual Saf.
2021
;
6
(
2
):
e393
. PubMed doi: 10.1097/pq9.0000000000000393
6
Pestka
EL
,
Hatteberg
DA
,
Larson
LA
,
Zwygart
AM
,
Cox
DL
,
Borgen
EE
Jr.
Enhancing safety in behavioral emergency situations
.
Medsurg Nurs.
2012
;
21
(
6
):
335
341
. PubMed
7
Dahnke
D
,
Mulkey
MA
.
Using a behavioral response team on non-psychiatric nursing units
.
Medsurg Nurs.
2021
;
30
(
4
):
229
234
. PubMed
8
Valtis
YK
,
Stevenson
KE
,
Murphy
EM
, et al
.
Race and ethnicity and the utilization of security responses in a hospital setting
.
J Gen Intern Med.
2023
;
38
(
1
):
30
35
. PubMed doi: 10.1007/s11606-022-07525-1
9
Nash
KA
,
Tolliver
DG
,
Taylor
RA
, et al
.
Racial and ethnic disparities in physical restraint use for pediatric patients in the emergency department
.
JAMA Pediatr.
2021
;
175
(
12
):
1283
1285
. PubMed doi: 10.1001/jamapediatrics.2021.3348
10
Noah
A
,
Andrade
G
,
DeBrocco
D
, et al
.
Patient risk factors for violent restraint use in a children’s hospital medical unit
.
Hosp Pediatr.
2021
;
11
(
8
):
833
840
. PubMed doi: 10.1542/hpeds.2020-000273
11
Tang
AS
,
Shieh
MS
,
Pekow
PS
,
Prentiss
KA
,
Lindenauer
PK
,
Westafer
LM
.
Treatment of pediatric behavioral health patients with intravenous and intramuscular chemical restraints: results from a nationwide sample of emergency departments
.
Acad Emerg Med.
2023
;
30
(
10
):
1029
1038
. PubMed doi: 10.1111/acem.14754
12
Mental health disorder codes
. Children’s Hospital Association. Updated December 8,
2023
. Accessed August 22, 2024. https://www.childrenshospitals.org/content/analytics/toolkit/mental-health-disorder-codes
13
Foster
AA
,
Porter
JJ
,
Monuteaux
MC
, et al
.
Disparities in pharmacologic restraint use in pediatric emergency departments
.
Pediatrics.
2023
;
151
(
1
):
e2022056667
. PubMed doi: 10.1542/peds.2022-056667
14
Carreras Tartak
JA
,
Brisbon
N
,
Wilkie
S
, et al
.
Racial and ethnic disparities in emergency department restraint use: a multicenter retrospective analysis
.
Acad Emerg Med.
2021
;
28
(
9
):
957
965
. PubMed doi: 10.1111/acem.14327
15
Cowden
JD
,
Flores
G
,
Chow
T
, et al
.
Variability in collection and use of race/ethnicity and language data in 93 pediatric hospitals
.
J Racial Ethn Health Disparities.
2020
;
7
(
5
):
928
936
. PubMed doi: 10.1007/s40615-020-00716-8
16
Tan-McGrory
A
,
Bennett-AbuAyyash
C
,
Gee
S
, et al
.
A patient and family data domain collection framework for identifying disparities in pediatrics: results from the pediatric health equity collaborative
.
BMC Pediatr.
2018
;
18
(
1
):
18
. PubMed doi: 10.1186/s12887-018-0993-2
17
Mitchell
M
,
Newall
F
,
Williams
K
.
Behavioural emergencies in a paediatric hospital environment
.
J Paediatr Child Health.
2022
;
58
(
6
):
1033
1038
. PubMed doi: 10.1111/jpc.15896
18
McGuire
TG
,
Miranda
J
.
New evidence regarding racial and ethnic disparities in mental health: policy implications
.
Health Aff (Millwood).
2008
;
27
(
2
):
393
403
. PubMed doi: 10.1377/hlthaff.27.2.393
19
Addison
HA
,
Richmond
TS
,
Lewis
LM
,
Jacoby
S
.
Mental health outcomes in formerly incarcerated Black men: a systematic mixed studies review
.
J Adv Nurs.
2022
;
78
(
7
):
1851
1869
. PubMed doi: 10.1111/jan.15235
20
Rajwani
A
,
Clark
N
,
Montalvo
C
.
Understanding best practices in implementation of behavioral emergency response teams through a scoping review
.
J Am Psychiatr Nurses Assoc.
2023
;
29
(
5
):
375
388
. PubMed doi: 10.1177/10783903221114335
21
Liang
D
,
House
SA
,
Moriates
C
.
Improving healthcare value: the need to explicitly address equity in high-value care
.
J Hosp Med.
2024
;
19
(
4
):
316
319
. PubMed doi: 10.1002/jhm.13280
22
Cheng
AW
,
Nakash
O
,
Cruz-Gonzalez
M
,
Fillbrunn
MK
,
Alegría
M
.
The association between patient-provider racial/ethnic concordance, working alliance, and length of treatment in behavioral health settings
.
Psychol Serv.
2023
;
20
(
suppl 1
):
145
156
. PubMed doi: 10.1037/ser0000582
23
Samalik
JM
,
Goldberg
CS
,
Modi
ZJ
, et al
.
Discrepancies in race and ethnicity in the electronic health record compared to self-report
.
J Racial Ethn Health Disparities.
2023
;
10
(
6
):
2670
2675
. PubMed doi: 10.1007/s40615-022-01445-w
24
Owosela
BO
,
Steinberg
RS
,
Leslie
SL
, et al
.
Identifying and improving the “ground truth” of race in disparities research through improved EMR data reporting. A systematic review
.
Int J Med Inform.
2024
;
182
:
105303
. PubMed doi: 10.1016/j.ijmedinf.2023.105303

Supplementary data