Physical and Pharmacologic Restraint in Hospitalized Children With Autism Spectrum Disorder

This was a retrospective cohort study of all pediatric patients aged ≥5 to ≤21 years admitted to pediatric inpatient floors, including medical, intensive, subspecialty, and surgical care floors, in a large urban pediatric hospital over 5 years (October 1, 2016–October 1, 2021). Previous studies in psychiatric units assessing restraints use a variety of age ranges, with several selecting a lower limit of 5 years.²⁴  Age 21 is the cutoff for admission to Boston Children’s Hospital with minimal exceptions for young adults with certain chronic conditions. Psychiatric admissions and ED visits were excluded. Psychiatric units have different policies and procedures regarding behavioral management and different unit configurations compared with inpatient pediatric units. ED visits tend to be brief and follow different protocols than inpatient settings. This strengthens external validity and generalizability because all hospitals that care for children in an inpatient setting do not have dedicated pediatric EDs or pediatric inpatient psychiatric units. Children admitted to inpatient pediatric units awaiting psychiatric placement were included. The study was approved by the institutional review board of Boston Children’s Hospital.

We collected demographic data on all children hospitalized on inpatient floors, including age, sex, race and ethnicity, and primary spoken language. International Classification of Diseases, 10th Revision, codes were used to identify ASD and cooccurring conditions. All children were identified on the basis of billing diagnoses associated with at least 1 hospital encounter during the study period. Codes are listed in Supplemental Table 4. We extracted year of hospitalization and length of stay. We identified the reason for admission entered by the medical care team for each hospitalization and coded these reasons as either medical or behavioral health (Supplemental Table 4). We extracted cooccurring mental health, behavioral health, and developmental conditions other than ASD (Supplemental Table 4).

The primary outcome was involuntary physical or pharmacologic restraint because of violent or self-injurious behavior. In this article, restraint refers to restraint for violent or self-injurious behavior. Involuntary restraint is categorized as either violent or nonviolent. Nonviolent restraints, a protective intervention to support medical or surgical care and healing, such as to avoid removing a line, were excluded. A clinical order for restraint for violent or self-injurious behavior associated with at least 1 hospital encounter during the study period classified a subject as having experienced restraint. Within this clinical order, clinicians indicate whether restraint was physical or pharmacologic. Physical restraint was defined as a clinical order indicating physical restraint including a cuff, physical restraint/hold, seclusion, or enclosure bed. Although some children with ASD use enclosure beds at home, use of this equipment paired with a restraint order will capture its use as an involuntary violent restraint and exclude use as part of existing care. Pharmacologic restraint is defined as a clinical order for olanzapine, haloperidol, or lorazepam paired with a restraint for violent or self-injurious behavior order. These medications are typically administered intramuscularly in the context of restraint. Oral administration, if paired with the appropriate clinical order, was also considered pharmacologic restraint. Consistent with the Joint Commission recommendations,¹¹  hospital policy requires a new order for physical restraint to be placed every hour, after face-to-face evaluation of the patient by the ordering clinician. Pharmacologic restraint must be a 1-time order, with a new order placed for each medication administration intended as pharmacologic restraint. Because restraint was a relatively rare event, we combined physical and pharmacologic restraint into a single variable of restraint.

Univariate analyses compared children with and without ASD with independent samples t test, assuming normality and equality of variances. χ² tests assessed differences in frequencies for categorical variables. An unadjusted odds ratio (OR) assessing the relationship between ASD and experiencing restraint was obtained using the omnibus χ² test. An assessment of whether the incidence of restraint in any child, restraint in children with ASD, and hospitalization in children with ASD increased during the study period was performed with tests of level using the analysis of variance model, with year as the independent variable. If the omnibus F-test was significant, Tukey’s post hoc tests were engaged. The level of significance was 5% for a 2-tailed test.

To ensure differences in restraint were not a function of ASD demographics, 2 equivalent samples on key variables were created using PSM via the R package Matchit.²⁵  Matching adjusted for differences in age, number of visits, sex, race and ethnicity, primary language, and year of admission between ASD and non-ASD groups using the nearest neighbor method. Out of the 2040 ASD cases, 2027 were included in PSM, (N = 13 excluded for missing primary language) and were matched with non-ASD cases using 1:1 matching. If a child was restrained during any hospitalization, only the restrained hospitalizations were included to avoid the same individual representing a restrained and nonrestrained subject. If a nonrestrained child had multiple hospitalizations, the first hospitalization was selected to ensure nonbiased selection and a specific child being overrepresented in the sample. Table 1 shows mean/percentage estimates using the unmatched data documenting significant differences in age, sex, race and ethnicity, language, and year of hospitalization, followed by nonsignificant differences with the matched data (except for hospitalization likely reflecting a type-I error).

Logistic regression analysis was performed with the primary binary outcome variable of restraint/no-restraint adjusted for age, including predictors of reason for admission, length of stay, and presence of a mental health disorder, a developmental disorder, or a behavioral disorder. The level of significance was 5% for a 2-tailed test. Models were analyzed using R and Stata.

