Pediatric Hospitalizations at Rural and Urban Teaching and Nonteaching Hospitals in the US, 2009-2019

Key Points Question How have pediatric inpatient hospitalizations at US urban and rural general hospitals and freestanding children’s hospitals changed from 2009 to 2019? Findings In this cross-sectional analysis of a national data set representing an estimated 23.2 million inpatient pediatric hospitalizations from 2009 to 2019, birth hospitalizations decreased by 10.6%, whereas nonbirth hospitalizations decreased by 28.9%. The largest decreases were at urban nonteaching and rural hospitals, most of which had 25 or fewer nonbirth pediatric hospitalizations in 2019. Meaning These findings suggest that clinical and policy initiatives to support hospitals with low pediatric volumes may be needed to maintain pediatric hospital access and capabilities, particularly in rural communities.


Introduction
In 2019, 5.2 million pediatric hospitalizations in the US incurred costs in excess of $46 billion, making inpatient care among the most costly components of pediatric health care. 1,24][5] An analysis of approximately 4400 hospitals surveyed annually by the American Hospital Association (AHA) found that less than half had pediatric units, with an almost 20% decrease from 2008 to 2018. 69][10] Although pediatric unit closures have received substantial media attention, the extent to which children continue to be hospitalized in general hospitals, including rural hospitals and urban teaching and nonteaching centers, is not well understood. 11,129][20] To address this knowledge gap, this study describes changes in the number and proportion of birth and nonbirth pediatric hospitalizations and health care costs at urban teaching, urban nonteaching, and rural hospitals compared with FCHs from 2009 to 2019; estimates the number and proportion of hospitals providing inpatient pediatric care; and characterizes changes in resource utilization, payer mix, and clinical complexity during the period.

Study Design and Data Sources
We conducted a retrospective cross-sectional analysis of the 2009, 2012, 2016, and 2019 Kids' Inpatient Database (KID), a nationally representative data set typically published every 3 years by the Healthcare Cost and Utilization Project (HCUP), Agency for Healthcare Research and Quality. 21The KID includes short-term, nonfederal, general, and specialty hospitals and excludes rehabilitation, long-term, and psychiatric hospitals. 21In 2009 and 2012, 44 states contributed to the KID; in 2016 and 2019, the KID included data from 47 and 49 states, respectively.Each data set represents a calendar year of inpatient hospitalization data for patients aged 20 years and younger, including a 10% systematic random sample of uncomplicated birth hospitalizations and an 80% sample of other birth and pediatric discharges.Data sets also include discharge weights for each hospitalization based on the universe of community nonrehabilitation hospitals surveyed annually by the AHA.
Birth and nonbirth hospitalizations among children and adolescents younger than 18 years were included in this analysis, because those aged 18 years or older are recognized as adults in most states and are typically admitted to adult beds at general hospitals. 22We excluded 4766 hospitalizations (<0.1%) that were missing a principal International Classification of Disease, Ninth Revision (ICD-9) or

International Statistical Classification of Diseases and Related Health Problems, Tenth Revision
(ICD-10) discharge diagnosis.
HCUP databases are limited data sets, and the Dartmouth-Health institutional review board determined that this study did not constitute human participants research; therefore, informed

JAMA Network Open | Pediatrics
Pediatric Hospitalizations at Rural and Urban Teaching and Nonteaching Hospitals consent was not needed, in accordance with 45 CFR §46. 23The methods and results reported here adhere with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies and HCUP data use agreement requirements. 24

Hospital Characteristics
We categorized hospitals into 4 mutually exclusive types: rural, urban nonteaching, urban teaching (together comprising general hospitals), and FCHs.Given the rarity of rural teaching hospitals, rural hospitals are not stratified according to teaching status in the KID. 21Teaching hospitals are defined as having Accreditation Council for Graduate Medical Education-approved residency programs, membership in the Council of Teaching Hospitals, or a ratio of full-time interns and residents to beds of greater than or equal to 0.25. 21FCHs are assigned to a separate stratum in the KID.We estimated the annual number of birth and nonbirth hospitalizations at each hospital, applying weights to generate hospital-level estimates.For example, for nonbirth hospitalizations, the number of hospitalizations per hospital was determined by multiplying the number of encounters by 1.25, because the KID includes an 80% sample of these discharges. 21Additional hospital characteristics included geographic region based on US Census Regions and hospital ownership.

Hospitalization Characteristics
For each hospitalization, we examined patient age at admission in years, binary sex (male or female as reported in the KID), race and ethnicity reported to HCUP by partner organizations and collected according to each hospital's standard operating procedures, 25 expected primary payer, and median household income of each patient's zip code (in quartiles).[28] Birth hospitalizations were identified using the KID in-hospital birth indicator variable and were categorized as uncomplicated or complicated according to Diagnosis Related Groups. 210][31] Similarly, the Children with Disabilities Algorithm was applied to identify those with disabilities, and the Child and Adolescent Mental Health Disorders Classification System was used to identify those with 1 or more mental health diagnoses in any ICD-9 or ICD-10 position. 32,33r each hospitalization, we determined length of stay (LOS) in days, total costs adjusted to 2019 dollars, whether the hospitalization followed a transfer from another acute care hospital (which includes transfers from their emergency departments), and whether the hospitalization resulted in a transfer to another short-term hospital. 21Transfers from other hospitals were limited to nonbirth hospitalizations.Costs were examined using data on total charges (total amount billed), which were converted to costs (expenses incurred in the provision of care) using HCUP cost-to-charge ratios. 34 adjusted costs from 2009 to 2016 to represent 2019 dollars using the Personal Consumption Expenditures index using a guide published by the Agency for Healthcare Research and Quality. 35

