Pediatric Capabilities in US Hospitals

Abstract

Objective:

To evaluate how availability of pediatric-capable hospitals has changed over 2 decades.

Methods:

We studied US acute care hospitals from 2003–2022 using the Healthcare Cost and Utilization Project Kids’ Inpatient Database. We determined the annual proportion of hospitals by pediatric capability level from 1–4, defined by which of 24 services were provided (Level 1: broadest range, Level 4, minimal services). We used Poisson models adjusting for hospital characteristics to evaluate changes from 2003 to 2022. We also measured changes in provision of each specific service.

Results:

We analyzed a mean 3,927 hospitals per year, which in 2022 were 67.3% non-profit, 39.7% urban teaching, 47.0% small, and 37.4% in the South. The number of high-capability hospitals decreased over time. Adjusting for hospital characteristics, Level 1 hospitals decreased 38% (95% confidence interval [CI] 27–48), Level 2 hospitals decreased 54% (95% CI 47, 60), Level 3 hospitals decreased 48% (95% CI 43, 53), and Level 4 hospitals increased 137% increase (95% CI 119, 156). For 17/24 (71%) capabilities, fewer hospitals were capable in 2022 than 2003. The steepest declines of hospital-level specific services were for moderate-intensity services including appendectomy (−50.5%), pneumonia hospitalization (−42.3%), and asthma hospitalization (−41.1%). The smallest changes were for extracorporeal membrane oxygenation (+0.8%), organ transplant (0.0%), and Fontan surgery (−0.1%).

Conclusions:

Low-capability hospitals more than doubled over two decades, corresponding to substantial losses in hospitals at all higher levels of pediatric capability. This would be expected to limit child access to care across a range of conditions and complexity.

INTRODUCTION

Approximately 70% of pediatric hospitalizations occur in general hospitals that care for both children and adults.¹ Since the 2000s, there has been a decreasing propensity to provide definitive care among general hospitals, evidenced by increasing transfer rates from emergency departments. Transfers have become more common for general pediatric complaints (such as abdominal pain) and conditions (such as asthma and gastroenteritis).^2,3 This is at least partially due to closures of pediatric inpatient units, with 30% closing from 2008 to 2022.⁴ This has resulted in regionalization of pediatric inpatient care,⁵ general hospital divestment from pediatrics in the United States,⁶ and a declining share of hospitalizations at non-children’s hospitals.⁷

Hospitals provide a wide variety of pediatric services, known as capabilities. Previous studies of regionalization have often relied upon structural measurements of capability such as inpatient beds. However, the presence or absence of certain structures does not guarantee that hospitals will provide a range of services. For example, having pediatric inpatient beds does not mean that a hospital will hospitalize a child with congenital heart disease. Lacking pediatric inpatient beds does not mean that a hospital will not hospitalize a child on an adult ward.⁸ Capabilities range from basic services such as laceration repair (present in nearly all hospitals) to highly specialized care such as complex congenital heart surgery or organ transplant. We recently derived and validated a system that can assign each hospital a level based on which services are provided (Table 1).⁹ Level 1 represents high-capability hospitals providing many advanced pediatric services, Levels 2 and 3 provide moderate-intensity pediatric services, and Level 4 represents low-capability hospitals providing virtually no pediatric services outside of the emergency department or obstetrical unit. Advantages of this system over measuring structural capabilities (e.g., presence of inpatient beds) is that it is more pragmatically based on what hospitals do rather than what they have, is more granular, and helps policymakers understand what real capabilities exist within the system.

TABLE 1:

Definitions of hospital levels based on validate prior evidence⁹.

