Pediatric Physical Restraint Coding in US Hospitals: A 2019 Kids Inpatient Database Study

This analysis used the 2019 edition of the Kids’ Inpatient Database (KID), from the Healthcare Cost and Utilization Project of the Agency for Healthcare Research and Quality. The KID is an administrative claims database including information for patients aged 20 or younger receiving inpatient care at community hospitals in the United States. The KID samples 80% of non-newborn discharges from 3998 nonfederal acute care hospitals in 49 states and is produced every 3 years. Data from 2019 is the most recent available. Data are provided for hospitalizations in a deidentified format, with no mechanism for linking database entries to identified individuals. KID data are not separated by hospital unit, and so data for general wards, ICUs, and psychiatric units within acute care hospitals are included. As this is a deidentified, publicly available database, this study was determined to be Not Human Subjects Research by the Institutional Review Board.

Hospitalizations involving physical restraint coding were identified using International Statistical Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) code Z78.1 “physical restraint status.” This code includes all episodes of physical restraint and does not distinguish the duration of restraint, type of restraint, or primary indication for restraint. Discharges were included in this analysis for patients from ages 1 to 18 years old if Z78.1 was included among the discharge diagnosis codes.

Sociodemographic characteristics are presented descriptively based on data from the KID. The KID is derived from state-level data, which is then recoded into uniform categories standardized during the production of the KID. As a result, there is variability in the underlying sociodemographic data as different hospitals and states report data differently, and it is not possible to know how the racial category was assigned for any given hospitalization in the KID. Race and ethnicity are social constructs subjective to those reporting. The KID combines race and ethnic categories, with values of white, Black, Hispanic, Asian or Pacific Islander, Native American, other, and missing. Given small sample sizes in certain KID race categories, to maintain individual patient privacy as required by the KID data use agreement, the categories of “Asian or Pacific Islander,” “Native American,” and “other” were combined into a single category of “other.”

Primary hospital diagnoses were identified from the discharge diagnoses and classified according to the Pediatric Clinical Classification System (PECCS),¹⁹  which groups the 72 446 ICD-10-CM diagnosis codes into 834 clinically distinctive categories for pediatric medical conditions.

The KID is based on a survey design in which hospitals are stratified based on geographic region, hospital ownership, teaching status, and bed size and then discharges are sampled within strata. As a result, all reported variables come with an associated variance. This variance is used to present sample uncertainty for the total number of discharges with a physical restraint code, whereas weighted point estimates are reported for all other analyses. All analyses were conducted on data weighted according to the appropriate KID discharge weight to obtain national estimates.

For the primary statistical analysis, a logistic regression incorporating the KID survey design was performed on the binary outcome of a hospitalization, including discharge a billing code for physical restraint status (compared with hospitalizations not including a physical restraint code). In this model, age (z score), sex (male versus female), race or ethnicity (Black, Hispanic, or other versus white), primary mental health or substance abuse diagnosis (yes versus no), region (Northeast, Midwest, or South versus West), income quartile of the patient’s ZIP code (first, second, or third quartile versus highest quartile), primary payor (Medicare or Medicaid; or other versus commercial insurance), and patient severity (defined based off of All Patient Refined-Diagnosis Related Groups′ Severity of Illness)^20,21  were descriptor variables. A term for race was included in this regression given prior literature indicating disparities in race utilization based on race. All analyses were conducted using SPSS (version 29; IBM Software, Inc, Armonk, NY). This study is reported in accordance with the Reporting of Studies Conducted Using Observational Routinely-collected Health Data statement.²² 

In the 2019 KID, 8893 hospitalizations (95% confidence interval [CI]: 8227–9560) for patients 1 to 18 years old involved a discharge code for physical restraint status and 1 409 614 (95% CI: 1 367 522–1 469 494) did not, for an overall rate of 0.63% of pediatric hospitalizations involving a code for physical restraints. Demographically, 57.2% of hospitalizations with a restraint code were for male patients and 42.8% for female patients, whereas 47.7% of overall hospitalizations were for male patients. Full demographic information for hospitalizations with and without restraint coding is given in Table 1. Physical restraint coding varied with age, from a low of 0.27% of hospitalizations for 4-year-old patients to a maximum of 0.91% of hospitalizations for 15-year-old patients (Fig 1).

