Skip to main content
Full access
Reviews & Overviews
Published Online: 11 December 2019

Psychiatric Readmission of Children and Adolescents: A Systematic Review and Meta-Analysis

Abstract

Objective:

To investigate predictors of psychiatric hospital readmission of children and adolescents, a systematic review and meta-analysis was conducted.

Methods:

Following PRISMA statement guidelines, a systematic literature search of articles published between 1997 and 2018 was conducted in PubMed/MEDLINE, Google Scholar, and PsycINFO for original peer-reviewed articles investigating predictors of psychiatric hospital readmission among youths (<18 years old). Effect sizes were extracted and combined by using random-effects meta-analysis. Covariates were investigated with meta-regression and subgroup analyses.

Results:

Thirty-three studies met inclusion criteria, containing information on 83,361 children and adolescents, of which raw counts of readmitted vs. non-readmitted youths were available for 76,219. Of these youths, 13.2% (N=10,076) were readmitted. The mean±SD study follow-up was 15.9±15.0 months, and time to readmission was 13.1±12.8 months. Readmission was associated with, but not limited to, suicidal ideation at index hospitalization (pooled odds ratio [ORpooled]=2.35, 95% confidence interval [CI]=1.64–3.37), psychotic disorders (ORpooled=1.87, 95% CI=1.53–2.28), prior hospitalization (ORpooled=2.51, 95% CI=1.76–3.57), and discharge to residential treatment (ORpooled=1.84, 95% CI=1.07–3.16). There was evidence of moderate study bias. Prior investigations were methodologically and substantively heterogeneous, particularly for measurement of family-level factors.

Conclusions:

Interventions to reduce child psychiatric readmissions should place priority on youths with indicators of high clinical severity, particularly with a history of suicidality, psychiatric comorbidity, prior hospitalization, and discharge to residential treatment. Standardization of methods to determine prevalence rates of readmissions and their predictors is needed to mitigate potential biases and inform a national strategy to reduce repeated child psychiatric hospital readmissions.

HIGHLIGHTS

This systematic review and series of meta-analyses investigated predictors of psychiatric readmission among children and adolescents.
Indicators of clinical severity at both the patient and the hospital levels were relatively more predictive of hospital readmission, compared with sociodemographic characteristics.
Standardization of methods to determine prevalence rates of child psychiatric readmissions and their predictors is needed to inform a national strategy to reduce repeated child psychiatric hospital readmissions.
Child psychiatric hospitalizations are common and costly. Over the past 15 years, pediatric hospitalizations for mental disorders have increased by over 50%, with total expenditures of $11.6 billion (1). Between 2005 and 2014, pediatric hospitalizations in U.S. children’s hospitals rose five times more among children with a psychiatric diagnosis than among children without a psychiatric diagnosis (2). Recent studies suggest that one in four youths is readmitted to a psychiatric hospital within 1 year of discharge (36), with most readmissions occurring within 3 months (7). Repeat hospitalizations disrupt social support and school performance and result in greater stigmatization for youths and their families (8, 9). A national indicator of risk of poor quality of care is the rate of repeat hospitalizations within 7 and 30 days (10). Nevertheless, questions have been raised as to whether repeated child psychiatric hospitalizations reflect hospital-level factors, ecological factors, or the complexity of the clinical presentation of the child served and the psychosocial factors involved (11, 12).
Sociodemographic characteristics appear to influence the risk of readmission, although findings are mixed (13, 14). For example, studies suggest that younger children (4, 13, 15) as well as older children (14, 16) may be more frequently readmitted, whereas other studies find no relationship with age (17, 18). Similarly, the posited relationships between readmission and gender (13, 16, 19), race-ethnicity (7, 13, 20), guardianship status (6, 16, 19, 21), early life stress and child abuse (6), and socioeconomic status (16, 22) are multidirectional. The role of disease severity also remains unclear, with variable findings regarding primary diagnosis and symptom severity (23), comorbidity (3), self-injury, and suicidality (5, 6, 23, 24).
Characteristics of hospitalization may also predict readmission, such as length of stay (6, 14) (LOS), access to inpatient case management services (23), and postdischarge care, including aftercare services (17), disposition to residential treatment or partial hospitalization (23), and discharge medications (21, 22, 25). Identifying predictors of readmission at this level can inform health policies and quality improvement interventions to mitigate cost and burden to systems and families (26). However, to our knowledge, no quantitative meta-analysis of the literature on child psychiatric hospital readmissions has been conducted.
To address this gap, this systematic review and series of meta-analyses investigated predictors of psychiatric readmission among children and adolescents in all available recent scientific literature. Predictor categorization was adapted from prior narrative reviews (24, 27) and divided into the following clusters: patient and family sociodemographic characteristics; indicators of clinical severity, including safety and diagnosis; and characteristics of index hospitalization. To address substantive and methodological heterogeneity in the literature, clusters were further explored by using meta-regression and subgroup analyses.

Methods

Search Strategy

A systematic search was performed in PubMed/MEDLINE, Google Scholar, and PsycINFO databases from January 1, 1997, through June 1, 2018. The study timeline was restricted to these years in light of substantial changes to systems of practice in hospitals, including the advent of managed care, so as to ensure relevance of our results to current providers. Search terms included (psych* OR mental health OR schizophreni* OR schizoaffective OR psychosis OR psychotic OR bipolar OR manic OR mania OR depress* OR mood OR autism OR attention deficit hyperactivity disorders OR neurodevelopmental) AND (readmi* OR rehosp* OR utilization OR hospital* OR admi*) AND (adolesc* OR child* OR youth). (Search strings are detailed in a table in an online supplement to this article.) The search strategy was augmented via review of the reference lists of each of the included manuscripts, citation tracking, and, when necessary, contact with study authors.