Statistical Analysis
After applying the KID survey weights to generate national estimates, we summarized sociodemographic and hospital characteristics using counts and percentages with associated 95% CIs.Using point and interval estimators for survey data, we determined the sampling probabilityweighted number and proportion of birth and nonbirth hospitalizations and inpatient health care costs at each hospital type each year and plotted these against year with associated 95% CIs.To characterize changes in clinical complexity over time, we determined the proportion of complicated birth hospitalizations at each hospital type, as well as the proportion of nonbirth hospitalizations with 1 or more complex chronic disease diagnosis, disability diagnosis, and mental health diagnosis.
Taylor series methods were used to approximate SEs and construct 95% CIs, 36

Hospitalization Characteristics
Across the 4 years of data, the KID included 12 809 624 million unweighted hospitalizations

Changes in Clinical Complexity Over Time
From 2009 to 2019, there were significant monotonic increases at all hospital types in the proportion of complicated birth hospitalizations, and in nonbirth hospitalizations with complex chronic disease, disability, and mental health diagnoses (Figure 2; eTable 5 in Supplement 1

Changes in LOS, Costs, and Interfacility Transfers
In all years, geometric mean LOS was longest at FCHs (

Discussion
This retrospective, cross-sectional analysis of nationally representative data found significant decreases in both birth and nonbirth pediatric hospitalizations from 2009 to 2019, with shifts in where care was provided.In urban communities, birth and nonbirth hospitalizations shifted from nonteaching to teaching hospitals, while the number of hospitalizations at FCHs remained relatively stable.In rural communities, birth hospitalizations decreased by approximately 25%, while nonbirth hospitalizations decreased 4-fold.These findings align with national birth statistics and obstetricsfocused research. 18,19,37Throughout this period, an increasing proportion of hospitalizations in all settings were experienced by children with disabilities, mental health diagnoses, and complex chronic diseases.In 2019, approximately two-thirds of hospitals in the data set provided birth hospitalization care, and nonbirth pediatric hospitalizations occurred in more than 3500 hospitals.
These findings build on a prior analysis 6 of AHA data that found a substantial decrease from 2008 to 2018 in the number of US hospitals with pediatric units, from 1753 to 1418 hospitals.Using data from a similar period, the present study illustrates that pediatric hospitalizations occurred at a much larger number of hospitals, including 1256 rural hospitals, 1333 urban teaching hospitals, and 843 urban nonteaching hospitals, in addition to FCHs.Prior single-state analyses 38 have shown that even following pediatric unit closures, almost one-quarter of hospitals continue to admit children at volumes similar to those experienced before unit closure.This national study supports these findings, because the number of hospitals admitting children exceeded the number of hospitals

Figure 1 .
Figure 1.Birth and Nonbirth Pediatric Hospitalizations and Total Hospitalization Costs at Rural Hospitals, Urban Nonteaching and Teaching Hospitals, and Freestanding Children's Hospitals, 2009-2019

accounting for JAMA Network Open | Pediatrics Pediatric
Hospitalizations at Rural and Urban Teaching and Nonteaching Hospitals Regression analyses were used to compare health care utilization in 2019 compared with 2009, using Poisson regression for LOS, linear regression for costs, and logistic regression for interfacility transfers, clustering hospitalizations within hospitals and computing SEs that were robust to misspecification of the within-hospital covariance matrix by using survey design estimators.Analyses were conducted using R statistical software version 4.1.3(R Project for Statistical Computing), SAS statistical software version 9.4 (SAS Institute), and Python programming language version 3.8.10(Python Software Foundation).The threshold for statistical significance was 2-sided P < .05.We reported the amount of missing data in the tables and figures and used pairwise deletion as appropriate.Data were analyzed from February to June 2023.

Table 1 .
Sociodemographic, Clinical, and Hospital Characteristics of Pediatric Hospitalizations, Weighted National Estimates, 2009-2019 (continued) Other race is a category used by the Kids' Inpatient database; additional description about this variable at the national level is not provided.Starting in 2014, more hospitals were categorized as urban teaching hospitals because there was an increase in facilities with approved residency programs in the American Hospital Association Annual Survey, from which hospital characteristics are derived.The Accreditation Council for Graduate Medical Education became the primary body for residency approval around this time.
c e Quartile ranges for median household income at home zip code vary annually (eg, g h In 2009, there were 238 819 hospitalizations (3.7%) with missing data.

Changes in Hospital Pediatric Volume Over Time The
proportion of hospitals in the KID with birth hospitalizations was not significantly different in 2019 vs 2009, but the absolute number decreased from 2784 to 2666, despite a larger number of states represented in the 2019 KID (Table3).Median birth hospitalization volumes per hospital remained stable overall.The largest decrease from 2009 to 2019 was observed at urban nonteaching hospitals, where the number of hospitals with births decreased from 1149 (65.6%) to 615 (54.9%),

Table 2 .
Length of Stay, Hospitalization Costs, and Interfacility Transfer Rates by Hospital Type, Weighted National Estimates, 2009-2019