A total of 21 275 unique patients were hospitalized between October 2016 and October 2021, with a total of 35 270 hospital encounters included in the analysis. The number of hospitalizations for any individual ranged from 1 to 40. Of the 21 275 unique patients, 1187 (5.6%) had a diagnosis of ASD and 367 (1.7%) experienced restraint. Of the 367 patients who experienced restraint, 88 (24.0%) had a diagnosis of ASD. The number of hospital encounters identified of children with ASD did not vary significantly over the study period using a χ² test (P = .22).

Of the 35 270 hospital encounters, 927 (2.6%) were identified where a child experienced physical or pharmacologic restraint for violent or self-injurious behavior. Children may have been restrained during >1 hospital encounter. The overall number of restraints for an individual child ranged from 1 to 95. There was 1 restraint for violent or self-injurious behavior order placed in 491 (52.9%) of the hospital encounters with children experiencing restraint, and 847 (91.4%) of the hospital encounters of children experiencing restraint had 10 or less restraint orders. The number of hospital encounters identified of any child who experienced restraint significantly varied over the study period (P < .001), with 2016 having significantly fewer encounters compared with all other years, likely because of the length of time of data collection during 2016. No significant differences were observed when contrasting other years.

Children with ASD had a total of 2040 (5.8%) hospital encounters. Of the 927 hospital encounters with associated restraint for violent or self-injurious behavior orders, 192 (20%) of these encounters were in children with ASD. The number of hospital encounters identified in children with ASD who experienced restraint varied significantly over the study period (P < .001), only in that 2016 had fewer encounters than all other years, likely given shortest duration of data collection. No other significant differences between years existed.

Bivariate unadjusted analyses showed children with ASD were predominantly male, white, and reported English as primary spoken language (Table 1). Children with ASD had longer lengths of stay (205.37 vs 167.36 hours, P < .001); were more likely to have additional mental health (32.5% vs 21.6%, P < .001), behavioral health (27.5% vs 8.0%, P < .001), and other developmental disorders aside from ASD (51.5% vs 16.5%, P < .001); were more likely to have multiple hospitalizations during the study period (57.6% vs 54.6%, P < .01); and were more likely to be admitted for behavioral health reasons (9.1% vs 2.7%, P < .001) than children without ASD. Children with ASD were more likely to experience restraint for violent or self-injurious behavior during any hospitalization during the study period (9.4% vs 2.2%, P < .001). Of the restrained children, children with ASD had higher average numbers of restraint orders than children without ASD (9.47 vs 2.30, P < .001).

The ASD and non-ASD matched groups were adequately matched across the matching variables without significant mean differences regarding age, sex, race and ethnicity, primary language, or year of hospitalization. The matched sample using PSM is described in Supplemental Table 5.

The unadjusted OR for children with ASD who experienced restraint for violent or self-injurious behavior compared with children without ASD was 4.4 (P < .001, 95% confidence interval 3.2–6.1) (Table 2).

The adjusted OR estimate for the presence of variables, reason for admission, length of stay, and the presence of mental health, developmental, and behavioral disorders, was 2.3 (P < .001, 95% confidence interval 1.6–3.4) (Table 3). A behavioral health reason for admission, a mental health disorder, a developmental disorder other than ASD, and a behavioral disorder were noted to have an independent association with odds of restraint for violent or self-injurious behavior.

This is 1 of the first studies to examine rates of restraint in inpatient hospitalized children with ASD. Children with ASD were more than twice as likely as nonautistic children to experience involuntary physical or pharmacologic restraint for violent or self-injurious behavior after accounting for demographics, characteristics of hospital encounter, and cooccurring diagnoses. Children with ASD had longer lengths of stay; were more likely to have cooccurring mental health, developmental, and behavioral disorders; were more likely to have multiple hospitalizations; and were more likely to be admitted for behavioral health reasons than children without ASD.

Previous studies examining ASD in the context of inpatient pediatric hospital admissions have focused on identifying risk factors for behavioral dysregulation²⁶  or strategies to approach the general care of children with ASD during hospitalization.⁹  One study showed that children with ASD experiencing agitation during inpatient pediatric hospitalization were more likely to experience restraint than children with ASD who were not agitated,²⁶  though there was no typically-developing group of hospitalized children included for comparison.

Limited studies have shown that children in EDs and psychiatric units experience greater rates of restraint for violent or self-injurious behavior. Consistent with the results of this study, children with ASD are more likely to experience physical and pharmacologic restraint in the ED.^12–14  Children with ASD hospitalized on psychiatric units are also more likely to experience physical and pharmacologic restraint^15,16  for aggressive behaviors than children without ASD.¹⁶ 

These findings, along with the results of our study, suggest that children with ASD are at risk for physical and pharmacologic restraint at multiple possible points in a hospital experience, including the ED,^12–14  inpatient psychiatric care,^15,16  and inpatient pediatric care. Given the increasing health care visits because of serious mental health difficulties in children that predated and have been accentuated by the coronavirus disease 2019 pandemic onset^18,19  along with the overrepresentation of children with ASD in this group,¹⁴  there is a pressing need to improve the health care setting for people with ASD.

