Authors: Max J. Rolison, Motunrayo Adu, Isaac V. Faustino, Anusha Kumar, Shiqi Huang, Emily Powers, Veronika Shabanova, Ambrose H. Wong, Pamela Hoffman, Gunjan Tiyyagura
Categories: Brief Report, restraints, behavioral health, psychiatry, child protective services, autism
Source: Journal of the American College of Emergency Physicians Open
Authors: Max J. Rolison, Motunrayo Adu, Isaac V. Faustino, Anusha Kumar, Shiqi Huang, Emily Powers, Veronika Shabanova, Ambrose H. Wong, Pamela Hoffman, Gunjan Tiyyagura
To examine how structural factors, such as child protective services (CPS) involvement, prehospital interactions with police or emergency medical services (EMS), and clinical factors, such as autism diagnosis, contribute to physical restraint use among pediatric patients presenting to the emergency department (ED) for behavioral health concerns.
In this retrospective cohort study, we reviewed pediatric ED encounters from January 1, 2021, to October 31, 2023, at a tertiary care children’s hospital. Multivariable logistic regression was used to assess associations among autism diagnosis, CPS involvement, and arrival mode (police/EMS) and physical restraint use, adjusted for demographic variables.
Among 6288 behavioral health encounters, physical restraints were used in 124 (1.97%; 95% CI, 1.69, 2.58) encounters. Children arriving by police or EMS were 3 times more likely to be restrained than those arriving by car or walk-in (adjusted odds ratio, aOR = 3.07, 95% CI, 2.01-4.69). Children with CPS involvement were almost twice as likely to be restrained (aOR = 1.91; 95% CI, 1.26-2.88). Children diagnosed with autism were 7 times more likely to be restrained (aOR = 7.25, 95% CI, 3.61-14.55). Black children were more likely to be restrained than White children (aOR = 1.78, 95% CI, 1.12-2.84).
CPS involvement, transport by police or EMS, autism diagnosis, and Black race were independently associated with increased physical restraint use in pediatric ED patients. These findings emphasize the role of both structural and child-level factors in contributing to physical restraint in emergency behavioral health care, highlighting the need for a multifactorial approach to reduce restraint use.
The Bottom LineThe use of physical restraints in pediatric emergency departments is not well described. In this retrospective cohort study of 6288 behavioral health encounters from 2021 to 2023, restraints were used in approximately 2% of visits. Using multivariable logistic regression, we found that children transported by police or emergency services had threefold higher odds of restraint. Involvement with child protective services nearly doubled the odds, autism diagnosis increased the odds more than sevenfold, and Black children were more likely to be restrained than White children. These findings underscore the structural and clinical factors contributing to restraint use and the need to address disparities.
Children frequently present for emergency psychiatric care due to acute agitation and emotional outbursts.^1^ Best practice guidelines recommend using behavioral de-escalation strategies to manage acute agitation and avoid restraints,^1^ yet approximately 6% to 10% of youth seeking psychiatric care in emergency departments (EDs) are restrained.^2^^,^^3^ Although physical restraints are employed to maintain safety for youth harming themselves or others, they can lead to injuries to both staff and patients, psychological distress, and, in rare cases, death.^1^
Prior work in one health system demonstrated that Black children were 1.8 times more likely to be physically restrained than White children. Additionally, males were 1.95 times more likely to be physically restrained than females.^3^ In adults, studies have demonstrated that structural factors like absence of private insurance, police transport, and homelessness, as well as specific clinical risk factors like substance use disorders, are also associated with increased odds of physical restraint use.^4^ It remains necessary to identify unique clinical and structural factors that contribute to restraint use in the pediatric ED. Clinical conditions, such as autism spectrum disorder (ASD), may make children particularly vulnerable to restraint use in the pediatric ED, given that ED settings and health care workers are commonly ill-equipped to care for their unique needs.^5^^,^^6^ Additionally, it remains unknown how EMS or police transport influences pediatric restraint use.
There has been an increase in the number of pediatric patients presenting to emergency departments for psychiatric concerns over the past decade, with increasing attention and scrutiny paid to the use of physical restraints.^7^ Identifying early signs of agitation in an ED can be challenging, sometimes leading to the need for physical restraints. Understanding the factors that increase restraint use is essential for system-level interventions.
