Healthcare Use in the Year Following Bronchiolitis Hospitalization

This retrospective multicenter observational cohort study was approved by the local institutional review board. To determine a multicenter 1-year readmission proportion among patients admitted with bronchiolitis, patients were included if they (1) had a first inpatient hospital admission with a diagnosis of bronchiolitis, (2) at an age of less than 24 months old, (3) between January 1, 2010 and December 31, 2019. To analyze outpatient healthcare utilization following bronchiolitis hospitalization, a nested single-center analysis was performed. The above cohort was filtered to include patients who additionally received ≥1 primary care visit within 0 to 12 months of hospitalization and ≥1 primary care visit ≥12 months after hospitalization within our healthcare system to eliminate loss to follow up. There were no exclusion criteria.

The cohort was identified by querying the Pediatric Health Information Systems (PHIS) database for inpatient encounters with International Classification of Diseases (ICD) diagnostic codes for bronchiolitis (ICD-9 466.1 or ICD-10 J21). PHIS is a quality-controlled administrative data warehouse of children’s hospitals maintained by the Children’s Hospital Association.¹⁶  Readmissions were identified by querying PHIS. Institutional records were linked to the PHIS encounters using a decryption key. Inpatient clinical variables were extracted from an enterprise data warehouse (EDW; Oracle, Austin, Texas) synchronized with the inpatient electronic health record (Cerner, North Kansas City, Missouri). Outpatient encounters for health system primary care clinics, specialty clinics, urgent care facilities, and emergency departments included data from the institution’s outpatient electronic health record (Epic, Verona, Wisconsin). Records were linked using the enterprise master patient index. Zip-code median income was extracted from PHIS and neighborhood area deprivation index (ADI) was extracted from the EDW. Outpatient encounters and readmissions were obtained for January 1, 2010 to December 31, 2020 and filtered to within 1 year of the index bronchiolitis hospitalization.

Vital signs, laboratory values, pharmacy prescriptions, and oxygen therapy underwent validation assessment before analysis. Demographics were described with summary statistics. For the multicenter cohort, mechanical ventilation was determined by ICD codes and includes both invasive and noninvasive ventilation. For the nested single-center cohort, mechanical ventilation was more precisely divided into invasive and noninvasive ventilation based on charting by respiratory therapists. Encounters were described using summary statistics. Monthly outpatient encounters following discharge were plotted. Patients were stratified according to level of respiratory support, and monthly healthcare utilization was compared using 1-way Analysis of Variance (ANOVA) or Kruskal-Wallis rank-sum test. Because there was an increase in outpatient healthcare use during the first 4 months following bronchiolitis hospitalization compared with previously published population norms,¹⁷  patients were stratified into quartiles based on the number of outpatient healthcare visits in the first 4 months following discharge and compared using Kruskal-Wallis rank-sum test, χ² test, or Fischer exact test. Readmission proportions were compared using χ² or Fischer exact test. All statistical analyses were performed using RStudio (RStudio, Boston, MA) and R version 4.1.3 (R Foundation for Statistical Computing, Vienna, Austria).

To clarify whether outpatient healthcare utilization was related to bronchiolitis hospitalization or other underlying illness, we conducted a sensitivity analysis excluding patients with complex chronic conditions at the time of their index hospitalization.¹⁸ 

There were 308 306 admissions for bronchiolitis among 271 115 patients across 47 hospitals in PHIS between 2010 to 2019. Demographics are shown in Supplemental Table 2. Briefly, the median (interquartile range [IQR]) age at admission was 5 (2 to 11) months. 18.5% (50 049 of 271 115) had a complex chronic condition. Specifically, 4.1% (11 200 of 271 115) were premature or had a neonatal condition and 3.3% (8814 of 271 115) had a respiratory condition. 27.3% (74 001 of 271 115) were admitted to an ICU and 10.8% (29 378 of 271 115) received mechanical ventilation. Survival to discharge was 99.8% (270 611 of 271 115).

17.8% (48 332 of 271 115) of patients were readmitted within 1 year after discharge a median (IQR) of 1 (1–2) times, for a total of 85 893 readmissions. The median (IQR) time to readmission was 1.9 (0.7–4.8) months. 6.0% (16 167 of 271 115) of patients were readmitted within 30 days after discharge. 9.9% (26 712 of 271 115) of patients were readmitted for bronchiolitis, with the remainder being readmitted for other diagnoses. The median (IQR) hospital-level 1-year readmission proportion was 18.3% (16.1% to 19.6%; Fig 1A). Those that received ICU care during their index bronchiolitis admission included 27.3% (74001 of 271115) of patients, and 10.8% (29378 of 271115) received mechanical ventilation. The 1-year readmission proportion was 22.9% (16 919 of 74001) among patients admitted to an ICU and 26.8% (7865 of 29378) among patients undergoing mechanical ventilation, respectively (Fig 1B).