Hospitalizations with a coded restraint diagnosis included a wide variation of primary discharge diagnoses, which differed among pediatric patients of different ages. Primary discharge diagnoses, based on PECCS codes, are listed in Table 2. Overall, mood disorders (a category that includes the diagnoses for mood disorders because of known physiologic conditions, other or unspecified mood disorders, and disruptive mood dysregulation disorder) was the most common PECCS category (790 hospitalizations; 9.1%), followed by major depressive disorder (710 hospitalizations; 8.2%), suicide and intentional self-inflicted injury (687 hospitalizations; 7.9%), and respiratory failure insufficiency or arrest (384 hospitalizations; 4.4%). Among patients ages 1 to 5, the most common diagnostic categories for hospitalizations with a restraint code are all primarily medical (with respiratory failure, seizures, sepsis, and bronchiolitis as the top 4 categories), whereas for all other age ranges, the top diagnostic categories are primarily psychiatric in nature, with mood, conduct, and psychotic disorders being among the most frequent. Raw ICD-10-CM codes for primary diagnoses for hospitalizations with a physical restraint code are listed in Supplemental Tables 5–9.

Among primary hospital diagnoses, there is significant variability in the fraction of hospitalizations involving co-occurring physical restraint coding. The primary discharge diagnoses categories that most commonly had a concurrent diagnosis of physical restraint status are listed in Table 3. Among the 1148 hospitalizations involving alcohol-related disorders, 87 (7.6%) also involved a code for physical restraint status, which is the highest rate among all diagnostic categories. Other primary diagnoses most frequently involving physical restraint coding include autistic disorder (6.8%), conduct disorder (6.3%), and intermittent explosive disorder (5.1%). In total, 17 diagnostic categories had a co-occurring physical restraint code for more than 3% of hospitalizations, whereas 117 categories had a restraint utilization rate of ≤0.3% and 600 PECCS categories had no hospitalizations with co-occurring physical restraint coding.

To explore factors associated with the outcome of having a physical restraint discharge code, we performed a logistic regression on the binary outcome of a hospitalization including a code for physical restraint status (yes or no). In this model, increasing age (adjusted odds ratio [aOR] 1.30; 95% confidence interval: 1.22–1.40), male sex (aOR 1.56; 95% CI: 1.47–1.65), Black race (aOR 1.43; 95% CI: 1.33–1.55), a primary mental health or substance diagnosis (aOR 7.13; 95% CI: 6.42–7.90), Medicare or Medicaid insurance (aOR 1.33; 95% CI: 1.24–1.43), and more severe illness (aOR 2.83; 95% CI: 2.73–2.94) were associated with higher odds of a hospitalization involving a physical restraint code. Hispanic ethnicity (aOR 0.87; 95% CI: 0.79–0.97), and Midwest (aOR 0.78; 95% CI: 0.66–0.92) or South (aOR 0.76; 95% CI: 0.63–0.91) regions were associated with lower odds of physical restraint coding, whereas household income of the patient’s ZIP code was not associated with physical restraint coding (Table 4).

Among hospitalizations for patients 1 to 18 years old in the United States in 2019, 8893 (95% CI: 8227–9560) involved a discharge code for physical restraint status, representing 0.63% of overall hospitalizations for patients in this age range. Physical restraints were coded for a variety of primary diagnoses, with psychiatric illnesses including mood, psychotic, and conduct disorders among the most common diagnoses overall for hospitalizations involving a physical restraint code (Table 2). There was an age dependence in the frequency of diagnostic categories associated with restraint, with older patients more likely to have a mental health discharge diagnosis and younger patients more likely to have a medical diagnosis.

The frequency of physical restraint coding found in this study can be contextualized by findings in prior studies focusing on specific areas of care delivery. For instance, in the emergency department (ED) setting, a total of 0.1% of encounters for patients aged 0 to 16 in a New England health care system involved physical restraint, with patients of Black race (relative to white race) and males (relative to females) displaying a higher odds of physical restraint utilization.¹⁵  A further analysis of data from this system found an intersectional interaction between race and sex, with larger racial disparities in restraint rates in girls relative to boys.¹⁶  A specialized pediatric psychiatric ED in the United States reported an average physical restraint prevalence of 1.94% over a 6-year period, with restraints more common among male patients.²³  A further Scandinavian survey showed that 79% of emergency departments use temporary physical restraints for painful pediatric procedures.²⁴  In the psychiatric setting specifically, systematic literature review of psychiatric pediatric inpatients suggests a weighted mean physical restraint prevalence across multiple studies of 29.4%.²⁵  Among critically ill children in the PICU. Sixty eight percent of such units in the United Kingdom employ physical restraint of some type.²⁶  Prevalence of physical restraint use in Japanese PICUs has been estimated at 53% of patients, most often in the form of wrist restraint bands.²⁷  The estimate of the prevalence of physical restraint coding among community hospital inpatient pediatric hospitalizations in the KID is similar to the rate of 0.59% found in a 5-year study of inpatient hospitalizations of patients aged 5 to 20 in a single health system comprised of academic children’s hospitals.¹⁷  This difference in estimates is likely attributable to a strong association between the need for physical restraints and either acute agitation secondary to psychiatric illness (in the case of a specialized psychiatric units) or to prevent interference with life-sustaining interventions such as mechanical ventilation (in the case of the ICU).