Study Selection

The study adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement guidelines (28) and followed a predefined protocol registered in PROSPERO (29) (CRD42018117268). Original studies published in peer-reviewed English-language journals were reviewed. The following inclusion criteria were applied. First, study participants were restricted to children and adolescents (age <18). Studies including adults (≥18 years old) were included only if the study reported data separately for youths. Second, eligible studies had to include a group of youths admitted to a psychiatric inpatient treatment setting (thus studies involving admission to day treatment programs and group homes were excluded). Studies solely targeting emergency psychiatric evaluation services were excluded if no data on hospitalization were available. Exclusion criteria were as follows: publications not including quantitative measures, such as reviews or book chapters; studies not published in English; studies that included adult participants without reporting separate data for youths; studies solely assessing eating disorders and chemical dependency units (because of substantial differences in LOS, medical acuity, and pathology); and studies that did not assess at least one potential predictor of psychiatric readmission.

Data Extraction

The effect size (ES), or statistic from which an ES could be calculated, was extracted for each predictor of readmission. In addition we extracted the following data when available: study author(s), study year, sample size, mean age of participants, sex or gender, and race-ethnicity of participants, study location, study design, institutional study site, level of care, study-specific exclusion criteria (including medical acuity and presence of an autism spectrum disorder [ASD] or intellectual disability [ID] diagnosis), most prevalent diagnoses of the study population, whether the study specifically targeted suicidal youths (attempt or ideation), percentage of suicidal youths, percentage of youths in foster care, provision and type of aftercare services (follow-up appointments, wraparound services, and case management), mean LOS (days), mean time to readmission (months), and study duration (days).
The literature search, title and abstract screening, decision to include full-text articles, and data extraction were independently reviewed by two authors (J.E. and B.Z.). Disagreements were resolved via consensus ratings. In the absence of consensus, a third author (M.S. or B.L.) was consulted.

Risk of Bias

Assessment of the quality of included studies was conducted with an eight-item form adapted from previously published quality criteria for systematic reviews (30, 31). The items assessed whether the study had a nationally representative sample, used a representative sampling time frame, and used random selection. The items also assessed the quality of data collection, use of a sufficient timeline over which to assess readmissions, use of an appropriate sampling size, and presence of multivariate analyses. The last item was an overall assessment of the risk of bias of the study.
Bias was ascertained independently by two authors (J.E. and B.Z.). Each item was rated as low, moderate, or high risk of bias. When studies were missing information, the item was scored as high risk of bias. Thus the overall measurement tended to overestimate risk of bias. Disagreements between raters were resolved via discussion.

Statistical Analyses

Studies were combined if at least three independent studies reported an ES for a given predictor. Because studies reported both continuous and dichotomous measures, ES estimates were converted to the natural logarithmic transformation of odds ratios (ORs). This was done because conversion to a dichotomous ES is more conservative and less prone to biases than conversion from a dichotomous to a continuous ES (32). Effects were pooled by using the DerSimonian-Laird random-effects method of meta-analysis to account for variation between studies and to allow for generalization beyond the study population (33, 34). Following pooling, the log ORs were converted to ORs for ease of interpretation. Because survival data cannot be readily statistically combined with dichotomous cross-sectional data, survival analyses were tracked and reported separately. Each predictor was assessed for publication bias by using Egger’s test (35, 36). Funnel plots were inspected for asymmetry.
Heterogeneity was measured via the I2 statistic and Cochran’s Q test (33, 35). Potential sources of heterogeneity were explored by using subgroup meta-analyses (categorical covariates) or random-effects meta-regression (continuous covariates). Meta-regression is a well-established methodology in meta-analysis in which an outcome variable (the effect estimate) is predicted according to the values of one or more explanatory variables (characteristics of studies that might influence the size of the effect estimate) (37). Subgroup analyses were performed if at least five studies evaluated the same predictor. The following variables were considered in subgroup analyses: study location (U.S. versus international); study design (retrospective versus prospective); study site (single versus multisite); inclusion versus exclusion of individuals with ASD or ID, mood disorders, or suicidality; and inclusion versus exclusion of previously hospitalized children. The following variables were considered in univariate meta-regression: sample size, study year, mean age, gender (percentage female), race-ethnicity (percentage Caucasian because insufficient data were available for other races and ethnicities), mean LOS, mean days to readmission, follow-up interval, and percentage of sample hospitalized for suicidality.
All analyses were performed with the metan package of Stata MP software, version 15.1 (38, 39). Meta-regressions were performed via the metareg command. An alpha level of 0.05 conferred statistical significance, and p values were two-tailed. The study did not meet criteria for review by an institutional review board.

Results

Characteristics of Included Studies

After removal of duplicates, the titles and abstracts of 5,601 unique references were evaluated for eligibility. A total of 4,733 were excluded, and 412 full-text articles were extracted and screened for eligibility. Of these, 33 original studies met eligibility criteria. (A PRISMA flowchart for study inclusion is provided in the online supplement.)
Of the 33 studies, 11 studies were conducted outside the United States. Ten studies were prospective, and the others were retrospective. Nineteen were multisite studies, and the remainder were single site. Five studies focused predominantly on suicidal youths (admitted either for a suicide attempt or ideation), and four excluded youths with ID or ASD. Sixteen studies excluded previously hospitalized children (see online supplement for detailed characteristics of included studies). Across studies, information on 99 predictors of acute psychiatric hospital readmission was extracted (see online supplement). Of these predictors, 29 were reported by at least three studies, and thus combinable via meta-analysis. Survival data are provided separately (see online supplement) .

