Children With Medical Complexity and Mental and Behavioral Disorders in the Emergency Department Free

We performed a multicenter cross-sectional study using 2 data sources to evaluate the medical needs and overlap of patients with MBH disorders and those with medical complexity and to study the association between a MBH diagnosis and an outcome of admission or transfer.

1. First, we used the Pediatric Health Information System (PHIS) dataset, an administrative database coordinated by the Children’s Hospital Association (Lenexa, Kansas). We included data from 45 PHIS hospitals that contributed ED data continuously during the study period.

2. Second, we used the COMPdata dataset, an administrative database of inpatient discharge and outpatient records maintained by the Illinois Health and Hospital Association. COMPdata provides encounter-level data from participating community, tertiary care, and academic hospitals, encompassing approximately 90% of nonfederal hospitals within the state.¹⁷ 

Although PHIS allows for a detailed evaluation of resource utilization with patients with complex chronic conditions (CCC) using a generalizable sample of children’s hospitals, COMPdata allows for an evaluation of statewide data to nonpediatric facilities, which represent where most children obtain acute care in the United States.¹⁸  This study was approved by the Institutional Review Board at (place redacted for review). From each dataset, we included patients ≥2 and <21 years of age and with an ED encounter for the years 2016 to 2021. We did not include patients <2 years of age given the that these patients were much more likely to have diagnoses related to developmental delay, which would make comparisons with the typical MBH diagnoses more difficult to interpret.

We used the CCC algorithm, which assigns patients into categories of chronic disease based on procedure and diagnosis codes, but which does not provide a dedicated grouping for patients with MBH disorders at the encounter level. We used the latest version of the CCC system, which uses International Classification of Disease, 10th revision (ICD-10) diagnosis codes.¹⁹  We elected to use the CCC as it is a widely used method to identify patients with medical complexity in the ED setting for a variety of health services research and as it can be applied at the encounter level, but which does not have a dedicated subgrouping for MBH disorders, allowing for the identification of overlapping groups. This contrasts with other grouping systems for medical complexity, such as the Pediatric Medical Complexity Algorithm which does have mental health diagnoses in the algorithm.²⁰ 

To identify patients with MBH conditions, we used a rubric for these diagnoses using the Child and Adolescent Mental Health Disorders Classification System, which utilizes the International Classification of Diseases Tenth Revision, Clinical Modification diagnoses to classify child mental health disorders into 21 child mental disease categories and aligns with The Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition diagnosis groups.²¹ 

We identified 4 mutually exclusive subgroups: (1) patients with CCC without a MBH, (2) patients with a MBH disorder without a CCC, (3) patients with both a CCC and MBH disorder, and (4) patients with neither a MBH disorder nor CCC.

For patients with a CCC, we identified the number and types of CCCs present. Age was classified into groups of 2–5 years, 6–11 years, 12–18, and 19–21 years.²²  We included the social constructs of race and ethnicity as identified in administrative data given recognized disparities based on these variables^23,24  but did not incorporate this variable in multivariable modeling. Payor type was classified based on the presence of private insurance, public insurance, and other (which included categories of charitable, self-pay, and unknown). Length of stay was recorded in days. We evaluated costs from individual components of patient care and the performance of neuroimaging or laboratory testing. From PHIS, we acquired data with respect to medication use, including use of any medication, use of a psychotropic medication, and use of an intravenous medication and data with respect to subsequent hospital utilization, including inpatient hospitalization, ICU hospitalization, 1-year mortality, and aggregated standardized costs.

Our primary outcome of interest was a composite outcome of admission and/or transfer. This was done because some hospitals admit first to a medical bed and then transfer to a psychiatric bed, particularly in situations of capacity limits or if medical clearance is needed before psychiatric admission.

We evaluated the overlap between patients with CCC and MBH disorders. Among patients with MBH and a CCC, we characterized the type and number of CCCs. We evaluated counts of patients with CCC, MBH, and both CCC and MBH by fiscal quarter. We describe demographics and ED disposition by subgroup. We identified if a primary discharge diagnosis of an MBH was independently associated with a composite outcome of admission or transfer to another institution using a multivariable linear mixed model adjusting for number of CCCs, age, year of presentation, and payor status. We used the age group of 12 to 18 years as the reference group for age as this was the largest category. We present results as odds ratios with 95% confidence intervals (CI). Analyses were performed in R, version 4.1.1 (R Foundation for Statistical Computing, Vienna, Austria).

To further evaluate the role of specific MBH disorders on an outcome of admission or transfer among patients, we constructed an additional model for each dataset wherein we identified the type of MBH listed in the principal diagnosis (again keeping no MBH primary diagnosis as the reference group) and in which we adjusted for individual types of CCC. We grouped the granular Child and Adolescent Mental Health Disorders Classification System groupings into 11 larger groupings described by the Clinical Classifications Software.²¹ 

To evaluate longitudinal resource utilization related to patients with a concomitant MBH or CCC and compare it to children with only a single or neither disorder, we modified our PHIS inclusion criteria to be limited to the first encounter per patient from the years 2016 to 2020 and reported the following longitudinal outcomes within 365 days of their index discharge: ED encounters, admissions, and admissions requiring the ICU and cost. We performed our logistic regression model limited only to these unique patients.

