Factors Associated With Psychiatric Readmission of Youths in a Racially Diverse and Urban Hospital Setting

We used data that were extracted from the EHR system of a large, urban pediatric hospital to conduct our secondary analyses. Per the hospital’s institutional review board, this study qualified as research that is exempt from review (see the online supplement to this report for data extraction procedures).

We identified patients (4–17 years old) who were discharged from either the child or adolescent inpatient psychiatric unit between July 1, 2017, and July 31, 2019 (N=1,674). To best represent baseline admissions, we excluded 70 patients who had an admission within 1 year before their earliest admission in our data set (total sample, N=1,604).

From the extracted data, we calculated how many readmissions each patient had within 1 year of the patient’s earliest discharge date. The number of readmissions was recoded into binary (“any” vs. “none”) as well as categorical (“none” vs. “once” vs. “more than once”) outcome variables to explore possible distinctions between patients who were readmitted once and those who were readmitted multiple times within a year of discharge.

We computed descriptive statistics for all variables of interest to check for normality of distributions, outliers, or both. We used chi-square analyses to test bivariate associations of demographic and clinical factors with any readmission. The factors significantly associated with any readmission at the bivariate level were simultaneously entered into a logistic regression model to estimate their independent effects. To explore predictors of multiple readmissions, the same significant factors were entered into a random-effects multinomial logistic regression to estimate odds ratios for being readmitted once versus never during the study period, readmitted more than once versus never during the study period, and readmitted once versus readmitted more than once. Analyses were conducted in SPSS, version 27, with statistical significance set at p<0.05.

Most patients were adolescents (ages 13–17) (67%, N=1,078), Black (59%, N=948), English speaking (88%, N=1,407), and insured by Medicaid (61%, N=972). Most patients’ first admissions lasted 7 days or fewer (70%, N=1,117). About half did not have documented suicidality (50%, N=806), and about two-thirds had documented mood disorder diagnoses (65%, N=1,035). Most patients were never readmitted within a year of discharge (85%, N=1,355), although 11% (N=175) were readmitted once and 5% (N=74) were readmitted more than once (see the online supplement).

Chi-square tests indicated that race was significantly associated with any readmission. Higher percentages of both Black (18%, N=174 of 948) and Hispanic or Latino (14%, N=44 of 321) patients were readmitted at least once compared with White patients (9%, N=27 of 293) or patients of other or unknown race (10%, N=4 of 42). Insurance type was also significantly associated with any readmission, with a higher percentage of patients insured by Medicaid (18%, N=178 of 972) being readmitted compared with those with commercial or other insurance (12%, N=71 of 599) or who were self-insured (0%). Finally, a higher percentage of patients with hospital stays of more than 7 days (20%, N=96 of 487) were readmitted than were those with shorter stays (14%, N=153 of 1,117). When these variables were simultaneously entered into a multivariate logistic regression, Black race (adjusted odds ratio [AOR]=1.90, p=0.007), Medicaid insurance (AOR=1.40, p=0.04), and LOS more than 7 days (AOR=1.58, p=0.002) remained significant independent predictors of readmission (see the online supplement).

In the multinomial logistic regression model, including the same three predictors that were significant at the bivariate level, Black youths’ odds of being readmitted more than once (compared with not being readmitted) were 2.5 times those of White youths (p<0.05). Similarly, among youths insured by Medicaid, the odds of being readmitted more than once (compared with not being readmitted) were almost twice the odds of those with commercial or other insurance (p=0.04). Longer LOS was associated with significantly greater odds of one readmission (AOR=1.43, p=0.04) as well as multiple readmissions (AOR=1.92, p=0.01) relative to patients who were hospitalized for 7 or fewer days. There were no significant predictors of being readmitted more than once compared with only once (see the online supplement).

The current study examined the likelihood of psychiatric readmission within 1 year of discharge among a sample of racially diverse, mostly Black youths at an urban pediatric hospital. The relationship between multiple readmissions and race was particularly striking, with Black youths having 2.5 times greater odds of multiple readmissions within a year of discharge compared with the reference group of White youths. Although this association is consistent with some findings in the literature, its strength exceeds what has been demonstrated previously (7). The association between public insurance and increased likelihood of readmission is also largely consistent with prior studies; past meta-analyses have shown that patients insured by Medicaid have a greater likelihood of readmission compared with those with private insurance (4, 6). Although LOS is often conceptualized as a marker of illness severity, studies have found that it may also indicate care quality and accessibility as well as quality of discharge planning, all of which may be affected by insurance (6). Other demographic factors we examined, such as age, language, suicidality, and diagnostic category, were not significantly predictive of readmission likelihood.