Multiple demographic factors were associated with differential restraint coding in an adjusted model. Older patients were more likely to have a physical restraint diagnostic code, with 0.81% to 0.91% of hospitalizations in teenagers having a restraint code compared with 0.27% to 0.38% among children aged 1 to 6. The rate of physical restraint coding in teenagers was higher than any adult age range based on a recent administrative claims study.²⁸  The primary diagnostic categories associated with age also differed in this study. Psychiatric disorders including mood disorders, suicidality, conduct disorders, and psychotic disorders were most common in adolescents, whereas among patients aged 1 to 5 respiratory failure, seizures, sepsis, and bronchiolitis were the most common diagnostic categories. This may suggest that restraint use in younger children, although less common than in older children, is associated with critical illness, whereas restraint use in older children is associated with psychiatric illness. This idea is supported by ICU-specific physical restraint data, in which restraints were most commonly used in patients aged 1 to 3.²⁷  In the psychiatric ED, restraint use occurred at similar rates in adolescents versus younger children.²³  As hospitalizations in this study were in the community hospital setting, practices, such as boarding of psychiatric patients in the community hospital settings while awaiting mental health placement, may contribute to increased psychiatric acuity in units not specifically designed for such care,^29,30  which may place patients at increased risk of restraint.

A significant association was also present between race and physical restraint as a discharge diagnosis, with Black patients having higher odds of physical restraint coding compared with white patients and Hispanic patients having a lower odds relative to white patients. This result was significant, even controlling for other demographic factors and disease severity. Similar disparities in restraint utilization have been identified in the adult emergency department setting.^31,32  Data drawn from pediatric ED visits replicates this relationship, with increased use of restraints in Black patients relative to white patients, but not Hispanic patients relative to white patients.¹⁵  These data suggest that race-related bias may play a role in utilization of restraints in pediatric populations, highlighting the importance of interventions to reduce disparities in high-risk restraint use. Multidisciplinary quality improvement efforts have been successful at reducing the use of restraints within pediatric general hospitals without increase in staff injury,^33,34  and incorporation of such efforts, including standardized admission order sets, increased collaboration with multidisciplinary behavioral health teams, and individualized safety planning, hold promise for sustained reduction in restraint utilization among patients with behavioral health diagnoses.³⁵  It remains unclear, however, whether such efforts mitigate racial and gender-based disparities in restraint utilization, and further research into this is required.

This study is based on, and therefore limited by, observational, routinely collected administrative data. This type of data are nationally representative and reflects the reality of healthcare delivery in the United States through 2019. Strengths of this approach include the minimization of bias compared with studies focusing on single centers, geographic regions, or payment sources. However, the generation of data using real world datasets inherently results in limitations on generalizability and interpretability of conclusions.^36–38  Because the KID data set is based on claims, it reflects codes used for billing rather than the actual events of each hospitalization, and discrepancies may exist between these 2 records. For example, if physical restraint use was not documented in the patient record or included in the final coding for billing postdischarge, the KID database would not accurately reflect that use of restraints. This may be an alternate explanation for the lower prevalence estimate of restraint use in our study compared with prior pediatric psychiatric or ED-focused studies. Most likely, the presence of coding for restraint use in final billing is a specific indicator that restraints were used, but the absence of such coding is not sensitive for lack of restraint use. Further work to validate these results would require examination of this discrepancy. Additionally, restraint duration and frequency are not captured by billing codes. For example, although clinically, the brief use of physical restraint or a physical hold for 1-time venipuncture is markedly different from prolonged mechanical restraint in the setting of agitation, these differences are not captured by billing codes. Further elucidating the differences between ICU, psychiatric, ED, or general inpatient use of restraints in pediatrics will require understanding of these differences, as well as potential differences in restraint practice or coding in different regions. Moreover, each hospitalization represents a distinct entity in the KID record, and therefore a given individual may represent multiple entries for multiple hospitalizations, introducing a source of bias into our estimate, particularly for demographic data. Finally, indicators for race in the KID database are based on hospital-recorded values, and protocols for collecting this vary by hospital and state. Although not specifically studied for the KID, in other administrative databases there were significant differences between self-reported race and administrative documentation,^39,40  which confounds analysis of potential disparities in restraint application.

A coded diagnosis of physical restraint status was present for 8893 (95% CI: 8227–9560) hospitalizations in the 2019 KID among individuals aged 1 to 18 years, or 0.63% of overall hospitalizations for patients in this age range. Restraint coding varied by age, sex, race, region, diagnosis, and disease severity. These results highlight potential disparities in restraint utilization and emphasize the importance of a standardized and equitable approach to physical restraint use, particularly in the context of increasing boarding of patients requiring subspecialty mental health care on general pediatrics units. Validation of restraint coding in claims data are, however, a critical further step in extending this research.