The aim of this study was to explore clinical and structural factors associated with physical restraint use in children. Our hope is that identification of these factors can deepen our understanding of pathways to restraint use in the pediatric ED setting and inform preventive strategies for higher-risk children presenting to the emergency care setting.
In this retrospective cohort study, we identified and reviewed all pediatric ED encounters with behavioral health conditions (defined as the presence of mental health, intoxication, and/or poisoning-related chief concerns or diagnostic codes associated with psychiatric disorders, intoxications, or ASD) from the electronic health record (EHR) between January 1, 2021, and October 31, 2023, at a tertiary care children’s hospital (Table 1). The Yale University Human Investigation Committee considered this study exempt from IRB review.Table 1Behavioral health presentations based on chief complaint.Chief complaintsPsychiatric evaluationAggressive behaviorSuicidalAgitationBehavior problemCombativeHomicidalHallucinationsSuicide attemptPsychotic symptomsAnxietySelf-mutilationDepressionManic behaviorPanic attackMental health problemHyperventilatingTicsParanoidPsychosisRunawayPoisoningAlcohol intoxicationOverdose-intentionalDrug problemOverdose-accidentalDrug/alcohol assessmentAddiction problem
This study was conducted in the pediatric ED at a tertiary care children’s hospital in New England. The pediatric ED has an annual volume of approximately 38,979 children, with approximately 1849 annual visits of children younger than 16 years old presenting with behavioral health concerns. Our pediatric ED has a care pathway for managing acute agitation, emphasizing verbal de-escalation and limiting physical restraints to situations of imminent harm.
We included all patients 5 years of age or older with a behavioral health encounter in the pediatric ED. In this health system, children under 16 seek behavioral health care in the pediatric ED, whereas those 16 and older go to the adult psychiatric ED, unless there is a medical concern (eg, overdose, intoxication). For this study, we included all patients evaluated in the pediatric ED.
We a priori identified primary exposure variables, including clinical (autism diagnosis) and visit-level variables (child protective services [CPS] involvement and means of arrival [car/walk-in or police/EMS]), with demographic variables as covariates (child’s age, legal sex, race and ethnicity, primary language, and insurance status). Autism diagnosis was determined by “autism” or “ASD” appearing on the patient’s list of active medical conditions or visit diagnosis. CPS involvement was defined as being under CPS custody (identified when CPS was listed as the emergency contact) or having a CPS report filed during the encounter (documented in social work flowsheets). EMS or police transport to the ED was based on the structured field “arrival method” captured in the electronic medical record. To validate the results of the “arrival method” responses, we performed a manual chart review and identified that police were commonly involved even in cases where the structured field assessing transportation was marked for EMS only; therefore, the categories for police and EMS were collapsed. Race and ethnicity classification reflected either self-identification or hospital registration assignment.
Categoric variables were summarized as count (percentage, %) and continuous variables as mean (standard deviation, SD). Categoric variables included legal sex (male/female), arrival method (car or walk-in/police and EMS/not specified), CPS involvement (no/yes), autism diagnosis (no/yes), insurance (private/public), and race/ethnicity (White non-Hispanic/Black non-Hispanic/Hispanic/Asian/other/unknown).
Physical restraint use was defined based on the presence of an order for violent physical restraints during an ED visit. At our institution, this order type follows the Centers for Medicare and Medicaid Services definition and can only be entered when a patient poses an imminent risk of harm to self or others.
Descriptive statistics included means with standard deviations for normally distributed continuous variables and frequencies with percentages for categoric variables. We used marginal multivariable logistic regression models with robust standard errors via the sandwich estimator to examine the association of exposure variables (autism diagnosis, arrival method, and CPS involvement) with our primary outcome and physical restraint use.^8^ Given the rarity of the outcome, results were summarized as adjusted odds ratios (aORs), which approximated risk ratios, with surrounding robust 95% CI. We assessed model fit using an extension of the Hosmer-Lemeshow test and evaluated multicollinearity with variance inflation factors (all ≈1). To assess the independent and combined effects of exposures, we used a manual stepwise approach. We first constructed separate models for each exposure of interest (autism diagnosis, CPS involvement, arrival method) adjusted for demographics. Next, we built pairwise models including 2 exposures at a time. Finally, we fit a combined model including all 3 exposures (Table S1). This approach allowed us to examine the consistency of associations across domains and evaluate whether effect sizes shifted when exposures co-occurred. We performed an additional sensitivity analysis, excluding patients older than 16 years of age, because they are not universally evaluated in the pediatric ED, ensuring our results were unbiased (Table S2).