Sample

In total, studies provided data on 83,361 participants. Raw counts of readmitted children were available for 76,219 participants, of whom 10,076 (13.2%) were readmitted during the study duration, where mean±SD duration of follow-up for readmission was 15.9±15.0 months (N=29 studies). On average, participants were 14.0±2.2 years old (N=25), 51.4% were female (N=33), and 38.2% were from racial-ethnic minority groups (N=20). Mean LOS of index hospitalization was 16.6±7.8 days (N=19). Average time to readmission was 13.1±12.8 months (N=11). Table 1 presents the pooled OR estimates indicating the association between predictor variables and psychiatric readmission.
TABLE 1. Pooled odds ratios and heterogeneity statistics for each predictor of psychiatric readmission of children and adolescents in the studies revieweda
PredictorN of studiesReadmitted (N)Not readmitted (N)ORb95% CIQcI2 (%)dEgger's teste
Demographic        
 Age177,25564,4721.02.88–1.18134.088.1.58
 Gender197,20956,0461.13.98–1.3071.574.8.09
 Race102,0035,103.99.90–1.107.66.0.16
Family characteristic        
 Parent is primary caregiver3167528.63.38–1.04.06.0.26
 Family psychiatric history361149.93.36–2.385.7465.2.13
 History of abuse or neglect61,4492,8651.18.91–1.5212.4459.8.01
Safety factor        
 Suicide attempt53248341.47.97–2.2211.0963.9.02
 Nonsuicidal self-injury4812942.17.80–5.8812.0275.0.02
 Suicidal ideation41455692.351.64–3.37.68.0.03
Diagnosis        
 Psychotic disorder107,94357,8421.871.53–2.2832.7976.9.81
 Bipolar disorder55,57849,7521.441.23–1.686.1334.8.73
 Mood disorder, unspecified63,89814,9511.181.01–1.3813.1969.7.01
 Substance use disorder128,04559,432.75.59–.9646.0176.1.07
 Attention-deficit hyperactivity disorder55,55649,8691.161.00–1.353.79.0.67
 Autism spectrum disorder or intellectual disability56,60651,0981.491.16–1.919.6258.4.86
 Eating disorder32,7027,1731.641.18–2.272.7427.0.95
 Personality disorder72,6869,4601.641.23–2.1916.3263.2.1
 Internalizing disorder43611,5341.35.87–2.1114.9579.4.39
 Depressive disorder65,65050,0541.01.85–1.2114.1564.7.39
 Anxiety disorder97,73558,747.94.80–1.1014.5551.9.21
 Externalizing disorder75,28646,8241.01.92–1.101.6643.7.54
 Oppositional defiant disorder41,8126,521.96.67–1.372.84.0.94
 Conduct disorder3192876.98.43–2.262.893.8.6
 Posttraumatic stress disorder51,8076,2921.47.77–2.8215.4374.1.73
 Any psychiatric disorder51,9154,472.98.68–1.4137.2189.3.94
Hospitalization characteristic        
 Prior hospitalization61,2077,1662.511.76–3.5713.362.4.01
 Length of stay101,97111,2241.141.00–1.3042.8779.00.08
 Discharge to residential treatment89013,7711.841.07–3.164.7882.80.01
 Aftercare82,08011,097.95.60–1.4911.4793.7.87
a
Excludes studies reporting survival analyses.
b
Effect sizes for the individual test measures are in the original test direction—i.e., an odds ratio greater than 1 indicates that readmission is higher in the group of interest, whereas an odds ratio less than 1 indicates that readmission is lower in the group of interest.
c
Cochran’s Q is the weighted sum of the squared differences between individual study effects and the pooled effect across study.
d
The I2 statistic describes the percentage of variation across studies that is due to heterogeneity rather than to chance.
e
p value in Egger’s test for publication bias.

Sociodemographic Characteristics

Demographic characteristics.

In total, 17 studies provided a separate ES for the relationship between participant age and readmission (4, 5, 14, 15, 1724, 4043). An ES for the relationship between gender and readmission was reported in 19 studies (36, 15, 1724, 27, 40, 4245), and ten studies provided an ES for race-ethnicity and readmission (5, 15, 17, 18, 21, 23, 27, 43, 46, 47). When ESs were pooled, no significant differences in readmission by age, gender, or race-ethnicity were observed (see figure in online supplement). Heterogeneity was high among studies measuring readmission by age (I2=88.1%) and gender (I2=74.8%) and low among studies measuring readmission by race-ethnicity (I2=0.0%).

Family characteristics.

Although 20 family-related risk predictors were identified across all studies, only three predictors (primary caregiver, family psychiatric history, and history of child abuse or neglect) were measured in three or more studies and thus combinable via meta-analysis (Figure 1). Youths whose primary caregiver was a parent were at lower risk of readmission, compared with those with a nonparent caretaker (6, 14, 19), and this relationship approached significance (ORpooled=0.63, 95% confidence interval [CI]=0.38–1.04). Heterogeneity was very low (I2=0.0%). Family psychiatric history (5, 6, 48) and history of abuse or neglect (46, 18, 24, 42) were not significant predictors of readmission. Meta-regression suggested that older studies tended to find a stronger relationship between abuse or neglect and readmission (β=–0.09, p=0.025). Subgroup analyses suggested that the relationship between abuse or neglect and readmission was significant among studies in which the youths had a primary diagnosis of an externalizing disorder versus those without an externalizing disorder (ORpooled=2.92, 95% CI=1.38–6.18).
FIGURE 1. Forest plots for meta-analysis of odds of psychiatric readmission of children and adolescents, by family factora
aCoding was such that the right side of the diagram indicates higher odds of readmission when the predictor was present and the left side indicates lower odds of admission when the predictor was present.

Clinical Characteristics

Safety factors.

Four studies measured the relationship between suicidal ideation at index hospitalization and readmission (6, 24, 44, 49). Suicide attempt at index hospitalization (N=5 studies) (6, 23, 24, 44, 50) trended toward prediction of increased risk of readmission (ORpooled=1.47, 95% CI=0.97–2.22). Nonsuicidal self-injury (N=4) (5, 6, 24, 50) was not significantly associated with readmission. ESs for suicide attempt (63.9%) and nonsuicidal self-injury (75.0%) had moderate to high heterogeneity. Youths with suicidal ideation were at significantly higher risk of readmission, compared with those without suicidal ideation at index hospitalization (ORpooled=2.35, 95% CI=1.64–3.37). Heterogeneity was low (I2 =0.0%) (Figure 2).
FIGURE 2. Forest plots for meta-analysis of odds of psychiatric readmission of children and adolescents, by safety factora
aCoding was such that the right side of the diagram indicates higher odds of readmission when the predictor was present and the left side indicates lower odds of admission when the predictor was present.
Meta-regression analyses suggested that studies with older youths (β=0.29, p=0.06) and more female youths (β=0.03, p=0.04) tended to find a stronger relationship between suicide attempt and readmission. Subgroup analyses indicated that single-site studies (ORpooled=1.45, 95% CI=1.04–2.03), studies excluding those with ASD and ID (ORpooled=2.18, 95% CI=1.05–4.54), and studies restricted to children with a primary diagnosis of suicidality (ORpooled= 2.18, 95% CI=1.05–4.54) all found a significant relationship between suicide attempt at index hospitalization and readmission. Newer studies (β=0.23, p=0.03) with older (β=2.6, p=0.01) and more male youths (β=–0.61, p=0.04) tended to find a stronger relationship between nonsuicidal self-injury and readmission.

