Inpatient Use and Outcomes at Children’s Hospitals During the Early COVID-19 Pandemic

This multicenter, cross-sectional study of hospitalized children (inpatient and observation status stays) used the Pediatric Health Information System (PHIS). PHIS is an administrative database of 49 tertiary care pediatric hospitals affiliated with the Children’s Hospital Association (Lenexa, KS). The PHIS database includes patient-level data, including demographic data, diagnosis and procedural codes, and resource use (eg, laboratory and diagnostic imaging). Encryption of patient identifiers allows for tracking of individual patients across multiple hospital visits. Four hospitals were excluded for data quality issues, including failure to report billing data through the end of the study period and for length of stay (LOS) data not reported in hours; the remaining 45 hospitals were included.

In the overall cohort, we included children 0 to 18 years of age with an index hospitalization (medical or surgical) at a PHIS-participating hospital. We excluded birth hospitalizations and maternal obstetrical hospitalizations because most of these stays are not related to acute illness, and children’s hospitals have variable admission practices for these patients. We excluded transfers in and out of children’s hospitals, those who left against medical advice, and hospitalizations with data quality concerns because of the inability to capture use data for the entire episode of care and readmission period. We excluded children with a complex chronic condition to identify a generally healthy cohort of children who would be of similar risk of hospitalization and readmission.¹⁶  Finally, because we were examining hospitalizations and readmissions across different years, we included only the first admission and readmission (if one occurred) per patient within a given index year and excluded subsequent readmissions within the same year as the index admission year.

The COVID-19 period was defined as index hospitalizations and discharges occurring between March 15, 2020, and August 29, 2020. Notably, the beginning of the COVID-19 period corresponded approximately to the US declaration of a national emergency, which occurred on March 13, 2020. Index admissions that occurred after August 29th were excluded from analyses to allow a full 30-day readmission window. To define the pre-COVID-19 period, we examined hospitalizations during the same time frame within the preceding 3 years (ie, March 15 to August 29 of 2017, 2018, and 2019). To account for fluctuations in inpatient admissions and severity of diseases across seasons and years, we describe the pre-COVID-19 period using the median number of inpatient admissions for the preceding 3 years of hospitalizations. Because we examined the same dates over multiple years, we normalized all dates using 7-day increments (ie, day of the week was not considered in describing weeks).

In addition to describing overall hospitalizations, we describe hospitalizations for 15 selected medical All Patient Refined Diagnosis Related Groups (APR-DRGs) (version 36; 3M Corporation, St Paul, MN). We defined hospitalizations within remaining APR-DRG categories as “all other.” APR-DRGs are a patient classification scheme that groups patients on the basis of the principal diagnosis for medical patients and incorporates severity of illness and risk of mortality by using demographics and comorbidities.¹⁷  Individual APR-DRG categories were chosen on the basis of prevalence within the top 80% of discharges annually and/or frequency of admissions across a majority of hospitals during the study period (eg, APR-DRGs with admissions in >90% of included hospitals, Supplemental Table 4). Within the COVID-19 period, we also describe the number of children hospitalized with COVID-19.

Our primary outcome was the number of inpatient admissions. In addition to examining numbers of admissions, we identified a group of clinical and financial outcomes. The clinical outcomes included LOS measured in hours, 14- and 30-day readmission rates within the same major diagnostic category, ICU use, mortality, and mechanical ventilation. We examined LOS for the index admission and the total episode of care (index and readmission) and describe the days between index admissions and readmissions. Both initial ICU use (ie, admission to the ICU on day 0) and any ICU use were examined. The financial outcomes included standardized costs in US dollars for the index hospitalization and the total episode of care (index hospitalization and any readmission). Standardized costs remove the high interhospital variation in item costs and thus allow for a meaningful comparison of resource use and costs across hospitals. We used previously described methods for calculating standardized costs using billing data.¹⁷  In PHIS, standardized unit costs are calculated yearly, and, for this analysis, standardized costs from previous years were inflated to 2020 dollars by using the medical component of the consumer price index.