For both datasets, inclusion is provided in Supplemental Fig 2 and demographics are provided in Table 1. Characteristics of CCCs and MBH disorders among patients with both conditions are provided in Table 2. Outcomes are provided in Table 3. Results of logistic regression modeling are provided in Supplemental Table 4 and 5.

From the PHIS dataset, we identified 11 880 930 encounters for patients. Of these, 80 452 (0.7%) had both a CCC and MBH. An MBH diagnosis was present among 7.8% of encounters with a CCC. The most common CCC was neuromuscular (35.3%), followed by technology dependence (31.3%) and gastrointestinal disease (30.6%). The most common type of MBH disorders were anxiety disorders (42.5%), depressive disorders (34.9%), and attention deficit hyperactivity disorder (31.5%).

Overall, 86.5% of encounters with an MBH and CCC had the composite outcome of admission or transfer, which was higher than the proportion admitted or transferred with an MBH alone (57.7%), with a CCC alone (52.0%), or without a CCC nor MBH (9.5%; Table 3). These encounters were almost always admitted to the hospital (98.8%) compared with those with an MBH alone (84.2%). Only 12.7% of patients with CCC and MBH were transferred, compared with 29.4% of patients with an MBH disorder only and 7.6% of patients with CCC only.

Among patients with overlapping MBH and CCC, 16 942 (21.1%) had an MBH primary diagnosis. In multivariable analysis, these patients had 1.41 higher odds of requiring admission or transfer compared when the primary diagnosis was related to an MBH problem alone (95% CI 1.33–1.50).

An analysis of the encounters within the PHIS dataset identified 5 290 300 unique patients between 2016 to 2020. Within this sample, 4 920 241 (93.0%) patients did not have an MBH or CCC, 262 501 (4.9%) had only a CCC, 88 473 (1.7%) had only an MBH diagnosis, and 19 085 (0.4%) had both an MBH and CCC. Patients with overlapping CCC and MBH had similar resource needs, mortality and ICU admission rates over 1 year compared with those with a CCC alone (Supplemental Table 6). For example, patients with overlapping MBH and CCC had 1-year readmission rates (33.9%) that were similar to those with CCC (31.2%), but higher than those with an MBH disorder only (15.8%). A logistic regression model among children with both a CCC and MBH performed on this subset again demonstrated a significant association with having a primary MBH diagnosis and a disposition of admission or transfer (adjusted odds ratio 1.53, 95% CI 1.36–1.72).

A model which adjusted for types of MBH and CCC identified that the association with the composite outcome of admission or transfer was strongest for diagnoses related to mental health history or mental health screening and mood disorders (Supplemental Table 7).

From the COMPdata dataset, 5 362 701 encounters were included. Of these, 11 724 (0.2%) had both a CCC and MBH. Among patients with a CCC, 6.4% had a concomitant MBH diagnosis. The most common types of CCC identified in this subgroup were metabolic (30.9%), cardiovascular (25.1%), and neuromuscular (19.8%). The most common MBH diagnoses were for anxiety (59.8%), depressive disorders (53.2%), and ADHD (29.6%). Patients with a CCC and MBH had a higher need for admission or transfer (61.8%) compared with those with a MBH disorder without a CCC (42.8%) and those with a CCC without MBH disorder (27.3%). Patients with an MBH disorder and CCC who met the outcome measure were more frequently admitted (90.3%) compared with patients with an MBH disorder (68.2%), but not compared with those with a CCC without an MBH disorder (91.8%). Patients with a concomitant MBH disorder and CCC were less frequently transferred (19.7%) compared with those with an MBH disorder alone (36.5%).

Overall, 4289 (36.3%) of patients with a CCC and MBH had an MBH primary diagnosis. We identified a multivariable odds of 3.47 (95% CI 3.13–3.84) of requiring admission or transfer when the primary diagnosis was for an MBH disorder. In a sensitivity analysis, a model which adjusted for types of MBH and CCC identified that the association with the composite outcome of admission or transfer was again strongest for mood disorders, ADHD, and diagnoses related to mental health history or mental health screening (Supplemental Table 8).

In both datasets, encounters for patients with an MBH, CCC, and both an MBH and CCC demonstrated an increase over time, superimposed over cyclical seasonal trends (Fig 1). All groups demonstrated a decrease during the second quarter of 2020, followed by a subsequent rise.