Diagnoses.

Forest plots for significant diagnostic predictors are presented in Figure 3. In regard to serious mental illness, 10 studies measured the relationship between presence of a psychotic disorder and readmission (3, 4, 6, 14, 19, 22, 40, 42, 51, 52). Youths with a psychotic disorder were at significantly higher risk of readmission, compared with those without a psychotic disorder (ORpooled=1.87, 95% CI=1.53–2.28). Heterogeneity was high (I2=76.9%). Meta-regression analyses suggested that studies with greater mean days until readmission tended to find a stronger relationship between psychotic disorders and readmission (beta=0.02, p<0.001). Subgroup analyses suggested that the significant relationship between psychotic disorders and readmission persisted when studies were divided by location (U.S. versus international) and study design (prospective versus retrospective; single site versus multisite).
FIGURE 3. Forest plots for meta-analysis of odds of psychiatric readmission of children and adolescents, by diagnosisa
aCoding was such that the right side of the diagram indicates higher odds of readmission when the predictor was present and the left side indicates lower odds of admission when the predictor was present.
Five studies measured the relationship between presence of bipolar disorder and readmission (4, 14, 19, 40, 51). Youths with bipolar disorder were at significantly higher risk of readmission, compared with those without bipolar disorder (ORpooled=1.44, 95% CI=1.23–1.68). The relationship between bipolar disorder and readmission reached significance only for U.S. studies (not for international studies).
Six studies measured the relationship between presence of unspecified mood disorder and readmission (6, 22, 42, 45, 51). Youths with a mood disorder were at higher risk of readmission, compared with those without a mood disorder (ORpooled=1.18, 95% CI=1.01–1.38). Older studies (β=–0.10, p=0.01) and studies with a longer mean LOS (β=0.16, p=0.003) reported a stronger relationship between unspecified mood disorder and readmission.
In regard to substance use disorders, 12 studies measured the relationship between presence of a substance use disorder and readmission (46, 19, 2224, 40, 42, 44, 45, 51). Youths with a substance use disorder were at significantly lower risk of readmission, compared with those without a substance use disorder (ORpooled=0.75, 95% CI=0.59–0.96). Heterogeneity was high (I2=76.1%). Studies with a smaller sample size (β=–0.00002, p=0.04) and a longer mean LOS (β=–0.10, p<0.001) tended to find a stronger relationship between presence of a substance use disorder and readmission. The relationship between substance use disorder and readmission was stronger for studies with a higher percentage of white youths (β=0.02, p=0.04) and older youths (β=0.11, p=0.009) and reached significance in prospective and multisite studies but not in retrospective or single-site studies.
Regarding neurodevelopmental disorders, five studies measured the relationship between presence of attention-deficit hyperactivity disorder (ADHD) and readmission (6, 14, 24, 40, 51). Youths with ADHD were at significantly higher risk of readmission, compared with those without ADHD (ORpooled=1.16, 95% CI=1.00–1.35). Heterogeneity was low (I2 =0.0%). Five studies measured the relationship between presence of ASD or ID and readmission (14, 19, 41, 42, 51). Youths with ASD or ID were at significantly higher risk of readmission, compared with those without ASD or ID (ORpooled=1.49, 95% CI=1.16–1.91). All studies were retrospective. Older studies (β=–0.05, p=0.003), studies with fewer female participants (β=–0.03, p=0.03), and studies with a longer LOS (β=0.03, p=0.004) tended to find a stronger relationship between ASD or ID and readmission.
Regarding other disorders, three studies measured the relationship between presence of an eating disorder and readmission (24, 42, 51). Youths with an eating disorder were at significantly higher risk of readmission, compared with those without an eating disorder (ORpooled=1.64, 95% CI=1.18–2.27). Seven studies measured the relationship between presence of a personality disorder and readmission (46, 22, 42, 45, 52). Youths with a personality disorder were at significantly higher risk of readmission, compared with those without a personality disorder (ORpooled=1.64, 95% CI=1.23–2.19). Studies with more female youths (β=0.02, p=0.02) and a shorter follow-up interval (β=–0.01, p=0.01) tended to find a stronger relationship between personality disorder and readmission.
Depressive disorders (N=6) (4, 6, 14, 40, 44, 51), anxiety disorders (N=8) (4, 6, 22, 24, 40, 42, 45, 51), internalizing disorders (N=4) (20, 43, 46, 50), oppositional defiant disorder (N=4) (14, 24, 27, 51), conduct disorder (N=3) (14, 24, 27), unspecified externalizing disorders (N=7) (20, 40, 4244, 46, 50), posttraumatic stress disorder (N=5) (4, 6, 14, 24, 51), and the presence of a formal psychiatric diagnosis (N=5) (14, 18, 23, 42, 45) were not significantly associated with risk of readmission (see figure in online supplement).

Hospitalization Characteristics

Forest plots for hospitalization characteristics are presented in Figure 4.
FIGURE 4. Forest plots for meta-analysis of odds of psychiatric readmission of children and adolescents, by hospitalization characteristica
aCoding was such that the right side of the diagram indicates higher odds of readmission when the predictor was present and the left side indicates lower odds of admission when the predictor was present.

Prior hospitalization.