The strong association between race and readmission likelihood is notable in the context of this study’s predominantly Black sample. The observed differences in readmission by race and insurance type may be framed in the context of historic disparities, systemic racism, and inequities in health care. Although other indicators of health care inequities are improving (e.g., healthy living, patient safety), disparities in access to and quality of health care related to socioeconomic status, race and ethnicity, and geography have been highly resistant to change (8). Since the late 1990s, studies have shown that Black and Hispanic youths use mental health services at lower rates than do White counterparts, even after controlling for relevant factors (9, 10). The Centers for Disease Control and Prevention has noted that barriers to access to adequate pediatric mental health care include an insufficient number of pediatric mental health care providers, lack of or limited transportation, long waitlists, and concerns about treatment cost (11). Prior research has also attributed racial disparities in pediatric mental health care to systemic causes, such as insufficient preventive efforts for populations at risk of poor mental health outcomes; high mental health care needs among youths with adverse childhood experiences; and factors related to treatment itself, including service use, response bias, mental health stigma, cultural differences in help-seeking behavior, and potential for undertreatment (12). Because establishing follow-up care is often a primary goal following hospitalization, it is particularly troubling that prior research has indicated that Black children have a lower likelihood, compared with all other races, of participating in outpatient follow-up, possibly because of inadequate infrastructure (7).

Findings indicate the need for prevention- and intervention-level efforts to promote accessible, high-quality mental health care for Black youths and youths insured by Medicaid, especially because the combination of these groups represents about 42% of American children (13). Recent research has identified that school-based mental health care and integrated mental health care models are linked to greater service use and acceptability; policy mechanisms addressing health insurance reform also show promise for increasing affordability of services (14). Although it is beyond our scope to explore these factors, consideration of the complexity of the existing landscape is crucial to understanding group-based differences in mental health care.

Our findings have limited generalizability because data were collected from one hospital that serves a unique population of patients who are predominantly Black and English speaking. However, it is valuable to understand predictors of readmission in this group, which historically has been and remains underserved by the mental health care system. It is important to note that the current study’s method of assessing the primary outcome of readmission within 1 year of discharge may underestimate readmissions. Although there was only one similar facility in the study service area, patients could have been readmitted to a different hospital that was not captured in the existing data set; moreover, the impacts of this underestimation could vary by racial-ethnic group.

This study was limited in its measurement of some predictor variables, given that it was a secondary analysis of EHR data. We used Medicaid status as a proxy indicator for socioeconomic status because direct measures of family income were unavailable, but Medicaid status is not a one-to-one substitute for inferring income; other factors affect Medicaid eligibility, and Medicaid may also lower financial barriers to care. Suicidality may not have been fully captured in the EHR because of changes in clinical coding practices over the course of this study, so this data set likely underestimated the true prevalence of suicidal ideation or attempts. Diagnostic category determinations were made on the basis of a single diagnosis and may not reflect a patient’s full presentation, especially for those with comorbid diagnoses. Although LOS is of independent clinical importance, we used LOS as a proxy indicator for illness severity. However, LOS may also be related to the complexity of discharge planning or to social and medical factors separate from illness severity. Other potentially important clinical predictors not assessed in this study include mental health care use pre- and postdischarge, comorbid medical diagnoses, out-of-home placement, and voluntary versus involuntary admissions.

Additional studies of predictors of psychiatric readmission in racially diverse samples are needed. If predictors of readmission can be reliably identified, then subsequent interventions to reduce their likelihood can be thoughtfully developed and implemented to mitigate established risks for specific populations, especially for groups that have been negatively affected by systemic health care inequities. Interventions that are focused on improving care coordination represent a promising avenue for readmission prevention, because receipt of aftercare services following discharge has been associated with a significant reduction in youths’ psychiatric readmissions (15). The prominence of stable demographic factors, such as race, as predictors of readmission in our findings underscores the need to study and address the root causes of inequities in the health care system. Future studies should further explore the nature of underlying inequities and systemic racism in health care to identify meaningful opportunities for systemic change. Although our study was unable to assess the quality of care postdischarge, this factor represents a promising direction of future research. Future studies should explore postdischarge factors as predictors of readmission risk and the relationship of or potential interaction between preadmission factors and racial-ethnic health disparities. Each presents an important opportunity for future clinical intervention and study because these factors, unlike demographic characteristics, are modifiable and may be responsive to interventions that could ameliorate the disparities in readmission likelihood found in the current study.

This study makes a novel contribution to the literature because it is, to our knowledge, the first investigation of factors associated with psychiatric readmissions in a sample of predominantly Black youths. Findings showed that the likelihood of readmission was increased, compared with respective reference groups, for Black youths, youths insured by Medicaid, and youths with an LOS of more than 7 days. Effects of these predictors were even greater for youths readmitted multiple times compared with those not readmitted. Findings underscore the importance of studying readmission predictors in racially diverse samples to inform intervention development, improve initiatives such as care coordination and aftercare, and address underlying inequities in the health care system.

The authors acknowledge the contributions of Jaeho (Thomas) Chang and Anqing Zhang, Ph.D., to this study. This content is solely the responsibility of the authors and does not represent official views of these organizations.