We identified distinct patterns in the ED distribution of CCC and MBH, both within pediatric hospitals and from a statewide sample of EDs. MBH disorders occurred in ∼ 5% of children in the ED with a CCC. Among this segment, an MBH disorder was the primary diagnosis among a quarter to one third of patients. Children with concomitant CCC and MBH disorder had a greater need for admission or transfer during their ED encounter relative to other subgroups of patients.

MBH considerations are an important challenge in the care of medically complex children. In a study evaluating trends and resource use for children with psychiatric emergencies in children’s hospitals, patients with greater medical complexity had a higher proportion of also having a concomitant psychiatric condition, which corroborates with our analysis.²⁵  The MBH diagnoses encountered by children with CCC may require differing approaches compared with those without medical complexity, particularly as a high proportion of patients with medical complexity (approximately 70% in one study) have developmental disabilities.²⁶  The care of these children likely requires dedicated mental health expertise, including the availability of mental health services which can support the needs of children who use nonverbal forms of communication, facilities which can simultaneously handle medical and behavioral issues, and use of pharmacologic aids for the deescalation of agitation and other behavioral emergencies. This may explain the higher association with admission or transfer among those with a primary diagnosis of a mood disorder and CCC, which includes depressive and bipolar disorders.

Our findings contribute important information on the MBH needs of children with medical complexity. Data from the National Survey of Children with Special Health Care Needs suggest that half of children with medical complexity had a mental health condition.²⁶  An evaluation of the National Health Interview Survey Disability Supplements reported that approximately 14% of children with disabilities receive mental health services.¹⁶  Our results corroborate the high demand for mental health services required by children with CCCs and demonstrates the high requirement for admission or transfer in this subgroup relative to patients with CCC not having MBH disorders.

Our findings fit into the context of literature demonstrative of the challenges of taking care of children with medical complexity within the healthcare system, where prior work has reported the unique challenges faced by parents and caregivers.^11,27  Children, with or without CCCs, face challenges in obtaining psychiatric care for multifactorial reasons, including limited access to child psychiatrists and a shortage of inpatient beds.²⁸  These challenges are likely compounded in medically complex children who have a concomitant MBH diagnosis. A longer term stay in the ED, termed “boarding,” can lead to issues of overcrowding, and the ED is poorly suited toward delivering psychiatric care. Alternatively, transfer of medically complex children to a distant place of care may limit the ability of families to participate in the care of the child.

Our comparison of children’s hospital data to statewide data also highlights important differences in decision making and resources at these 2 respective sites of care. Both datasets identified a high rate of admission or transfer among patients with overlapping CCC and MBH conditions. Among those meeting this outcome measure, admission to freestanding children’s hospitals was comparatively higher compared with those noted from the COMPdata hospitals. These data suggest that children’s hospitals almost uniformly admit patients with MBH in children with concomitant medical complexity. However, some pediatric facilities may not be licensed for psychiatric care and may at times have fewer mental health resources compared with psychiatric institutions. In contrast, the COMPdata patients with these disorders had lower acuity and had a higher rate of transfer, suggesting that a greater proportion of these facilities may be inadequately suited toward delivering care to children with CCCs.

Further work is needed to delineate best-practice guidelines for these children. The high proportion of children who are admitted or transferred in this subgroup suggests a need for improved outpatient care targeted toward optimization of mental health needs, caregiving needs, and family health. Important topics relevant to children with MBH disorders and CCCs identified in this study by the National Research Agenda on Health Systems for Children and Youth With Special Health Care Needs include those related to caregiving, family health, and principles of care (which includes offering medical home services, integration with community entities and resources, and facilitation of high-quality care transitions).²⁹  Hospital priorities include targeted questions to caregivers to identify optimal supports, the design of facilities to reduce unnecessary environmental stressors, and the consideration of organic causes for behavioral changes, which may be of greater relevance for medically complex children. One recent study using a multidisciplinary Delphi panel, for example, identified key quality metrics for the management of pediatric agitation, with a special emphasis on the management of children with developmental disorders.³⁰ 

Our findings are subject to limitations. This study was performed using administrative data that relied on diagnosis codes to identify CCC and an MBH disorders. As EDs may undercode some diagnoses, we suspect that findings may underestimate the frequency of these conditions among the study populations. We were unable to account for repeat encounters from the same patient in the COMPdata dataset. We were unable to ascertain relevant factors in the care of patients with MBH and CCC, such as ED or inpatient boarding, as we do not have reasons for admission or length of stay data in the ED in more granular units of hours. Although we used the primary diagnosis code to identify the primary reason for the encounter, chief complaint data are lacking in both datasets. Despite these limitations, the findings from this study demonstrate the unique healthcare needs of children with concomitant CCC and MBH disorders.

MBH disorders occur in ∼ 5% of encounters with a CCC who present to the ED. These patients have a higher requirement for admission or transfer relative to other patients. The presence of an MBH disorder as a primary diagnosis for these patients is associated with a need for admission or transfer. Further research is required to understand the specific types of MBH and identify best practice guidelines among this vulnerable group of children.