Six studies measured the relationship between prior hospitalization and readmission (5, 6, 19, 2123). Youths with a prior hospitalization were at significantly higher risk of readmission, compared with youths without a prior hospitalization (ORpooled=2.51, 95% CI=1.76–3.57). Heterogeneity was moderate (I2=62.4%). Meta-regression analyses were not significant. Subgroup analyses suggested that this relationship persisted across study designs (prospective versus retrospective; single site versus multisite) and location (U.S. versus international).

LOS.

Ten studies measured the relationship between LOS and readmission (5, 6, 14, 15, 18, 2224, 45, 46). Youths with a longer LOS were at significantly higher risk of readmission, compared with those with a shorter LOS (ORpooled=1.14, 95% CI=1.00–1.30). Heterogeneity was high (I2=79.0%). Meta-regression suggested that larger studies (β=0.00004, p<0.001) and studies with a longer mean LOS (β=0.03, p<0.001) tended to find a stronger relationship between LOS and readmission.

Discharge to residential treatment.

Six studies measured the relationship between readmission and discharge to residential treatment following index hospitalization (5, 6, 19, 2123). Youths discharged to residential treatment were at significantly higher risk of readmission, compared with those discharged to nonresidential treatment (ORpooled=1.84, 95% CI=1.07–3.16). Heterogeneity was high (I2=82.8%). Meta-regression analyses were not significant. All but one of the studies were retrospective. Only one study included children with ASD or ID.

Aftercare.

Eight studies measured the relationship between provision of aftercare services and readmission (3, 14, 18, 19, 2123, 25). There was no significant relationship between aftercare services and readmission. Heterogeneity was high (I2=93.7%), and studies were mixed with regard to whether aftercare services were defined as referral to services, patient engagement, or family engagement. Meta-regression analyses found that older studies (β=–0.09, p=0.01) and studies with greater mean days to readmission (β=0.03, p<0.001) tended to report a stronger relationship between aftercare and readmission.

Study Bias

Of the 33 included studies, 27 had a representative sampling time frame, 24 utilized random selection or enrolled all eligible participants, and 11 had a representative national sample. The method for data collection was assessed as having a low risk of bias in 15 studies. Twenty studies had an acceptable timeline for assessing readmission, 21 studies had appropriately defined numerators and denominators, and 24 studies accounted for covariates in the final analyses. Overall risk of bias was high in six studies, moderate in 16 studies, and low in 11 studies.
Egger’s test (35) was positive, suggesting a significant degree of bias in the published literature, for the following predictors: suicidal ideation, suicide attempt and nonsuicidal self-injury, unspecified mood disorder, history of abuse or neglect, prior hospitalization, and discharge to residential treatment. For these predictors, small studies were more likely to find significant results, compared with larger studies. This may suggest that for these predictors, small studies with insignificant findings (small ES) are not published and thus not included in the meta-analysis, whereas small studies with significant findings are more likely to be published.