Level 1
Capable of at least 7 of: - Cerebral ventricular shunt - Chemotherapy hospitalization - Congenital heart disease hospitalization - Cystic fibrosis hospitalization - Diabetic ketoacidosis hospitalization - Dialysis - Endoscopy - Extracorporeal membrane oxygenation - Fontan surgery - Gastrostomy - Inflammatory bowel disease hospitalization - Intensive care unit hospitalization - Metabolic disorder hospitalization - Organ transplant - Sickle cell anemia hospitalization - Ventricular septal defect surgery
Level 2
Does not meet level 1 criteria above AND Capable of at least 7 of: - Any capabilities from level 1 above - Appendectomy - Closed reduction of radius - Complex hospitalization - Lumbar puncture - Mechanical ventilation - Pneumonia hospitalization
Level 3
Does not meet criteria for levels 1–2 above AND Capable of at least 4 of: - Any capabilities from levels 1–2 above - Acute hospitalization - Asthma hospitalization
Level 4
Does not meet criteria for levels 1–3 above

Understanding the distribution and availability of pediatric capabilities as well as trends in gains or losses may inform planning in order to ensure high-quality pediatric care. Given that children will continue to visit emergency departments at nearly all acute care hospitals, access to high-quality care will depend on linking children with the capabilities needed to provide safe, timely, and efficient care. Our primary objective was therefore to evaluate changes in the proportion of hospitals at each of 4 levels of pediatric capability over 2 decades. Our secondary objective was to describe changes in the provision of each specific service..

METHODS

We studied hospitalizations at acute care hospitals in the United States using the Kids’ Inpatient Database (KID) from 2003–2022. KID is released approximately every 3 years, sampling 80% of all pediatric non-newborn hospitalizations nationally. The number of contributing states increased from 36 to 49 over the study period. All non-birth hospitalizations for children under 16 years old¹⁰ were eligible. KID allows national estimates of hospitalizations through its complex survey weighting.

Outcomes

The main outcome was pediatric hospital level, assigned as one of four tiers (Level 1, the highest capability to Level 4, the lowest capability) based on a previously-developed and validated system⁹ for categorizing hospitals based on the presence or absence of 24 capabilities. These capabilities range in complexity from asthma hospitalizations to performing congenital heart disease surgeries. Conceptually, Level 1 hospitals have a broad range of pediatric services and Level 4 hospitals have minimal capabilities. Of the 24 capabilities, 16 contribute to defining Level 1 hospitals, 6 contribute to Level 2 hospitals, and 2 contribute to Level 3 hospitals. Level 4 is defined by an absence of these capabilities (Table 1). Hospitals that performed a service >2 times based on the weighted count were considered as having performed it.

We extended the definitions to the International Classification of Diseases (ICD) 9th Edition since our analysis included years when the ICD-9 standard was used in the US (Supplemental Methods). Capability definitions are listed in Supplemental Table 1. To evaluate whether there were issues with varying sensitivity across the ICD-9 to ICD-10 transition in 2015, we formally evaluated immediate changes in hospital capabilities at the time of the transition (Supplemental Figures 1–2). Four (17%) capabilities significantly increased and 3 (13%) decreased at the time of the transition.

Secondary outcomes included each unique capability. The main exposure was year.

Variables

We obtained hospital characteristics from KID including hospital size (small, medium, large, or children’s hospital defined using KID definitions¹¹ that were region- and urbanicity-specific for size and based on data from the American Hospital Association and the National Association of Children’s Hospitals and Related Institutions for children’s hospital), ownership (public, not-for-profit, or for-profit), region (Northeast, Midwest, South, or West), and location/teaching status (rural, urban nonteaching, or urban teaching).

Analysis

We first determined the proportion of hospitals at each capability level by year. To evaluate the relative change in hospitals and hospitalizations from 2003 to 2022, we used Poisson regression to estimate incidence rate ratios and 95% confidence intervals (CI). To account for the increasing number of states in the sample over time, the model was estimated with an offset for the log total number of hospitals in the sample for each year. The inclusion of additional states in later years also changed the composition of the sample to increase the number of small hospitals, which would be expected to offer fewer services. To address sampling changes over time, we re-estimated the regression model, adjusting for hospital size, region, ownership, and location/teaching status. We performed three sensitivity analyses. First, to test the sensitivity of study conclusions to the requirement that a service be performed more than 2 times, we recalculated proportions of hospitals by capability level and year relaxing the capability requirement to include any provision of the service. Second, given changes in the hospital sample over time by region, we sought to fully isolate the region effect. To do so, we re-estimated the adjusted model per region. Third, we re-evaluated the change in hospital numbers by level by avoiding the ICD-9 to ICD-10 transition. In this analysis we repeated the adjusted model, comparing the earliest post-transition year (2016) to the latest (2022).