We examined demographic characteristics, including age, sex, race and ethnicity, primary payer, and median household income quartile. Race and ethnicity was included as a covariate within analyses because authors of multiple studies have reported the differential impacts of the COVID-19 pandemic based on race and ethnicity.^18–21  Race and ethnicity data are collected at the patient-level within PHIS. We examined patient characteristics, including the number of noncomplex chronic conditions and the mean Hospitalization Resource Intensity Score for Kids (H-RISK).²²  Noncomplex chronic conditions were identified by using the Healthcare Cost and Utilization Project chronic condition indicator, and patients were classified on the basis of the number of chronic conditions.²³  The chronic condition indicator describes conditions that last ≥12 months and result in the need for ongoing health care services. H-RISK was developed to quantify severity of illness among hospitalized children and assigns relative weights to each APR-DRG and severity of illness level, facilitating comparison across APR-DRG groups. We also examined hospitalization characteristics, including hospital region and source of admission (through the individual hospital’s ED versus direct admission).

Descriptive statistics described demographic, patient, and hospital characteristics. A χ² test compared differences in categorical characteristics and unadjusted outcomes between pre-COVID-19 and COVID-19 periods. A Wilcoxon rank test compared nonnormally distributed continuous variables between the COVID-19 and pre-COVID-19 eras, and an independent Student’s t test compared normally distributed continuous variables. Generalized linear mixed models (GLMMs) were used to examine the association of COVID-19 period and outcomes, adjusting for age, race and ethnicity, payer, median household income, presence of a chronic condition, admission through the ED, census region, and severity. GLMMs for readmission assumed an underlying binomial distribution and a logit link; GLMMs for LOS and cost assumed underlying log-linear distribution. All GLMMs included a random hospital effect to account for clustering. In a subanalysis, we divided the COVID-19 period into early (March 15 to May 31, 2020) and extended (June 1 to August 29, 2020) periods to assess for differences in outcomes over time. All statistical analyses were performed by using SAS 9.4 (SAS Institute, Inc, Cary, NC), and P values <.001 were considered statistically significant. The Office of Research Integrity at Children’s Mercy Kansas City (Kansas City, MO) deemed this study using deidentified data exempt from institutional board review.

Our study included a total of 516 388 hospitalizations for 501 459 children, of which 424 856 hospitalizations occurred in the pre-COVID-19 period (mean: 141 619 hospitalizations per year) and 91 532 hospitalizations in the COVID-9 period (Table 1, Supplemental Fig 2, and Supplemental Table 5 present pre-COVID-19 characteristics by year). During the COVID-19 period, we observed declines in the percentage of all hospitalizations classified as medical hospitalizations (pre-COVID-19: 76.2%; COVID-19: 72.4%). Overall, demographic and patient characteristics were similar between periods, with some small differences observed. For example, we observed a decreased proportion of hospitalizations during the COVID-19 period for children 1 to 9 years of age (Table 1). Small differences were also observed for patient characteristics, such as number of chronic conditions, with higher rates of hospitalizations observed during the COVID-19 period for children with no chronic conditions and for those with ≥2 chronic conditions. The mean H-RISK was greater among children presenting in the COVID-19 period (0.734 vs 0.642, P < .001). A total of 1062 children (1.2% of hospitalizations in the COVID-19 period) had COVID-19, with the majority of hospitalizations mapped to APR-DRG 137 (major respiratory infections). The 15 specific medical APR-DRG categories examined accounted for 37.4% of all hospitalizations (2017 N = 54 231; 2018 N = 54 307; 2019 N = 57 818; 2020 N = 26 585).

Compared with the median number of inpatient admissions in the pre-COVID-19 period, the overall admissions were reduced 35.1% during the COVID-19 period. Admissions began to decline after the 10th week of the year, corresponding to March 11 to 17 (Fig 1). The most substantial reductions in overall admissions were observed at approximately the 15th week of the year (April 8 to 14). By the end of the study period, overall admissions remained below pre-COVID-19 numbers.