Among 6288 behavioral health encounters involving 3860 unique individuals, 1629 were male (42.20%), with an average age of 12.7 years (5-24 years, SD 3.0), and 1051 (27.23%) patients had at least 2 behavioral health encounters in the pediatric ED. Physical restraints were used in 124 encounters (1.97%, 95% CI, 1.69, 2.58) (Table 2). In the study sample, 2674 encounters involved police or EMS (42.53%), 101 (1.61%) encounters involved children with ASD, 913 (14.52%) encounters involved children with CPS involvement, 387 (6.15%) encounters involved non-English speakers, and 2453 (39.01%) encounters involved individuals with public insurance.Table 2Demographic and visit characteristics of pediatric ED visits by the presence of a physical restraint order.VariableRestraint used (N = 124)No restraint used (N = 6164)Mean age, y (SD)12.33 (2.69)12.87 (2.78)Legal sex, n (%) Female61 (49.19%)3589 (58.23%) Male63 (50.81%)2575 (41.77%)Race/ethnicity, n (%) White non-Hispanic36 (29.03%)2628 (42.63%) Black non-Hispanic48 (38.71%)1423 (23.09%) Hispanic40 (32.26%)1798 (29.17%) Asian0 (0.00%)112 (1.82%) Othera0 (0.00%)18 (0.29%) Unknown0 (0.00%)185 (3.00%)Primary language, n (%) English speaking121 (97.58%)5780 (93.77%) Non-English Speaking3 (2.42%)384 (6.23%)Insurance, n (%) Private insurance30 (24.19%)2423 (39.31%) Public insurance94 (75.81%)3741 (60.69%)Arrival method, n (%) Car/walk-in30 (24.19%)3251 (52.74%) Police and EMSb90 (72.58%)2584 (41.92%) Not specified4 (3.23%)329 (5.34%)CPS involvement, n (%)38 (30.65%)875 (14.20%)Autism diagnosis, n (%)10 (8.06%)91 (1.48%)ED length of stay, hours (SD)47.81 (75.55)20.97 (31.61)Patients admitted, n (%)70 (56.45%)1938 (31.44%)Abbreviations: ED, emergency department; CPS, child protective services; EMS, emergency medical services; SD, standard deviation.aOther comprises Native American, Native Hawaiian or Pacific Islander, and others.b62% had prehospital police and EMS involvement, 37% arrived with EMS only, and 1% arrived with police only.
Children who were restrained had longer lengths of stay (restrained 47.81 h, SD 75.55; no restraints 20.97 h, SD 31.61, P < .001) and were more frequently admitted than those who were not restrained (56.4% versus 31.4%, P < .001).
Twenty-six (0.44%) children were restrained before arrival at the ED. These restraint decisions did not involve ED physicians. They comprised 20.8% of children for whom physical restraints were also applied in the pediatric ED. Only restraint episodes occurring within the pediatric ED were included in the analysis.