To evaluate how hospitalization counts occurring in different hospital levels changed, we counted the number of weighted hospitalizations at each level and using each capability by year. We used a survey-weighted contrast of the ratio of hospitalizations in 2022 compared with 2003. This hospitalization-level analysis did not require adjustment for hospital-level covariates because that was inherent to the complex survey design.

Next, we measured percentage of hospitals with each specific capability by year. To describe changes in capabilities we used the difference in proportions of hospitals between 2003 and 2022. CIs were calculated using bootstrapping with 1000 replicates for each computed risk difference. We repeated this analysis for each of the 24 specific capabilities.

We next examined which capabilities accounted for changes in hospital level. We compared the changes in performance of each capability for all Level 1 capabilities to all Level 2 capabilities using a Wilcoxon rank sum test. We repeated this pairwise between all capability levels.

Analyses were conducted using R version 4.4.0 and the survey¹² package. For all hospitalization-level analyses, we used weighted counts and statistics in accordance with the complex KID survey design. Statistical significance was set at alpha=0.05. This study was considered exempt by the local Institutional Review Board.

RESULTS

We analyzed a mean 3,927 hospitals per data year. Compared to 2003, the hospital sample in 2022 had higher proportions of nonprofit, urban teaching, small, and southern hospitals (Table 2). Seventy (1.8%) were children’s hospitals.

TABLE 2:

Demographic features of hospitals in 2003 and 2022.

	2003 N=3438	2022 N=3811
Ownership
Public	582 (34.9)	642 (16.8)
Nonprofit	686 (41.1)	2566 (67.3)
For profit	400 (24.0)	603 (15.8)
Location/Teaching
Rural	1408 (41.1)	1312 (34.4)
Urban nonteaching	1413 (41.2)	987 (25.9)
Urban teaching	605 (17.7)	1512 (39.7)
Size
Small	1135 (33.1)	1793 (47.0)
Medium	998 (29.1)	959 (25.2)
Large	1255 (36.6)	989 (26.0)
Children’s	42 (1.2)	70 (1.8)
Region
Northeast	420 (12.2)	519 (13.6)
Midwest	1250 (36.4)	1101 (28.9)
South	1068 (31.1)	1426 (37.4)
West	700 (20.4)	765 (20.1)

The proportion of high-capability hospitals declined over time (Figure 1), regardless of the threshold for defining capability (Supplemental Figure 3). There was a relative decrease in hospitals in capability Levels 1–3 though hospitalizations significantly decreased only in Levels 2 and 3 (Table 3). The largest change in hospital proportion occurred in the lowest-capability level (Level 4), increasing from 26.7% to 65.0% of all hospitals, a 137% (95% CI 119, 156) increase. After adjusting for hospital characteristics, the direction of the results was unchanged. There was no significant change in hospitalizations at Level 4 hospitals, with 99.0% and 98.1% of hospitalizations occurring in Levels 1–3 in 2003 and 2022 respectively. The direction of associations did not change when stratified by region (Supplemental Figure 4). In the sensitivity analysis comparing 2022 to 2016, Level 1 hospitals did not change significantly (−7%, 95% CI −21, 10), Level 2 hospitals decreased (−15%, 95% CI −27, −2), Level 3 hospitals decreased (−29%, 95% CI −35, −22), and Level 4 hospitals increased (18%, 95% CI 12, 25).

FIGURE 1:

Proportion of hospitals meeting criteria for each capability level by year. Alt text: Figure with lines depicting the percentage of hospitals at each of pediatric capabilities levels 1 to 4 from 2003 to 2022. The line for level 4 starts in the 20–30% range and increases steadily until it is above 60%. The lines for levels 2 and 3 each decline, while level 1 appears fairly stable over time. Level 1 is the lowest line, just below 10% of hospitals each year.

TABLE 3:

Relative changes in number of hospitals and number of patients visiting hospitals at each capability level. Adjusted changes in hospital counts accounted for hospital size, region, ownership, and location/teaching status.