Discussion

For pediatric mental health admissions, indicators of clinical severity emerged as predictive of repeat hospitalization. Measures of clinical severity at the patient and hospital levels were more predictive of readmission than were sociodemographic characteristics. Studies were heterogeneous with regard to methodology and practice setting, and analyses of bias suggested a need for larger controlled studies and standardization of methodology.
In multiple cases, demographic features significantly moderated the size of the relationship between predictors and readmission. For example, although gender was not a significant predictor of readmission itself, studies with more females tended to find a positive relationship between history of suicide attempt and readmission, whereas studies with more males tended to find a positive relationship between history of nonsuicidal self-injury and readmission. This may suggest that controlling for gender is important in identifying predictors of readmission. Moreover, screening for nonsuicidal self-injury, particularly among males, for whom self-injury is less common (53), may help identify children at elevated risk of readmission. The wide breadth of acute mental health care settings of the included studies may have contributed to the nonsignificance of demographic features (age, gender, and race-ethnicity) in directly predicting readmission when ESs were pooled. Of note, no studies reported data for gender nonbinary youths.
This meta-analysis identified 20 different measures of family functioning (e.g., family risk index, parent-child conflict, and parental involvement). However, only three of these measures (primary caretaker, family psychiatric history, and history of abuse or neglect) were reported by more than one study, and none reached significance when pooled. There was a trend toward significance that youths whose primary caregiver was a parent were at lower risk of readmission. Future standardization of measures pertaining to family and peer factors would be beneficial, because homegrown measures are difficult to combine or compare.
Across all predictors, youths with suicidality and more severe illness experienced an elevated risk of readmission. Youths with suicidal ideation were readmitted at twice the frequency of youths without suicidal ideation. Suicidality is a known predictor of psychiatric readmission among adults (54), and as psychiatric inpatient units are increasingly transformed into short-term crisis intervention facilities, rates of admission associated with suicidality have increased (55). Because criteria for admission frequently revolve around suicidality, impulsivity, and situational aggression, admissions for suicidality may be an indirect reflection of readmission risk as perceived by the provider. Serious mental illness (psychotic and mood disorders) was associated with increased risk of readmission in this analysis. It is striking that the relationship between psychotic disorders and readmission persisted even with the wide variation in number of study sites, study design, mean LOS and duration of follow-up, and study location.
Unexpectedly, substance use disorders were associated with a reduced risk of readmission. This finding may suggest that youths with substance use disorders are referred to substance use treatment, experience poorer access to care, or are diverted from psychiatric readmission—for example, through incarceration in juvenile detention facilities. Alternatively, youths with substance use disorders may have a higher probability of loss to follow-up because of transition between systems of care, with relatively few studies tracking readmissions across institutions. Beyond substance use disorders, readmission estimates for populations that may require specialty mental health services, such as youths with eating disorders, may be more difficult to interpret because of movement between primary and specialty mental health care.
ASD and ID were associated with increased risk of readmission. Insufficient data were reported to distinguish rates of readmission between youths with ASD or ID and ASD alone. Of the five available studies investigating this population, none were prospective, none reported data for the racial-ethnic composition of the study population, none reported mean time to readmission, and only three included youths ever previously hospitalized. The aggregated national costs of supporting children with ASD in the United States is $61 billion per year (56). Research has suggested that increased spending on respite care is linearly associated in adjusted analyses with decreased odds of hospitalization (57). The results of this meta-analysis suggest that the availability of robust aftercare services and respite care may be beneficial for this population in mitigating frequent readmissions; however, overall, more data are needed regarding risk of readmission in ASD and ID populations, which will, in turn, inform interventions to reduce acute care utilization.
Personality disorders were also a significant predictor of readmissions. Personality disorders may increase risk of readmission secondary to illness burden, comorbidity, or suicidality and other safety factors. Moreover, this association might reflect that youths with frequent readmissions and suicidality are more frequently diagnosed as having personality disorders. In turn, youths with personality disorders may also be more frequently categorized as treatment resistant or more severely ill (5).
The relationship between clinical severity and readmission was also evident at the hospital level: prior hospitalization, discharge to residential treatment, and longer mean LOS were all associated with readmission. Diagnostic and utilization-related predictors are likely inherently confounded—for example, youths with serious mental illness are more likely to have prior hospitalizations. Across varied inpatient practice settings and diverse cohorts, the studies suggest that youths with higher clinical severity are more likely to experience frequent hospital admissions and early readmission. This may justify increased implementation of resource-intensive postdischarge interventions for severely ill youths to mitigate the costlier alternative of rehospitalization.
The long mean LOS (N=16.6 days) of the included studies may be an indicator of differences in resources across hospital systems reporting data on readmissions compared with those absent from this sample. The LOS may also reflect an evolution in the prevalence of managed care over the past 20 years. Included studies ranged in data collection from 1997 through 2018, and newer studies may more accurately reflect current managed care practices. There was a significant negative correlation between year of study and LOS (r=−0.693, p=0.001), consistent with the evolution of managed care practices over the past two decades. Of note, hospital-level predictors were pooled across a wide range of practice settings, including 11 international studies (Australia, New Zealand, Canada, Norway, China, Denmark, and South Africa) conducted within diverse health systems.
Several limitations of this meta-analysis should be acknowledged. This study describes an association between individual predictors and readmission; no causal inferences or assumption of multivariate adjustment can be made. Although in some cases, individual studies attempted to control for confounding variables, many did not. Potential sources of heterogeneity were explored by using subgroup and meta-regression analyses. Differences in study population and methodology may have contributed to the moderate to high ES heterogeneity of multiple predictors. Results from meta-regression analyses should be interpreted with some caution because of the possibility of type I errors (36, 37). Study bias was evident in the existing literature, with most studies rated as being at moderate risk of bias, suggesting that negative studies are less frequently published.
Because meta-analysis is inherently limited to predictors previously reported in the literature, this review may be missing potentially important predictor variables, such as type of aftercare interventions, outpatient care, distribution of specific diagnoses in the study population, community resources, various socioeconomic factors and other social determinants of health, and processes of care on the inpatient unit (e.g., medication changes). Several studies reported unexpected results—for example, an inverse relationship between presence of child abuse or neglect and readmissions (24). Modality of biopsychosocial and diagnostic assessment varied between studies with regard to extent of standardization (e.g., structured versus unstructured interview), and thus sensitivity of assessments across multiple domains likely differed.

Conclusions

Overall, the findings of this meta-analysis suggest that severely ill children and adolescents may benefit from targeted interventions to reduce readmission. Measurement of sociodemographic and clinical characteristics of the study population as well as clinical disease severity, aftercare services, and community characteristics is critical if readmission rates are to be compared across institutions and care environments. Standardization of methods to determine prevalence rates of child psychiatric readmissions and their predictors is urgently needed to inform a national strategy to reduce repeated child psychiatric hospital readmissions. As predictive modeling of readmissions becomes more readily accessible with the advent of high-dimensional electronic health record data, a foundational understanding of the existing literature helps drive formation of models of readmission that are clinically informed and specific to pediatric mental health.

Acknowledgments

The authors gratefully acknowledge the UCLA Institute for Digital Research and Education statistical consulting services.

Supplementary Material

File (appi.ps.201900234.ds001.xlsx)