Within the fully adjusted model, children arriving by police or EMS were 3 times more likely to be restrained than those arriving by car or walk-in (Table 3; aOR = 3.07, 95% CI, 2.01, 4.69), and children with CPS involvement were almost 2 times more likely to be restrained (aOR = 1.91, 95% CI, 1.26, 2.88) than those without CPS involvement. Children with autism were over 7 times more likely to be restrained (aOR = 7.25, 95% CI, 3.61, 14.55) than those without autism diagnoses. Insurance status was not significantly associated with restraint use in the multivariable model. Black children were more likely to be restrained than White children (aOR = 1.78, 95% CI, 1.12, 2.84). Individual models for each exposure demonstrated similar results (Table S1). Sensitivity analyses excluding patients >16 years of age also demonstrated consistent findings (Table S2).Table 3Results from multivariable adjusted logistic model for associations with restraint usea (N = 6288 encounters in n = 3860 individuals)CharacteristicObserved, N (%)Restraint useEstimated restraint use, % (95% CI)baOR (95% CI)Arrival method Car/walk-in30 (0.91)0.89 (0.62, 1.25)Ref. Police and EMS90 (3.37)2.68 (2.08, 3.42)3.07 (2.01, 4.69) Not specified4 (1.20)1.13 (0.39, 2.62)1.13 (0.39, 3.25)CPS involvement No86 (1.60)1.3 (1.01, 1.65)Ref. Yes38 (4.16)2.47 (1.67, 3.50)1.91 (1.26, 2.88)Autism diagnosis No114 (1.84)1.38 (1.08, 1.76)Ref. Yes10 (9.9)9.76 (4.96, 17.68)7.25 (3.61, 14.55)Insurance Private30 (1.22)0.85 (0.55, 1.28)Ref. Public94 (2.45)0.93 (0.6, 1.37)1.10 (0.70, 1.72)Race/ethnicity White non-Hispanic36 (1.35)1.22 (0.85, 1.72)Ref. Black non-Hispanic48 (3.26)2.16 (1.50, 2.96)1.78 (1.12, 2.84) Hispanic40 (2.18)1.51 (1.06, 2.13)1.26 (0.78, 2.04) Asian0 (0.00)NENE Other0 (0.00) Unknown0 (0.00)Legal sex Female61 (1.67)1.30 (0.96, 1.72)Ref. Male63 (2.39)1.62 (1.19, 2.16)1.24 (0.86, 1.79)AgeN/A2.31 (0.94, 4.70)0.97 (0.91, 1.04)aOR, adjusted odds ratio; NE, not estimable due to 0 count.aGoodness of model fit was examined using an extension of the Hosmer-Lemeshow test to marginal regression models (P > .05) and multicollinearity with variance inflation factors, all approximately 1. See Table S1 for intermediate models and discussion in the footnote.bPredicted percentage restrained from the adjusted logistic regression model were estimated using marginal means (on the scale of the outcome via an inverse logit) with 95% CI estimated using robust standard errors using the sandwich estimator. Percentages and confidence intervals for a given covariate were calculated by using the average value of the other covariates.
Several limitations exist for our study. First, due to the limitations of EHR data, race and ethnicity data, or CPS involvement may not be fully accurate, and we could not reliably capture attempts at verbal de-escalation or offers of oral medications. Second, CPS involvement is limited to active custody or referrals made during the encounter and may not fully capture the population of children interacting with CPS. Third, we were unable to reliably separate instances where EMS or police were involved because they commonly co-occurred. Fourth, these results do not clarify all potential mechanisms leading to increased restraint use, such as differences in acuity of presentation, the range of intellectual delay, perception of safety concerns, or whether supportive adults were bedside. Finally, due to the cross-sectional nature of this study, we cannot establish causality between our predictors and outcome measure, only associations. However, these findings could serve as the basis for future studies designed to examine causality. Additionally, these findings are based on findings from a single site, which may limit generalizability.
Among children presenting with behavioral health problems to one pediatric ED, we found that CPS involvement, transport by EMS or police, and autism diagnosis were independently associated with a greater odds of being physically restrained. Among demographic variables, Black race was the only independent predictor.
Our findings demonstrate that prehospital interaction with EMS and law enforcement is associated with physical restraint use in the ED for pediatric patients. Violent or threatening behavior, prior to arrival at the ED, may have prompted caregivers to call emergency services to transport a child who presented as a danger to themselves or others, rendering physical restraint possibly necessary, and may reflect a higher level of clinical acuity. However, in this analysis, we focused exclusively on patients who were restrained in the ED. In the ED, patients transported by emergency medical services (EMS) or police may be perceived as more dangerous and bias health care workers to restrain patients. Further, being brought in by police or EMS could itself be distressing and lead to behavioral escalation, leading to restraint use. Law enforcement involvement may also reflect or contribute to greater exposure to adverse childhood experiences (ACEs).^9^ ACEs have been associated with increased mental and physical health problems throughout the lifespan,^10^ as well as greater externalizing behaviors that may lead to the use of physical restraints.
