Self-Injurious Behavior Among Adults With ASD: Hospitalizations, Length of Stay, and Costs of Resources to Deliver Care

Individuals diagnosed as having autism spectrum disorder (ASD) are at increased risk of experiencing psychological comorbidities (1–3), self-injurious behavior and ideation (4–7), and both suicide ideation and attempt (8–10). Although the etiology of self-injurious behavior and ideation in the ASD population is not entirely understood, it is theorized to arise from a combination of atypical sensory processing (11, 12), frustration with communication (13), and co-occurring psychological conditions (5, 14), with social determinants (e.g., bullying and social exclusion) perhaps mediating some of the association between ASD and psychiatric comorbidities (15). It is estimated that 28% of U.S. children with ASD experience self-injurious behavior (6). The percentage of U.S. children without ASD who exhibit self-injurious behavior has been estimated at around 8% (16). However, definitions of self-injurious behavior and ideation vary widely, limiting the ability to compare samples with and without ASD. Epidemiological research on self-injurious behavior and ideation among U.S. adults with ASD, rather than children, is limited, and little is known about the burden to the health care system.

In a 2008 study of predictors of hospitalization among children with ASD, self-injurious behavior and ideation was associated with an increase of 2.14 in the odds of hospitalization, and risk of hospitalization increased with age; however, the study used a relatively small convenience sample of caregivers, and age was truncated at 21 (17). Kalb and colleagues (18) analyzed data from the Healthcare Cost and Utilization Project (HCUP) National Emergency Department Sample (NEDS) to describe national rates of emergency department (ED) visits among children with and without ASD. This study demonstrated that the odds of psychiatry-related ED visits (e.g., suicide or self-harm or mood disorder) were ninefold higher among children with ASD compared with children without ASD. Vohra and colleagues (19) used HCUP’s NEDS data but restricted their analysis to adults ages 22–64. Their study also demonstrated that psychiatric conditions were the primary reason for ED visits for adults with ASD and that adults with ASD had a substantially higher rate of psychiatry-related ED visits compared with adults without ASD (15% versus 4%). Moreover, the average cost for ED visits was higher for adults with ASD compared with adults without ASD.

In a separate study, using 2008 HCUP NEDS data, Kalb and colleagues (20) found that youths with ASD were about five times more likely than youths without ASD to have an ED visit related to self-injurious behavior and ideation and to be hospitalized. In one of the few studies to focus on hospital utilization related to self-injurious behavior and ideation among adults with ASD, Kato and colleagues (21) utilized a small sample of 587 patients to compare characteristics of suicide attempts among ED visits of adults with and without ASD. Individuals with ASD were overrepresented among those who visited the ED for reasons associated with a suicide attempt and experienced longer hospital stays and higher costs. This study, however, was concentrated in only one hospital in Japan and was restricted to suicide attempts. Several other studies have corroborated the link between ASD and psychiatry-related ED visits and hospitalizations among children and adolescents, but they have not specifically focused on self-injurious behavior and ideation or on adults (22). To our knowledge, no nationally representative study has examined inpatient hospital utilization and costs related to self-injurious behavior and ideation among adults with ASD in the United States.

In this study, we used HCUP’s 2014 National Inpatient Sample (23) (HCUP-NIS) to compare rates of hospitalizations related to self-injurious behavior and ideation between adults with and without ASD, and we also identified differences in length of stay and costs. Given the potentially greater prevalence of self-injurious behavior and ideation among individuals with ASD compared with those without ASD, we hypothesized that compared with adults without ASD, adults with ASD would be more likely to have a hospitalization related to self-injurious behavior and ideation, would have longer stays for such hospitalizations, and would have higher costs. Because previous research has focused primarily on the ED setting and on children with ASD, this study filled a gap in the literature by evaluating inpatient hospitalization (not ED visits) among adults with self-injurious behavior and ideation by using a nationally representative sample.

Methods

Data Source

Data for this study were derived from the HCUP-NIS, the largest all-payer, publicly available U.S. inpatient health care database. It contains data on approximately eight million hospital stays each year from about 1,000 hospitals. These data yield approximately a 20% stratified sample of U.S. community hospitals. The sample of hospitals was drawn from 44 states plus the District of Columbia and was divided into 60 strata based on geographic region, ownership, location, teaching status, and size. Detailed information on the design of the survey is available elsewhere (23).