Hospital capability level	Hospitals
	2003 N (%)	2022 N (%)	Unadjusted change (%)	Adjusted change (%)
All levels	3,438	3,811	*	*
Level 1	313 (9.1)	280 (7.3)	−19 (−31, −5)	−38 (−48, −27)
Level 2	800 (23.3)	323 (8.5)	−64 (−68, −59)	−54 (−60, −47)
Level 3	1,406 (40.9)	732 (19.2)	−53 (−57, −49)	−48 (−53, −43)
Level 4	919 (26.7)	2,476 (65.0)	143 (125, 162)	137 (119, 156)
	Hospitalizations**
	2003 N** (%)	2022 N** (%)	Change (%)
All levels	2,194	1,409	−36 (−43, −29)
Level 1	1,304 (59.4)	1,131 (80.3)	−13 (−27, 0)
Level 2	571 (26.0)	158 (11.2)	−72 (−76, −68)
Level 3	296 (13.5)	93 (6.6)	−69 (−73, −65)
Level 4	23 (1.0)	27 (1.9)	17 (−1, 35)

^*Change in absolute number of hospitals was not reported because it was driven by the change in number of states contributing data over the study period

^**Thousands of hospitalizations

Provision of specific services generally decreased from 2003 to 2022. For 17/24 (71%) capabilities, fewer hospitals provided these services in 2022 than in 2003. There was no significant change for 6 (25%) capabilities, and extracorporeal membrane oxygenation capability was available at more hospitals (Figure 2). The capabilities with the greatest loss of hospitals were appendectomy, which was performed in 72.0% of hospitals in 2003 and 21.5% in 2022 (change −50.5%, 95% CI −52.5, −48.5), pneumonia hospitalization (77.0% in 2003, 34.7% in 2022, change −42.3%, 95% CI −44.2, −40.3), and asthma hospitalization (77.9% in 2003, 36.8% in 2022, change −41.1%, 95% CI −43.0, −39.0). The capabilities with the least loss were extracorporeal membrane oxygenation (2.2% in 2003, 3.0% in 2022, change 0.8%, 95% CI 0.1, 1.5), organ transplant (2.1% in 2003, 2.1% in 2022, change 0.0%, 95% CI −0.7, 0.7), and Fontan surgery (2.5% in 2003, 2.4% in 2022, change −0.1%, 95% CI −0.8, 0.5). Among pediatric capabilities, national patient volume decreased for 9 (53%) of the corresponding services, had no significant changes in 2 (12%), and increased in 6 (35%) (Supplemental Figure 1).

FIGURE 2:

Change in proportion of hospitals providing each service from 2003 to 2022. Changes are shown as differences in percentage points with 95% confidence intervals. Alt text: Figure with bars depicting the change in the number of capable hospitals providing each service we measured. Provision of each service decreased between 2003 to 2022. In descending order of how much each service declined (steepest declines first): appendectomy, pneumonia hospitalization, asthma hospitalization, acute hospitalization, lumbar puncture, complex hospitalization, closed reduction of radius, diabetic ketoacidosis hospitalization, gastrostomy, endoscopy, congenital heart disease hospitalization, sickle cell anemia hospitalization, metabolic disorder hospitalization, cystic fibrosis hospitalization, mechanical ventilation, inflammatory bowel disease hospitalization, ICU hospitalization, cerebral ventricular shunt, dialysis, chemotherapy hospitalization, ventricular septal defect surgery, Fontan surgery, organ transplant, extracorporeal membrane oxygenation.

Compared with capabilities defining Level 1 hospitals, those defining Level 2 (p=0.004) and Level 3 (p=0.03) had larger decreases in hospitals performing them. There was no significant difference (p=0.64) between Levels 2 and 3 (Figure 2).

DISCUSSION

Over two decades, fewer pediatric services were provided across US hospitals. This resulted in Level 4 hospitals, those with the fewest pediatric capabilities, more than doubling, accounting for 27% of hospitals in 2003 and 65% in 2022, an adjusted increase of 137%. The change was largely driven by losses in each of the other capability levels. Our findings indicate that a small subset of hospitals provides the broadest range of services, and that the moderate range of pediatric services decreased most substantially.