References

1.
Torio CM, Encinosa W, Berdahl T, et al: Annual report on health care for children and youth in the United States: national estimates of cost, utilization and expenditures for children with mental health conditions. Acad Pediatr 2015; 15:19–35
2.
Zima BT, Rodean J, Hall M, et al: Psychiatric disorders and trends in resource use in pediatric hospitals. Pediatrics 2016; 138:e20160909
3.
Gearing RE, Mian I, Sholonsky A, et al: Developing a risk-model of time to first-relapse for children and adolescents with a psychotic disorder. J Nerv Ment Dis 2009; 197:6–14
4.
Bobier C, Warwick M: Factors associated with readmission to adolescent psychiatric care. Aust N Z J Psychiatry 2005; 39:600–606
5.
Barker D, Jairam R, Rocca A, et al: Why do adolescents return to an acute psychiatric unit? Australas Psychiatry 2010; 18:551–555
6.
McCarthy L, Pullen LM, Savage J, et al: Risk factors leading to increased rehospitalization rates among adolescents admitted to an acute care child and adolescent psychiatric hospital. J Child Adolesc Psychiatr Nurs 2017; 30:105–111
7.
Fontanella CA, Bridge JA, Campo JV: Psychotropic medication changes, polypharmacy, and the risk of early readmission in suicidal adolescent inpatients. Ann Pharmacother 2009; 43:1939–1947
8.
Moses T: Adolescents’ perspectives about brief psychiatric hospitalization: what is helpful and what is not? Psychiatr Q 2011; 82:121–137
9.
James S, Charlemagne SJ, Gilman AB, et al: Post-discharge services and psychiatric rehospitalization among children and youth. Adm Policy Ment Health Ment Health Serv Res 2010; 37:433–445
10.
Mangione-Smith R, Schiff J, Dougherty D: Identifying children’s health care quality measures for Medicaid and CHIP: an evidence-informed, publicly transparent expert process. Acad Pediatr 2011; 11(suppl):S11–S21
11.
Kocher RP, Adashi EY: Hospital readmissions and the Affordable Care Act: paying for coordinated quality care. JAMA 2011; 306:1794–1795
12.
Lyons JS, O’Mahoney MT, Miller SI, et al: Predicting readmission to the psychiatric hospital in a managed care environment: implications for quality indicators. Am J Psychiatry 1997; 154:337–340
13.
Arnold EM, Goldston DB, Ruggiero A, et al: Rates and predictors of rehospitalization among formerly hospitalized adolescents. Psychiatr Serv 2003; 54:994–998
14.
Romansky JB, Lyons JS, Lehner RK, et al: Factors related to psychiatric hospital readmission among children and adolescents in state custody. Psychiatr Serv 2003; 54:356–362
15.
Eltz M, Evans AS, Celio M, et al: Suicide probability scale and its utility with adolescent psychiatric patients. Child Psychiatry Hum Dev 2007; 38:17–29
16.
Foster EM: Do aftercare services reduce inpatient psychiatric readmissions? Health Serv Res 1999; 34:715–736
17.
Trask EV, Fawley-King K, Garland AF, et al: Do aftercare mental health services reduce risk of psychiatric rehospitalization for children? Psychol Serv 2016; 13:127–132
18.
Yampolskaya S, Mowery D, Dollard N: Predictors for readmission into children’s inpatient mental health treatment. Community Ment Health J 2013; 49:781–786
19.
Pieterse D, Temmingh H, Vogel W: Factors associated with readmission in South African adolescents discharged from two inpatient psychosocial rehabilitation units. J Child Adolesc Ment Health 2016; 28:199–212
20.
Fite PJ, Stoppelbein L, Greening L, et al: Child internalizing and externalizing behavior as predictors of age at first admission and risk for repeat admission to a child inpatient facility. Am J Orthopsychiatry 2008; 78:63–69
21.
Blader JC: Symptom, family, and service predictors of children’s psychiatric rehospitalization within one year of discharge. J Am Acad Child Adolesc Psychiatry 2004; 43:440–451
22.
Cheng C, Chan CWT, Gula CA, et al: Effects of outpatient aftercare on psychiatric rehospitalization among children and emerging adults in Alberta, Canada. Psychiatr Serv 2017; 68:696–703
23.
Fontanella CA: The influence of clinical, treatment, and healthcare system characteristics on psychiatric readmission of adolescents. Am J Orthopsychiatry 2008; 78:187–198
24.
van Alphen NR, Stewart JG, Esposito EC, et al: Predictors of rehospitalization for depressed adolescents admitted to acute psychiatric treatment. J Clin Psychiatry 2017; 78:592–598
25.
Carlisle CE, Mamdani M, Schachar R, et al: Aftercare, emergency department visits, and readmission in adolescents. J Am Acad Child Adolesc Psychiatry 2012; 51:283–293.e4
26.
Durbin J, Lin E, Layne C, et al: Is readmission a valid indicator of the quality of inpatient psychiatric care? J Behav Health Serv Res 2007; 34:137–150
27.
Chung W, Edgar-Smith S, Palmer RB, et al: Psychiatric rehospitalization of children and adolescents: implications for social work intervention. Child Adolesc Social Work J 2008; 25:483–496
28.
Moher D, Liberati A, Tetzlaff J, et al: Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. Ann Intern Med 2009; 151:264–269
29.
Booth A, Clarke M, Dooley G, et al: PROSPERO at one year: an evaluation of its utility. Syst Rev 2013; 2:4
30.
Walby FA, Myhre MØ, Kildahl AT: Contact with mental health services prior to suicide: a systematic review and meta-analysis. Psychiatr Serv 2018; 69:751–759
31.
Shamliyan TA, Kane RL, Ansari MT, et al: Development quality criteria to evaluate nontherapeutic studies of incidence, prevalence, or risk factors of chronic diseases: pilot study of new checklists. J Clin Epidemiol 2011; 64:637–657
32.
Cox DR: The analysis of multivariate binary data. Appl Stat 1972; 21:113
33.
DerSimonian R, Laird N: Meta-analysis in clinical trials. Control Clin Trials 1986; 7:177–188
34.
Rosenthal R: Meta-analysis: a review. Psychosom Med 1991; 53:247–271
35.
Egger M, Smith GD, Phillips AN: Meta-analysis: principles and procedures. BMJ 1997; 315:1533–1537
36.
Harbord RM, Harris RJ, Sterne JAC: Updated tests for small-study effects in meta-analyses. Stata J 2009; 9:197–210
37.
Harbord RM, Higgins JPT: Meta-regression in Stata. Stata J 2008; 8:493–519
38.
Sterne JAC, Bradburn MJ, Egger M: Meta-analysis in Stata, in Systematic Reviews in Health Care, 2nd ed. Edited by Egger M, Davey-Smith G, Altman DG. New York, Wiley, 2008
39.
Stata Statistical Software. College Station, StataCorp, 2017.
40.
Feng JY, Toomey SL, Zaslavsky AM, et al: Readmission after pediatric mental health admissions. Pediatrics (Epub Nov 3, 2017)
41.
Leon SL, Cloutier P, Polihronis C, et al: Child and adolescent mental health repeat visits to the emergency department: a systematic review. Hosp Pediatr 2017; 7:177–186
42.
Stewart SL, Baiden P: An exploratory study of the factors associated with medication nonadherence among youth in adult mental health facilities in Ontario, Canada. Psychiatry Res 2013; 207:212–217
43.
Fite PJ, Stoppelbein L, Greening L: Predicting readmission to a child psychiatric inpatient unit: the impact of parenting styles. J Child Fam Stud 2009; 18:621–629
44.
Czyz EK, Berona J, King CA: Rehospitalization of suicidal adolescents in relation to course of suicidal ideation and future suicide attempts. Psychiatr Serv 2016; 67:332–338
45.
Fadum EA, Stanley B, Qin P, et al: Self-poisoning with medications in adolescents: a national register study of hospital admissions and readmissions. Gen Hosp Psychiatry 2014; 36:709–715
46.
Leon SC, Stoner AM, Dickson DA: Does the hospital predict readmission? A multi-level survival analysis approach. Adm Policy Ment Health Ment Health Serv Res 2016; 43:514–523
47.
Stewart SL, Kam C, Baiden P: Predicting length of stay and readmission for psychiatric inpatient youth admitted to adult mental health beds in Ontario, Canada. Child Adolesc Ment Health 2014; 19:115–121
48.
Gearing RE, Charach A: Medication adherence for children and adolescents with first-episode psychosis following hospitalization. Eur Child Adolesc Psychiatry 2009; 18:587–595
49.
Enns MW, Cox BJ, Inayatulla M: Personality predictors of outcome for adolescents hospitalized for suicidal ideation. J Am Acad Child Adolesc Psychiatry 2003; 42:720–727
50.
Berona J, Horwitz AG, Czyz EK, et al: Psychopathology profiles of acutely suicidal adolescents: associations with post-discharge suicide attempts and rehospitalization. J Affect Disord 2017; 209:97–104
51.
Kagabo R, Hashibe M, Kim J, et al: The association between length of hospital stay and readmission for pediatric psychiatric patient. Journal of Psychiatry and Mental Health 2017; 2. doi
52.
Pedersen J, Aarkrog T: A 10-year follow-up study of an adolescent psychiatric clientele and early predictors of readmission. Nord J Psychiatry 2001; 55:11–16
53.
Claes L, Vandereycken W, Vertommen H: Self-injury in female versus male psychiatric patients: a comparison of characteristics, psychopathology and aggression regulation. Pers Individ Dif 2007; 42:611–621
54.
Verdoux H, Liraud F, Gonzales B, et al: Predictors and outcome characteristics associated with suicidal behaviour in early psychosis: a two-year follow-up of first-admitted subjects. Acta Psychiatr Scand 2001; 103:347–354
55.
Mellesdal L, Mehlum L, Wentzel-Larsen T, et al: Suicide risk and acute psychiatric readmissions: a prospective cohort study. Psychiatr Serv 2010; 61:25–31
56.
Buescher AVS, Cidav Z, Knapp M, et al: Costs of autism spectrum disorders in the United Kingdom and the United States. JAMA Pediatr 2014; 168:721–728
57.
Mandell DS, Xie M, Morales KH, et al: The interplay of outpatient services and psychiatric hospitalization among Medicaid-enrolled children with autism spectrum disorders. Arch Pediatr Adolesc Med 2012; 166:68–73