The HCUP-NIS contains more than 100 clinical and nonclinical data elements for each hospital stay, including the primary diagnosis, up to 24 secondary diagnoses, and up to 14 procedures coded with ICD-9-CM codes. Records also include admission and discharge status, patients’ sociodemographic characteristics (e.g., gender, age, race-ethnicity, health insurance, zip code, and median household income), clinical characteristics (based on the Elixhauser Comorbidity Index of the Agency for Healthcare Research and Quality [AHRQ]) (24, 25), hospital characteristics (e.g., U.S. region of hospital), and charges. The HCUP-NIS does not include unique patient identifiers; thus the unit of analysis is hospital discharge.

Sample

We used a validated ASD case identification algorithm to identify adults with ASD (26). From the discharge data, records with a primary or secondary diagnosis of ASD with ICD-9-CM codes 299.0–299.9 were extracted (these codes include autistic disorder; pervasive developmental disorders; other pervasive developmental disorders, such as Asperger’s; and unspecified pervasive developmental disorders). This extraction procedure yielded the sample of case records of persons with ASD. This method of ASD case ascertainment has been widely used (27). The control group of adults without ASD was matched by age and gender in a 1:3 ratio.

Measures

Dependent variables.

The main dependent variables were hospitalizations related to self-injurious behavior and ideation, length of stay for such hospitalizations, and costs. Hospitalizations were identified as related to self-injurious behavior and ideation if they had an ICD-9-CM code of E950.0–E959 (suicide and attempted suicide and self-inflicted injuries specified as intentional) and V62.84 (suicidal ideation). The length-of-stay variable was derived for each individual hospitalization associated with self-injurious behavior and ideation. The cost variable was derived from HCUP’s cost-to-charge ratio file and represents an estimate for resources utilized without physician fees. Cost was converted to log cost for use in the regression models.

Independent variable.

The main independent variable was an indicator of whether the discharge record did or did not list a primary or secondary diagnosis of ASD.

Covariates.

Model covariates included sociodemographic and clinical characteristics. Sociodemographic characteristics included age (18–24, 25–35, 36–49, and ≥50), type of health insurance (Medicare, Medicaid, private, and uninsured), median household income for patients’ zip code ($1–$38,999, $39,000–$47,999, $48,000–$62,999, and ≥$63,000), and region (Northeast, Midwest, South, and West). Clinical characteristics included having one or more of the comorbidities (identified by the AHRQ with standard methods of Elixhauser and colleagues [24) and number of procedures.

Statistical Analyses

Unadjusted sociodemographic characteristics and hospitalizations related to self-injurious behavior and ideation were compared by ASD status. Additional comparisons of clinical and service characteristics by ASD status were conducted among those with a hospitalization related to self-injurious behavior and ideation, including comorbidity, number of procedures, length of stay, and cost. A chi-square test was used to assess differences in frequencies for categorical variables. Means, standard deviations, and medians are reported for continuous variables; significance was tested with the Wilcoxon-Mann-Whitney test because of nonnormal distributions (28).

We used hierarchical mixed-effects models (29) to account for the clustering nature of the HCUP-NIS data (30). In this study, patients (level 1) were clustered within hospitals (level 2). Hierarchical mixed-effects logistic regression was used for the binary dependent variable (having a code for a hospitalization related to self-injurious behavior and ideation), and hierarchical mixed-effects linear regression was used for the continuous dependent variables (length of stay and cost). In addition, we evaluated the presence of any moderation by including interaction terms between ASD and the covariates in each model. Sampling weights were not used on the basis of AHRQ’s recommendation (30). All analyses were performed with Stata 14 MP (31).

This study was approved by the Brandeis University Institutional Review Board.

Results

In Table 1 we present bivariate, unadjusted data on sociodemographic and clinical characteristics of adults with and without a diagnosis of ASD who were age and gender matched in a 1:3 ratio. We identified 5,341 hospital admissions of adults with ASD and 16,023 admissions of adults without ASD. Compared with adults without ASD, adults with ASD were more likely to be non-Hispanic white and to have public health insurance (Medicare or Medicaid) and were less likely to be uninsured and living in a low-income area. Adults with ASD were more likely than adults without ASD to have a code indicating that the hospitalization was related to self-injurious behavior and ideation (12.7% versus 6.1%), to have longer lengths of stay (7.61 versus 5.43 days), and to have 36.8% higher costs ($6,788.37 versus $4,961.06).