In our study, CPS involvement was also associated with increased physical restraint use in the ED. Children with CPS involvement have experienced abuse or neglect and higher rates of trauma-related mental health disorders,^11^ which can contribute to episodes of agitation requiring restraint use. They may also be receiving care that is fractured and inconsistent, potentially leading to missed doses of medications or lack of effective treatment.^12^ Furthermore, involvement with CPS may, in itself, reflect and contribute to increased exposure to ACEs, such as abuse and neglect.^13^
Children with autism were significantly more likely to be physically restrained. Autistic children more commonly present to the ED for behavioral health concerns, and most health care workers are not adequately trained or resourced to provide appropriate care for this population.^5^ Further, the physical infrastructure of the ED with bright lights and loud sounds can often be overwhelming for some autistic individuals, which may contribute to behavioral dysregulation.^6^ Finally, communication differences for children with autism may make it more difficult to effectively engage in behavioral de-escalation strategies.^6^ In inpatient units, specialized care pathways and staff training have been effective in reducing physical restraints.^14^ Future work should focus on improving health care worker training and resources to more adequately care for this specific high-risk pediatric population in order to reduce physical restraint use. Utilizing standardized ratings of agitation to begin to understand differences in clinical decision making and outcomes based on acuity is an important next step.
Among children’s demographic variables, consistent with other studies,^3^ Black race was associated with higher odds of being physically restrained during an ED visit. Public insurance status was not associated with greater odds of restraint in the adjusted model. Although public insurance is often a proxy for poverty and has been associated with restraint use in other studies,^3^ other unmeasured variables such as family support, educational level, and access to social services, might be more closely related to restraint use in the ED. Therefore, it is possible that adjusting for demographic factors such as Black race and clinical factors such as CPS involvement may have attenuated the association between insurance and restraint use. Although poverty is often associated with poorer mental health outcomes, it may not confer the same risk for agitation or aggression that would lead to a physical restraint in the ED.
These findings provide the first evidence of greater odds of physical restraint use in the ED among children involved with CPS. Similar to adults, EMS or police transport to the hospital was also associated with a greater risk of restraint use among pediatric patients. Given the disproportionate involvement of Black youth with CPS and law enforcement,^15^^,^^16^ future work is needed to elucidate how these prehospital factors mediate racialized health inequity in pediatric restraint use and focus on the development of specific interventions to address these factors. Additionally, efforts should be made to improve emergency care for children with ASD and other neurodevelopmental disabilities. Future work should also consider how various factors, including those in this study, may be associated with the duration of time children are restrained.
Dr Max J. Rolison conceptualized and designed the project, interpreted and analyzed the data, and wrote the initial manuscript.
Dr Motunrayo Adu conceptualized and designed the project.
Isaac V. Faustino, Shiqi Huang, and Anusha Kumar performed the data analysis.
Dr Emily Powers performed the data collection from the electronic health record.
Dr Veronika Shabanova assisted with the data analysis and interpretation, as well as reviewed and edited the final manuscript.
Dr Ambrose H. Wong conceptualized and designed the project, assisted with the data analysis and interpretation, and reviewed and edited the final manuscript.
Dr Pamela Hoffman conceptualized and designed the project and interpreted the results.
Dr Gunjan Tiyyagura provided project supervision, conceptualized and designed the project, analyzed data and interpreted results, and reviewed and edited the final manuscript.
All authors reviewed and approved the submitted manuscript.
10.13039/100016899Yale-New Haven Hospital Health Equity, Patient Safety, and Quality Improvement Grant (Adu), and Yale Child Study Center Pilot Grant (Rolison). This work was supported in part by funds from the 10.13039/100000071National Institute of Child Health and Human Development K23 HD107178 (Tiyyagura), the 10.13039/100000025National Institute of Mental Health R25 MH077823, T32 MH18268 (Rolison), the National Institute of Mental Health K23MH126366, R01MH132605 (Wong), and the Patient-Centered Outcomes Research Institute BPS-2022C3-30381 (Wong). The contents of the manuscript are solely the responsibility of the authors and do not necessarily represent the official view of the NIH or PCORI.
All authors have affirmed they have no conflicts of interest to declare.