Because people with ASD often have invisible disabilities or disabilities that are less frequently understood, they may get less support in health care settings. There are ongoing efforts to improve access to health care by providing accommodations, environmental modifications, and staff training to support children with ASD and prevent restraint and foregone care. Hospital staff may have an ableist mindset that overvalues the typical and undervalues the neurodivergent. Hospitals often engage in medical or structural ableism that allows large systems to maintain discriminatory processes, policies, and institutions that favor able-bodied people.²⁷  Specific interventions that have been helpful include implementation of a care pathway for patients with ASD in psychiatric settings that resulted in a sustained decrease in use of pharmacologic and physical restraints.^28,29  Specialized psychiatric units compared with general child psychiatric units have demonstrated lower parental ratings of challenging behaviors and ED visits 2 months after discharge.³⁰  In pediatric inpatient settings, implementation of behavioral supports by an interdisciplinary team resulted in infrequent use of physical and pharmacologic restraint.^22,31  Training patient-facing nursing staff resulted in higher rates of adequate strategies to care for children with ASD,³²  suggesting that training health care providers may improve care and reduce adverse events like restraint. An interprofessional consensus statement including representation of autistic adults, and parents of children and adults with neurodevelopmental disabilities is being drafted. The Supporting Access for Everyone Initiative is designed to establish a standard of practice for supporting youth with neurodevelopmental disabilities in health care settings.³³ 

This study identified additional independent associations with odds of restraint. Hospitalized children admitted for behavioral health reasons, and with behavioral, mental health, and developmental disorders other than ASD were more likely to experience physical or pharmacologic restraint during hospitalization, consistent with other studies.³⁴  A study conducted in the ED demonstrated that having a diagnosis of attention-deficit/hyperactivity disorder was a predictor of restraint.¹³  Children with developmental disabilities and children with externalizing and internalizing disorders were more likely to experience restraint in inpatient psychiatric settings.^16,24,35  Having multiple cooccurring diagnoses further increased likelihood of restraint in the psychiatric setting.²⁴  In this study, 22% of children with ASD had cooccurring mental health conditions, slightly lower than other population estimates.^6–8  Further attention to the intersectionality of cooccurring ASD and mental health conditions in future studies is warranted, such as conducting a secondary analysis of this current study to compare children with ASD to children without ASD who were admitted for primary mental and behavioral health conditions.

Several limitations are considered in interpreting the findings from this study. The unmatched cohort of children with ASD identified in this study is predominantly male (75%), white (63.5%), and English speaking (90.4%). The unmatched study sample is consistent with known sex differences in ASD given the ∼3:1 male predominance.³⁶  Race and ethnicity were collapsed into 1 variable according to hospital demographic collection procedures. The racial and ethnic makeup of the study sample demonstrates a significant difference in ASD and non-ASD groups before matching. It is unlikely that true prevalence of ASD varies by race or ethnicity,^37,38  though it has traditionally been shown that white children are more likely to be identified with ASD across time and geographic regions.^39–42  Further attention to race and ethnicity in relation to ASD and restraint should be addressed in future studies. Additionally, given high incidence of children with concurrent public and private insurance coverage, insurance coverage was not included.

Given the retrospective approach, covariates were determined via health record extraction and based on billing diagnoses associated with hospital encounters. Therefore, misclassification and missing diagnoses are possible. Clinicians may not enter a physical restraint order in every instance during which physical restraint was used. Any medication order not accompanied by a restraint order, even if intended as pharmacologic restraint, is not captured. Any nonrecorded restraint events, however, would likely result in an underestimation of the true incidence of restraint. Recent studies, using a broader definition of pharmacologic restraint than this study, demonstrated higher rates of restraint, estimating 3.5% of pediatric mental health visits to the ED⁴³  and 12.6% of encounters for a primary mental health reason in inpatient pediatric units.³⁴ 

Restraint was a binary variable, which cannot account for differing experiences of restrained children. Physical and pharmacologic restraint are not differentiated, nor are we able to describe length of time restrained. Lastly, this is a single-institution study, limiting generalizability, and would benefit from replication and multicenter studies.

Children with ASD have significantly higher odds of experiencing involuntary physical or pharmacologic restraint for violent or self-injurious behavior compared with children without ASD during inpatient pediatric hospitalization after accounting for factors related to demographics, hospital stay, and cooccurring diagnoses. Work is needed to modify the hospital environment for children with ASD to reduce behavioral dysregulation that can lead to unnecessary and avoidable physical and pharmacologic restraint.

The study team thanks William Barbaresi, MD, Olivia Miller, BCBA, LABA, David Davis, MN, RN, and Lauren Herr, MBA, MPH, for their consultation on this study.

ASD

autism spectrum disorder

ED

emergency department

OR

odds ratio

PSM

propensity score matching