Inpatient admissions varied across the 15 specific APR-DRGs, with declines in hospital admissions during COVID-19 ranging from 8.5% to 81.3% compared with median admissions in the pre-COVID-19 period. The largest reductions were observed for asthma (81.3% reduction), bronchiolitis (80.1% reduction), pneumonia (71.4% reduction), and upper respiratory tract infections (URTIs) (68.9% reduction). The smallest admission reductions were observed for diabetes (8.5% reduction) and urinary tract infections (14.2% reduction).

In examinations of trends over time, we observed that changes in inpatient admissions varied by APR-DRG (Fig 1). Diagnoses such as diabetes and seizure experienced substantial initial declines with return toward pre-COVID-19 numbers of admissions between weeks 18 and 24, whereas diagnoses such as asthma, bronchiolitis, pneumonia, URTI, and nonbacterial gastroenteritis experienced sustained declines through the end of the study period.

Overall adjusted LOS (index and total) and costs were similar during the COVID-19 period compared with the pre-COVID-19 period (Table 2). ICU use, mortality, and mechanical ventilation also remained similar between time periods, although adjusted readmission rates were lower during the COVID-19 period at both 14 and 30 days. Unadjusted outcomes are presented within Supplemental Table 6. In a subanalysis of the early versus extended COVID-19 period, LOS and ICU use were slightly higher during the early COVID-19 period (Supplemental Table 7).

Index (adjusted ratio of means [aRM] 0.86–0.88, all P < .0001) and total episode of care LOS (aRM 0.85–0.89, all P < .0001) were shorter during the COVID-19 period for bronchiolitis, asthma, pulmonary edema and respiratory failure, and other gastrointestinal diagnoses, although absolute differences in LOS were small (Table 3). URTI was the only condition associated with longer index (aRM 1.12 [99.9% confidence interval (CI) 1.04–1.20]) and total episode of care LOS (aRM 1.10 [99.9% CI 1.02–1.19]) during the COVID-19 period. ICU use was lower during COVID-19 for bronchiolitis. Readmissions were similar during pre-COVID-19 and COVID-19 periods for all APR-DRG categories.

Costs of care were lower during the COVID-19 period for 3 specific APR-DRGs, including bronchiolitis (aRM 0.86 [99.9% CI 0.80–0.94]), pulmonary edema and respiratory failure (aRM 0.90 [99.9% CI 0.83–0.98]), and asthma (aRM 0.92 [99.9% CI 0.87–0.97]). Costs were higher for infections of the upper respiratory tract (aRM 1.13 [99.9% CI 1.05–1.22]) and nonbacterial gastroenteritis (aRM 1.07 [99.9% CI 1.01–1.14]).

In this multicenter study of pediatric hospitalizations, we observed considerable reductions in use during the COVID-19 pandemic, as seen by overall inpatient admissions declining nearly 40%. Despite reductions in overall admissions, the overall clinical and financial outcomes were similar between COVID-19 and pre-COVID-19 periods. In assessments of use by APR-DRG, we observed considerable variation in inpatient admissions by diagnosis between the two time periods, with notable declines for the common pediatric conditions of asthma, bronchiolitis, and pneumonia. Many of the reductions to inpatient admissions were sustained through the end of the study period.

Similar to global reports of decreased health care use during the early pandemic, we observed substantial declines in pediatric inpatient admissions among generally healthy children compared with previous years, especially in the early pandemic period. These data align with a study of hospitalization and ICU admissions from 5 pediatric hospitals around the world, in which researchers noted a 32% to 49% decrease in hospitalization rates and a 12% to 38% decrease in ICU admission (with the exception of a hospital in Paris, at which admissions increased by 2%) during the first 6 weeks of the pandemic.¹²  In a study of ED resource use within the United States, Hartnett et al¹¹  reported substantial declines in ED visits, with the greatest reductions observed for children ≤14 years of age in age-stratified analyses. Similarly, Wilder et al²⁴  describe declines in overall hospitalizations per week at a single tertiary care pediatric hospital. The substantial reductions that we observed likely reflect a complex interplay of factors, including, among others, the influence of fear on health-seeking behavior, adherence to social distancing policies, knowledge about when to seek health care, and the impact of social distancing and school closures on transmission of respiratory infections.