In Table 2 we present the hierarchical mixed-effects logistic regression results for the association between ASD status and having a hospitalization related to self-injurious behavior and ideation. We first present a main-effects model and then a full model with interactions. After controlling for model covariates, we found that adults with ASD had nearly twice the odds (odds ratio [OR]=1.99) of having a hospitalization related to self-injurious behavior and ideation, compared with adults without ASD. There were also significant interactions of ASD with gender, age, and insurance. Compared with men, women were less likely (OR=0.31) to have a hospitalization related to self-injurious behavior and ideation if they did not have ASD. However, among adults with ASD, hospitalizations related to self-injurious behavior and ideation did not differ significantly by gender. Among adults with ASD, those ages 25–35 were no less likely than those ages 18–24 to have a hospitalization related to self-injurious behavior and ideation; however, among adults without ASD, those ages 25 to 35 were significantly less likely (OR=0.69) than those ages 18 to 24 to have a hospitalization related to self-injurious behavior and ideation. Although the rate of hospitalization for self-injurious behavior and ideation was lower among adults ages ≥50 with ASD (odds=.022, 95% CI=.015–.032) relative to adults ages ≥50 without ASD (odds=.006, 95% CI=.004–.008). No interaction was found between ASD and the 36–49 age group. Finally, among adults with ASD, having Medicare, having Medicaid, and being uninsured, compared with private insurance, were not predictors of having a hospitalization related to self-injurious behavior and ideation. However, among adults without ASD, having Medicare coverage, compared with private insurance, was associated with a greater likelihood of having a hospitalization related to self-injurious behavior and ideation (OR=1.89), as was having Medicaid coverage (OR=1.62).

In Table 3 we present the hierarchical mixed-effects linear regression results for the association between ASD status and length of hospital stay among those with hospitalizations related to self-injurious behavior and ideation. After controlling for model covariates and clustering, we found that among adults whose hospitalization was related to self-injurious behavior and ideation, those with ASD spent 2.14 additional days in the hospital compared with adults without ASD. No significant interactions were found between ASD status and covariates in predicting length of stay among adults with hospitalizations related to self-injurious behavior and ideation.

In Table 4 we present hierarchical mixed-effects linear regression results for the association between ASD status and total log costs for hospitalizations related to self-injurious behavior and ideation. After controlling for sociodemographic factors only, we found that among individuals whose hospitalization was related to self-injurious behavior and ideation, those with ASD had higher costs than those without ASD (16.81% higher, 95% confidence interval=8.05%−26.29%). When the analysis controlled for comorbidity, number of procedures, and length of stay among those whose hospitalization was related to self-injurious behavior and ideation, adults with ASD had 7.48% (95% CI=1.05%–14.32%) higher residual costs compared with those without ASD. No significant interactions were found between ASD and covariates in predicting costs among those with hospitalizations related to self-injurious behavior and ideation.

Discussion

The results support our hypotheses, suggesting that adults with ASD are nearly twice as likely as adults without ASD to have inpatient admissions related to self-injurious behavior and ideation. Among adults whose hospitalization was related to self-injurious behavior and ideation, those with ASD had longer stays and, even after the analysis accounted for length of stay, they also had higher costs. These findings build on past research demonstrating that adults and children with ASD are more likely than those without ASD to frequent the ED for psychiatric treatment (18, 19), that self-injurious behavior and ideation is a predictor of hospitalization among children with ASD (17), and that self-injurious behavior and ideation is more prevalent among both adults and children with ASD (8–10). These findings specifically illustrate the burden presented to inpatient psychiatric units by adults with ASD and self-injurious behavior and ideation.

The relationship between gender, age, and insurance and hospitalizations related to self-injurious behavior and ideation among adults with and without ASD suggests that approaches to preventing and understanding high-risk self-injurious behavior and ideation might differ for adults depending on ASD diagnosis. We found no gender differences in hospital admissions for self-injurious behavior and ideation among those with ASD; however, among those without ASD, men had higher rates of hospitalization for self-injurious behavior and ideation compared with women. In the general population, epidemiological research on nonsuicidal self-injurious behavior and ideation has provided conflicting evidence on gender differences, with some research finding no differences by gender (32) and others finding a higher prevalence among females (33). Future research should investigate gender differences in self-injurious behavior and ideation, especially with respect to typology and severity.