There were 5326 admissions for bronchiolitis among 4791 patients at our center during the study period. Of those, 1438 of 4791 (30.0%) patients had primary care in our system and were included (Supplemental Fig 3). Demographics are listed in Supplemental Table 2. Briefly, the single-center cohort was younger (median [IQR] age 4 [1–9] versus 5 [2–11] months, P < .001) than the multicenter cohort. The single-center cohort also had a higher proportion of white patients (69.1% [993 of 1438] versus 55.1% [149 378 of 271 115], P < .001), complex chronic conditions (20.7% [297 of 1438] versus 18.5% [50 049 of 271 115], P = .033), and mechanical ventilation (15.4% [222 of 1438] versus 10.8% [29 378 of 271 115], P < .001) than the multicenter cohort.

Outpatient healthcare use following hospitalization for bronchiolitis is shown in Fig 2. There were a median (IQR) of 9 (6–13) total outpatient physician or advanced practice provider visits per patient within 12 months after discharge. Utilization was higher across all visit types for the first 4 months after discharge compared with the remainder of the study period (Fig 2). Outpatient healthcare utilization was right tailed, with a small number of patients utilizing a large volume of outpatient care (Supplemental Fig 4). Severity of illness was correlated with increased outpatient healthcare utilization for approximately 4 months following hospital discharge (Fig 2). Patients in the highest quartile of healthcare utilization had more usage across all visit categories, but most markedly in sick visits to primary care offices and specialist visits (Table 1). The median (IQR) age at admission was 5 (2–10) months in patients in the lowest quartile and 5 (2–9) in patients in the highest quartile. Patients in the highest quartile were more often male, white, and from areas with higher household income and lower ADI (Table 1). They were more likely to have complex chronic conditions, a history of prematurity, and technology dependence (Table 1). There was no significant difference in presenting respiratory rate, heart rate, hemoglobin oxygen saturation, or laboratory values (eg, pH, Pco₂, electrolyte values, and inflammatory markers) at index admission between the quartiles of healthcare utilization (data not shown).

There were 567 readmissions within 1 year among 293 of 1438 (20.4%) of patients in the single-center analysis. The readmission proportion in the single-center analysis was significantly higher than in the multicenter analysis (20.4% [293 of 1438] versus 17.8% [48 332 of 271 115], P = .013). Higher quartiles of outpatient healthcare utilization within 4 months following bronchiolitis hospitalization were associated with all-cause readmissions, as well as bronchiolitis readmissions (Table 1).

The sensitivity analysis excluding patients with complex chronic conditions (CCCs) included 1141 of 1438 (79.3%) of the single-center cohort. Demographics are listed in Supplemental Table 3. Patients without CCCs had a lower 1-year readmission proportion compared to all patients in both the multicenter analysis (14.8% [169 of 1141] versus 17.8% [48332 of 271115], P = 0.008) and the single-center analysis (14.8% [169 of 1141] versus 20.4% [293 of 1438], P < 0.001) (Supplemental Table 5). Patients without CCCs used fewer subspecialty visits after discharge, resulting in lower overall outpatient healthcare utilization (Supplemental Table 4). Similar to the single-center analysis, higher quartiles of outpatient healthcare utilization within 4 months following bronchiolitis hospitalization were associated with all-cause readmissions, as well as bronchiolitis readmissions (Supplemental Table 4).

This medical record and administrative database linkage study is the first to analyze both inpatient and outpatient healthcare utilization in the year following hospitalization for bronchiolitis. We report 2 major findings. First, readmissions are common in the first year after bronchiolitis, occurring in 17.8% of patients in this 47-hospital decade-long cohort. Second, outpatient healthcare use is increased for 4 months following bronchiolitis hospitalization in a large, integrated delivery and finance healthcare system in the United States.

Most prior studies have examined 30-day readmissions in patients with bronchiolitis.^12–14  The 30-day readmission proportion of 6.0% in the current study is in agreement with previous work.^12–14  However, the 1-year readmission proportion of 17.8% suggests that previous studies likely do not capture the long-term outcomes of these infants. Indeed, readmission rates rose relatively sharply for ∼ 4 months following hospitalization before reaching a constant level (Fig 1B). This is the same period during which we observed an increase in outpatient healthcare utilization in our single-center cohort (Fig 2). In both cases, the effect was most pronounced in more ill children; those requiring mechanical ventilation had higher outpatient healthcare use (Fig 2) and readmissions (Fig 1B).

