Association of Rural Residence With Surgical Outcomes Among Infants at US Children’s Hospitals

Infants (aged <1 year) who received surgical care for major congenital anomalies at children’s hospitals in the United States were evaluated using the Pediatric Health Information System (PHIS) from January 1, 2016, to June 30, 2019. The PHIS is an administrative database maintained by the Children’s Hospital Association (Lenexa, Kansas) that includes clinical and resource utilization data from >49 children’s hospitals in the United States.¹⁶  Specifically, infants who received surgical care for diagnoses of esophageal atresia/tracheoesophageal fistula (EA/TEF), gastroschisis (GS), Hirschsprung’s disease (HD), anorectal malformation (ARM), and congenital diaphragmatic hernia (CDH) were included. These conditions were selected because they represent surgical conditions that require both complex multidisciplinary care and continued close follow-up in the postoperative period.

The study population was identified through the use of relevant International Classification of Diseases, 10th Revision, diagnosis and procedure codes (Supplemental Table 5). Infants with an index hospital admission between January 1, 2016, and June 30, 2018, were included, such that all patients could have a complete year of postoperative follow-up for evaluation. Infants were included only if they had both a diagnosis code and a corresponding procedure code for surgical repair. For infants with EA, GS, and CDH to be included, a relevant procedure had to be performed within the first 30 days of life, because infants with these conditions generally undergo repair within the first several weeks, if not days, of life. For infants with ARM and HD, a relevant procedure had to be performed within the first 3 months of life to be included, given that infants with these conditions may initially be managed with dilations and irrigations, respectively, before undergoing delayed repair. In addition, for infants with ARM and HD whose initial procedure was an ostomy creation, definitive repair was required to have occurred within the first year of life. Children who had >1 of the surgical diagnoses were excluded.

Demographic and clinical characteristics including age, sex, race, insurance type, gestational age, and birth weight were evaluated. Socioeconomic status was examined using the median household income on the basis of home residence zip code. The Hospitalization Resource Intensity Scores for Kids (H-RISK) score was used as a measure of comorbidity and severity of illness.¹⁷  Children that underwent a cardiac procedure were identified using relevant cardiac All Patient Refined–Diagnosis Related Group codes.¹⁸  The utilization of Extracorporeal Membrane Oxygenation (ECMO) during the initial hospital encounter of infants with CDH was also noted using relevant ECMO procedure codes.

Rural versus urban residence was defined through rural-urban commuting area codes, which classify census tracts according to measures of population density, urbanization, and daily commuting.¹⁹  The original 4-tier rural-urban commuting area classification system (large urban area, small urban area, large rural town, small rural town/isolated rural area) was used to define urban and rural populations.²⁰  Rurality was defined as residence in a large rural town or small rural town/isolated rural area in the primary analyses. Secondary analyses examined rurality further by comparing:

1. residence in a small rural town/isolated rural area versus residence in any other areas; and

2. residence in a small rural town/isolated rural area versus residence in a large urban area.

Distance between home residence and hospital was calculated as the linear distance in miles between the centroids of zip codes of the home residence and the hospital. Hospital mortality and total hospital days were examined for both the index hospitalization and total 1-year hospitalizations at the index hospital where surgical care was received. Total hospital readmissions were examined at the 30-day and 1-year time points after surgery at the index hospital where surgical care was received.

Unadjusted comparisons of patient characteristics were performed using χ² tests of independence for categorical variables and t tests or Wilcoxon rank-sum tests for continuous variables. Generalized linear mixed effects models with appropriate distributions were used to evaluate for associations of various demographic, clinical, and geographic factors with outcomes. Hospital days were log-transformed before modeling. A random intercept term was used to adjust for center-level differences such as variability related to case volumes. Variance inflation factors were estimated to ensure the absence of collinearity between covariates. A stepwise forward selection process was used to select covariates for model inclusion, and alternate models were compared using log likelihoods. Models were adjusted for race, insurance status, cardiac procedures, H-RISK, birth weight, household income, travel distance, ECMO use (for CDH patients), and ostomy timing in relation to definitive repair (for ARM and HD patients).

To evaluate for possible effect modification, interaction terms between rurality and travel distance and between rurality and household income were introduced into the models, after which stratified analyses were explored as indicated. To control the error rate within each congenital condition cohort, Bonferroni correction was used. All statistical tests were 2-sided, and the statistical significance level was considered P < .05. Given multiple comparisons, the statistical significance level for comparisons of rural and urban infants based on distance and median household income strata was considered P < .004 and P < .003, respectively. Analysis was performed using SAS v.9.4 (SAS Institute, Cary, NC).

In total, 5732 patients underwent surgical repair of EA/TEF, GS, HD, ARM, and CDH, and had 1 year of postoperative follow-up (Table 1). Overall, 20.2% lived in rural areas. Rural and urban infants were similar with regard to sex, age at admission, and distribution of surgical diagnoses. Rural infants were significantly more likely than urban infants to be non-Hispanic white (69.7% vs 49.3%, P < .001) and have public or another nonprivate type of insurance (64.3% vs 59.9%, P < .001).