Age has a moderating effect on self-injurious behavior and ideation in the community. For the general population, young adults (18–24) are at substantially increased risk of self-injurious behavior and ideation compared with older adults (32). We found that among adults without ASD, older age decreased the risk of hospitalization for self-injurious behavior and ideation; however, among adults with ASD, those ages 25–35 were no less likely than those ages 18–24 to have a hospitalization related to self-injurious behavior and ideation. Further, adults ages ≥50 with ASD were more likely to have a hospitalization related to self-injurious behavior and ideation relative to adults ages ≥50 without ASD. This suggests that self-injurious behavior and ideation, or at least those cases requiring hospitalization, remains a constant phenomenon well into adulthood for those with ASD. We noted differences between those with and without ASD in how insurance type was related to hospitalizations for self-injurious behavior and ideation; however, it is unclear what these findings mean. Research examining the role of insurance coverage on health care access and utilization among children with ASD found that those with private insurance had a lower likelihood than those with public insurance of receiving therapy and higher out-of-pocket expenses (34). It is unclear how this phenomenon might affect utilization of crisis services among adults with ASD in particular.

Among adults with a hospitalization related to self-injurious behavior and ideation, those with ASD had longer stays than those without ASD, and after the analysis controlled for clinical factors and length of stay, they also had higher residual costs. Differences in costs not associated with comorbidity, number of procedures, or length of stay could be attributed to greater use of private rooms and to use of psychotropics. Previous research has suggested that polypharmacy in inpatient psychiatry contributes to longer stays and adverse events (35). Indeed, antipsychotic use and polypharmacy are highly prevalent among adults with autism (and developmental disabilities more broadly) in outpatient settings, especially among those with a history of psychiatric crisis and service utilization (36, 37). It is reasonable to suspect even greater use of polypharmacy in the inpatient context for persons with ASD (38, 39).

The evidence found in this study—for greater prevalence of hospitalizations related to self-injurious behavior and ideation, for longer stays, and for higher costs among adults with ASD compared with those without ASD—suggests the need to ensure that inpatient hospital environments are able to provide appropriate care. The environment of hospitals presents unique sensory, interpersonal, and organizational challenges that could exacerbate distress among adults with ASD (e.g., disruption in routine, harsh lighting, and communal settings). Providers’ capacity to accommodate the needs of adults with ASD could be limited (40). Therefore, efforts should be made to support providers’ ability to safely treat and deescalate distress among adults with ASD. Moreover, future research should focus on identifying interventions to prevent distress and hospitalization among those with ASD (41), with particular attention given not only to those of transition age (42) but also to young adults more broadly.

The study limitations warrant consideration. First, it is possible that some adults with ASD admitted to the hospital were not coded as having ASD, because the main reasons for the hospitalization and the related diagnosis were the focus and not the person’s ASD. In addition, there may be a bias in accuracy of coding toward those who present to the ED with symptoms of behavioral health conditions, because such patients might be more likely to be screened for behavioral health history and thus have their ASD identified and coded. This nonrandom variability in reliability of coding could have biased comparisons between adults with and without ASD in predicting self-injurious behavior and ideation, but it is likely to be of less concern in the comparisons between the ASD and non-ASD groups whose hospitalization was related to self-injurious behavior and ideation (i.e., length of stay and costs). Also, we used administrative data rather than data from chart reviews, which also limited our ability to observe all cases of both self-injurious behavior and ASD. Nevertheless, the process for extracting ASD codes has been validated and widely used (27).

Second, common with many secondary data analyses, some data were missing; however, multiple imputation was employed for variables with missing data, consistent with best practices (43, 44). Third, causality cannot be established because of the cross-sectional nature of the data. Fourth, we were working with a hospital-based denominator; future epidemiological research should more rigorously assess risk of hospitalization related to self-injurious behavior and ideation by using community-based denominators. Finally, we were unable to distinguish type and severity of self-injurious behavior and ideation or differences in presentation of ASD (e.g., verbal or intellectual disability), which limited the depth of our analyses.

Conclusions

This study has added to the literature by using a nationally representative sample to evaluate hospitalizations related to self-injurious behavior and ideation among adults with ASD. Adults with ASD might lack access to appropriate health care services, have difficulty communicating their symptoms to their health care providers, be more sensitive to stimuli and stressors, and lack a healthy social support network (45). Future research is needed to better understand prevention of hospitalization among adults with ASD, as well as organizational elements that might support hospital staff and providers in accommodating adults with ASD who are hospitalized with self-injurious behavior and ideation.