The greatest declines in inpatient admissions we observed were for asthma, bronchiolitis, pneumonia, and URTIs. Although reductions in hospitalizations for some specific APR-DRGs, such as asthma, may be expected with the provision of high-quality ambulatory pediatric care (eg, ambulatory care sensitive conditions), we observed declines greater than expected across many APR-DRG categories, and the degree of impact was not consistently observed across other ambulatory care sensitive conditions (eg, diabetes stays were reduced only 8.5% during the COVID-19 period). Our observation of declines in inpatient care for many infections and infection-mediated diseases (eg, asthma) are consistent with other studies in which researchers examined ED and inpatient resource use during COVID-19, including asthma, croup, bronchiolitis, and pneumonia.^24–26  Notably, the specific conditions with the greatest declines in our study are all conditions that may be affected by changes in community viral and bacterial transmission. Thus, social distancing and policies including school closures may have had a cumulative impact on declines in admissions in addition to the factors described above.

With overall patient and caregiver avoidance of health care facilities during the pandemic, one might expect a delay in presentation, leading to patients who are potentially sicker being admitted and worse patient outcomes (eg, increased ICU use). Previous studies of the COVID-19 pandemic in children have reported delayed presentations of pediatric solid tumors, diagnosis of new onset diabetes leading to increased incidence of diabetic ketoacidosis, and increased complications related to appendicitis.^13–15  Although the mean H-RISK (used as a proxy for severity of illness) was higher during the COVID-19 period in our study, we did not observe many differences in overall clinical outcomes (eg, ICU use) among previously healthy children when looking across the entire COVID-19 period compared with the pre-COVID-19 period, with the exception of fewer readmissions. Notably, ICU use was higher during the early compared with extended COVID-19 period, although other outcomes including LOS, readmission rates, mortality, and mechanical ventilation use remained similar during this early period. Overall reductions in inpatient admissions may have contributed to more focused care and improved hospital flow and/or reduced readmissions for admitted children; however, other factors including staffing changes could have also contributed, and future studies exploring these factors may shed further light on our observations.

Although differences in overall outcomes were small, when we examine outcomes by APR-DRG some important differences did emerge. Whereas several infection and infection-related conditions experienced reductions in inpatient use (eg, declines in LOS and costs for respiratory syncytial virus bronchiolitis and asthma), children with URTIs experienced longer LOS. Focused investigations of factors that contributed to differential outcomes for these subsets of infection and infection-mediated diseases may identify opportunities to mobilize ambulatory resources in future pandemics.

Our findings should be interpreted within the context of several limitations. First, in this study we used administrative data and errors in billing and coding could contribute to misclassification bias. Because PHIS is rigorously monitored for any data quality issues, the impact of misclassification bias is expected to be small. Differences in the prevalence of COVID-19 within individual communities could have also impacted some of the variability we observed. Previous reports describe the impact of the COVID-19 pandemic on socioeconomic factors, including an increased proportion of publicly insured individuals. Finally, within our models, we adjusted for sociodemographic and clinical factors, including payer, which may have adjusted away some of the indirect effects of the pandemic by biasing our results toward the null.

This study of US children’s hospitals revealed overall reductions in pediatric inpatient admissions, with substantial reductions observed for many infections and infectious-mediated processes (eg, asthma). Although we observed small reductions in overall LOSs, ICU use, and readmission rates during the early COVID-19 pandemic compared with the same time frame in previous years, the impact by diagnosis was variable.

APR-DRG

All Patient Refined Diagnosis Related Group

aRM

adjusted ratio of means

CI

confidence interval

COVID-19

coronavirus disease 2019

ED

emergency department

GLMM

generalized linear mixed model

H-RISK

Hospitalization Resource Intensity Score for Kids

LOS

length of stay

PHIS

Pediatric Health Information System

URTI

upper respiratory tract infection