Why are hospitals divesting from moderate-intensity pediatric services? These Level 2 and 3 services generally require a pediatric inpatient unit, the capability to perform common pediatric surgeries, or both. Prior research shows that many hospitals have closed their pediatric inpatient units and fewer are providing pediatric surgical services.^4,13 This demonstrates that the functional losses reported in this study correspond with previously-observed structural ones. The losses have been particularly acute in rural areas.¹⁴ The reasons for loss of pediatric inpatient and surgical capabilities are not clear and likely vary between hospitals. However, there are several forces that favor divestment from pediatrics. First, pediatric inpatient care has lower reimbursement compared to adults due to higher Medicaid enrollment rates, lower Medicaid reimbursement rates compared with private insurance and Medicare, and lower care intensity per bed day.^15–18 Second, pediatric staffing is difficult to maintain with specialized nurse and physician shortages, particularly after the COVID-19 pandemic.^19,20 Third, children with medical complexity comprise an increasing share of hospitalizations and may require broader services that are less available in community hospitals.²¹ Finally, hospitalizations for children have been declining, falling 26% from 2000 to 2019.⁷ While it is not yet clear whether this was a temporary response to the pandemic, the decreased demand may have driven a decreased need for supply. There are few incentives to bring back such capacity after it is lost.

Concentration of pediatric hospital services in fewer hospitals is an example of regionalization.⁶ Although regionalization may have benefits for certain pediatric subspecialties and outcomes²², declines in pediatric inpatient capacity can have negative ramifications for care delivery and access such as longer transfer times, reduced access, and burdens on families.^23,24 The impacts of pediatric inpatient capacity losses are unclear. When capacity is matched to typical demand, the system is less able to absorb volume surges. As an example, recent concurrent respiratory virus epidemics led to overcrowding at pediatric centers.²⁵ To date, regionalization has not shown signs of slowing and it remains to be seen whether there is a floor on pediatric capacity.

Improving Medicaid reimbursement could help reduce the financial motivation for hospitals to cater preferentially to adult patients.²⁶ Ensuring emergency departments are ready for children, and providing certification to facilities for high readiness could ensure hospitals without pediatric inpatient capability can stabilize children before transfer.^27,28 New directions are also possible, including enacting a program for pediatric critical access hospitals, deepening investments in safety net hospitals, or expanding pediatric emergency or inpatient teleconsultation.²⁹

This study had limitations. We used weighted counts to qualify hospitals as capable, and it is possible that some hospitals missed qualifying for being considered capable of a service if we undersampled qualifying hospitalizations. However, with an 80% per-hospital sample and a low threshold for being considered capable, we feel this is unlikely. The KID database does not include all US hospitals, states were added over the life of the study, and KID does not include a consistent hospital identifier between years to understand how hospitals entered or left the sample. These factors could lead to bias in estimates of the proportion of capable hospitals. This would be unlikely to affect trends, since analyses adjusted for region did not substantially differ from the unadjusted. Finally, we used updated definitions of the 24 capabilities because of the need to bridge the transition from ICD-9 to ICD-10 codes. That transition could have biased estimates of capability over time if ICD-9 codes were more sensitive. We may have undermeasured the 6 capabilities that used CPT codes in the original definitions, since CPTs, which are typically coded in ED discharges, were not available in this study. However, more conditions significantly increased in prevalence than decreased, suggesting coding changes were unlikely to be a major contributor to study findings of decreasing capability nationally.

Over two decades, low-capability hospitals more than doubled due to decreases in hospitals capable of moderate-and high-intensity services. Fewer hospitals provide a range of pediatric services to children, limiting national capacity to provide pediatric hospital care across a range of conditions and complexity.

What’s Known on This Subject

Pediatric inpatient units have closed, resulting in concentration of services in fewer US hospitals over time, but it is unclear the extent to which specific services have regionalized.

What This Study Adds

Among more than 3,000 US hospitals, 26.7% provided minimal pediatric services in 2003, which increased to 65.0% in 2022, the result of regionalization of a range of specific pediatric services in fewer hospitals.

Abbreviations:

CI

confidence interval

ICD

International Classification of Diseases

KID

Kids’ Inpatient Database