Information & Authors

Information

Published In

Go to Psychiatric Services
Go to Psychiatric Services
Psychiatric Services
Pages: 269 - 279
PubMed: 31822241

History

Received: 8 May 2019
Revision received: 1 August 2019
Accepted: 6 September 2019
Published online: 11 December 2019
Published in print: March 01, 2020

Keywords

  1. Admissions and readmissions
  2. Child psychiatry/general
  3. Readmission
  4. Hospitalization
  5. Meta-analysis

Authors

Details

Juliet Beni Edgcomb, M.D., Ph.D. [email protected]
Semel Institute for Neuroscience and Human Behavior, University of California, Los Angeles, Los Angeles (Edgcomb, Zima); Department of Pediatrics, Cincinnati Children’s Hospital, University of Cincinnati, Cincinnati (Sorter); Department of Psychiatry, University of Massachusetts Medical School, Worcester (Lorberg).
Michael Sorter, M.D.
Semel Institute for Neuroscience and Human Behavior, University of California, Los Angeles, Los Angeles (Edgcomb, Zima); Department of Pediatrics, Cincinnati Children’s Hospital, University of Cincinnati, Cincinnati (Sorter); Department of Psychiatry, University of Massachusetts Medical School, Worcester (Lorberg).
Boris Lorberg, M.D.
Semel Institute for Neuroscience and Human Behavior, University of California, Los Angeles, Los Angeles (Edgcomb, Zima); Department of Pediatrics, Cincinnati Children’s Hospital, University of Cincinnati, Cincinnati (Sorter); Department of Psychiatry, University of Massachusetts Medical School, Worcester (Lorberg).
Bonnie T. Zima, M.D.
Semel Institute for Neuroscience and Human Behavior, University of California, Los Angeles, Los Angeles (Edgcomb, Zima); Department of Pediatrics, Cincinnati Children’s Hospital, University of Cincinnati, Cincinnati (Sorter); Department of Psychiatry, University of Massachusetts Medical School, Worcester (Lorberg).

Notes

Send correspondence to Dr. Edgcomb ([email protected]).
Preliminary findings were presented at the annual meeting of the American Academy of Child and Adolescent Psychiatry, Seattle, October 22–27, 2018.

Competing Interests

The authors report no financial relationships with commercial interests.

Metrics & Citations

Metrics

Citations

Export Citations

If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download.

For more information or tips please see 'Downloading to a citation manager' in the Help menu.

Format
Citation style
Style
Copy to clipboard

View Options

View options

PDF/EPUB

View PDF/EPUB

Get Access

Login options

Already a subscriber? Access your subscription through your login credentials or your institution for full access to this article.

Personal login Institutional Login Open Athens login
Purchase Options

Purchase this article to access the full text.

PPV Articles - Psychiatric Services

PPV Articles - Psychiatric Services

Not a subscriber?

Subscribe Now / Learn More

PsychiatryOnline subscription options offer access to the DSM-5-TR® library, books, journals, CME, and patient resources. This all-in-one virtual library provides psychiatrists and mental health professionals with key resources for diagnosis, treatment, research, and professional development.

Need more help? PsychiatryOnline Customer Service may be reached by emailing [email protected] or by calling 800-368-5777 (in the U.S.) or 703-907-7322 (outside the U.S.).

Media

Figures

Other

Tables

Share

Share

Share article link

Share