During the first 4 months following hospitalization, patients with bronchiolitis used approximately 1.5 times as many outpatient visits as would be expected based on reports of primary care use among commercially-insured well children under 2 in the United States (Fig 2).¹⁷  Outpatient healthcare utilization was right-tailed, with a small number of patients incurring a large number of visits (Supplemental Fig 4), and frequent outpatient visits were associated with readmissions for bronchiolitis as well as all-cause readmissions (Table 1). Similar trends were seen in a sensitivity analysis excluding children with pre-existing complex chronic conditions (Supplemental Table 4 and 5), suggesting that there may be a subpopulation of children who have prolonged vulnerability to subsequent respiratory infections after discharge. Although prior studies have shown long-term differences in pulmonary function testing and more severe routine childhood illnesses in children with severe bronchiolitis,^8,9,11,19,20  it remains unclear whether this association is casual or mirrors underlying genetic predisposition. Prospective study is needed to determine whether patients with severe bronchiolitis may benefit from early evaluation by a pulmonologist or intermittent corticosteroids during acute illnesses to reduce severity.^21,22 

The association between outpatient visits and hospital readmissions has important policy ramifications. The AAP has specific interest in predicting the course of patients with bronchiolitis¹  and there is national interest in identifying “superusers” of healthcare as a method of cost-containment.^23,24  A recent large, randomized, controlled trial highlighted that adult patients who demonstrate increased healthcare utilization for a period surrounding hospitalization commonly regress to a mean level of utilization.²⁵  However, in our population, increased outpatient utilization was sustained for 4 months following discharge and associated with readmissions (Fig 2 and Table 1). This association is important, as previous studies have been unable to identify reliable predictors of readmission in bronchiolitis.^12,13 

In our study, patients in the highest quartile of healthcare utilization had generally more-severe illness (as assessed by higher rates of ICU admission and invasive mechanical ventilation). However, predicting “high-users” of healthcare at the time of index admission for bronchiolitis remains challenging. As shown in Table 1, only 30.3% (121 of 400) of patients admitted to an ICU and 40.2% (47 of 117) of patients undergoing invasive mechanical ventilation for bronchiolitis were in the highest quartile of healthcare utilization after discharge. Thus, follow up programs for every intubated patient with bronchiolitis will likely be low yield. The harmonization of inpatient and outpatient medical records to predict long-term readmissions is an important goal of learning health systems.

Outpatient healthcare utilization is an amalgamation of disease severity, care-seeking thresholds, and barriers to care. We note concerning disparities in care locations between racial groups and socioeconomic strata. Patients with high healthcare utilization were more likely to be white, have commercial insurance, and live in more-affluent neighborhoods with lower area deprivation indices (ADI; Table 1). Race and ADI are both known to be associated with access to primary care.^26,27  Because higher ADI is known to be associated with worse outcomes in other pediatric respiratory conditions,^28,29  higher healthcare utilization among wealthier patients in the current study likely reflects ongoing disparities in pediatric care.

This study highlights the potential of harmonizing multiple health system data sources to understand patient trajectories. Many pediatric conditions, such as bronchiolitis, have very low inpatient mortality. As mortality continues to decline, there is increased interest in examining long-term outcomes. The interaction between patients and the healthcare system is complex, with patients often changing context. For example, a patient is seen in urgent care, admitted to the hospital, and then has a follow up appointment in a primary care office after discharge. Prominent inpatient databases often do not contain outcome data after discharge. We leveraged the readily available search tools of a quality-controlled national pediatric database¹⁶  to locate patients of interest and obtain demographic information, then used the institutional data warehouse to obtain detailed clinical information from an inpatient encounter, and a system-wide EDW to obtain information about follow up at dozens of different outpatient locations. Crossplatform linkages, as demonstrated by our study, offer the potential to understand disease arcs and complex long-term associations in exciting new ways. Mapping of multiple hospital systems to common data ontologies and standards, such as the Observational Medical Outcomes Partnership (OMOP) and Fast Healthcare Interoperability Resources, offers the promise of rapidly scaling such projects to analyze outcomes across multiple institutions.^30,31 

Our study has important limitations. First, patients may move out of the service area of PHIS hospitals during the follow up period, resulting in an underestimate of the 1-year readmission proportion in the multicenter analysis. By contrast, patients in the single-center analysis could be filtered for ongoing primary care after 12 months, ensuring that they were not lost to follow up. However, this population represents patients who deliberately receive care within a quaternary children’s regional healthcare system, which may overestimate healthcare utilization relative to the population served by private practice clinics. Second, our definition of receiving healthcare within the system would exclude some patients with low healthcare utilization (less than 1 primary care physician [PCP] visit per year). Given that ∼ 5% of all children in the United States lack health insurance³²  and likely consume very little healthcare, this may lead to a nontrivial overestimation of outpatient visits in the current study. Third, previously published outpatient visits were among commercially insured children,¹⁷  which made up only 26.1% of the single-center cohort in the current study. Because commercial insurance was associated with higher healthcare utilization than government insurance, the current study may underestimate the true increase in outpatient healthcare use after bronchiolitis hospitalization. Fourth, it remains unclear whether sociodemographic factors, such as race and neighborhood ADI, truly influence biologic disease severity (for example, through airborne pollutant exposure), or reflect disparities in access to healthcare.

In conclusion, readmissions are common in the year following bronchiolitis hospitalization, and outpatient healthcare use is increased for approximately 4 months following discharge. Outpatient healthcare utilization is related to severity of illness during the index admission and associated with socioeconomic disparities. Prospective study is needed to track long-term outcomes of infants with bronchiolitis.