Rural infants lived farther from children’s hospitals and more frequently lived in areas with lower median household income (both P < .001). In fact, the majority of rural infants lived >60 miles away (65.7% vs 18.4% of urban infants) and in areas with median household income <$35 000 (64.9% vs 26.4% of urban infants). Demographic differences between rural and urban infants were similar when stratified by diagnosis (Supplemental Table 6). Urban infants with EA/TEF were slightly younger at the time of surgery than their rural counterparts; for other diagnoses, surgical repair occurred at a similar age between rural and urban infants.

Rural and urban infants had similar levels of comorbidity and severity as marked by H-RISK scores for all congenital conditions (Table 2). There were no significant differences between rural and urban infants regarding the need for a cardiac procedure or the level of prematurity by birth weight. Urban infants were more likely to have received an ostomy before their ARM repair (47.5% vs 37.3%, P = .02), but there was no such difference among infants with HD. For CDH, there was no significant difference between rural and urban infants in ECMO utilization, either pre- or post-repair.

In unadjusted analysis, rural infants with ARM had higher hospital mortality during both the index hospitalization (2.5% vs 0.4%, P = .007) and total 1-year hospitalizations (2.5% vs 0.7%, P = .04) (Table 3). Additionally, rural infants with ARM also had lower total 1-year hospital days (median 19.0 days [interquartile range (IQR) 10.0–36.0] vs 15.0 days [IQR 8.0–25.0], P < .001) and lower hospital readmissions within 1 year (71.6% vs 60.1%, P = .005) compared with urban infants. Rural infants with GS had lower total hospital days during the index hospitalization (median 38.0 days [IQR 26.0–59.0] vs 39.0 days [IQR 28.0–67.0], P = .046) compared with urban infants, but there was no difference in total 1-year hospital days. There were otherwise no significant unadjusted differences between rural and urban infants with HD, EA/TEF, and CDH regarding hospital mortality, hospital days, or hospital readmissions.

Findings were generally similar after multivariable adjustment. There were no significant differences in hospital mortality between rural and urban infants during the index hospitalization or total 1-year hospitalizations. Rural infants with ARM had lower adjusted hospital days over 1 year (incidence rate ratio [IRR] 0.76 [95% confidence interval (CI) 0.63–0.92], P = .004) (Table 4). However, there was no significant difference in total hospital days during the index hospitalization between rural and urban infants with ARM. Similarly, rural infants with GS had lower adjusted hospital days during the index hospitalization (IRR 0.91 [95% CI 0.83–0.98], P = .02) and over 1 year (IRR 0.91 [95% CI 0.84–0.99], P = .03). There were no significant differences in hospital readmissions between rural and urban infants, except among those with EA/TEF. Rural infants with EA/TEF had higher adjusted hospital readmissions within 30 days (odds ratio 1.63 [95% CI 1.04–2.55], P = .03).

Many of the interaction terms between rurality and travel distance and between rurality and household income were found to be statistically significant, and therefore additional stratified analyses were completed. However, inferences were generally unchanged when examining rural infants who would theoretically have the greatest challenges to accessing care, specifically those who lived ≥60 miles from their children’s hospital and those living in zip codes with median household income <$35 000. Specifically, there were no significant differences in adjusted hospital mortality, hospital days, or readmissions among those subgroups of rural infants compared with urban infants.

In secondary analyses further examining rurality, there were no significant differences identified in hospital mortality, hospital days, or hospital readmissions between those with:

1. residence in a small rural town/isolated rural area compared with residence in any other areas; and

2. residence in a small rural town/isolated rural area compared with residence in a large urban area, with the exception of infants with GS.

Rural infants with GS from a small rural town/isolated rural area had higher adjusted hospital days during the index hospitalization (IRR 1.17 [95% CI 1.02–1.35], P = .03) and over 1 year (IRR 1.20 [95% CI 1.04–1.39], P = .02) compared with infants from a large urban area.

This national study examining children with congenital surgical conditions evaluated whether rural infants have worse postoperative outcomes compared with urban infants. Rural infants lived farther from children’s hospitals and in areas with lower median household income, yet outcomes after surgical repair of EA/TEF, GS, HD, ARM, and CDH were largely similar apart from a higher adjusted risk of readmission within 30 days among rural infants with EA/TEF. Rural infants with the lowest household income and living farthest from a major children’s hospital also did not appear to be at higher risk of poor postoperative outcomes. In addition, infants from the most rural areas (those defined as small rural towns and isolated rural areas) had similar outcomes. Despite many known rural health disparities among children, rural infants undergoing surgery at major children’s hospitals appear to have similar outcomes, even with their significantly increased travel burden and limited financial resources. However, this study is limited by the variables and outcomes available in PHIS and therefore may not have detected differences in other clinically meaningful outcomes.

Disparities in health outcomes among rural children are widespread and persistent over time, including a 25% higher mortality risk compared with urban children as found by 1 study using the Centers for Disease Control and Prevention’s Wide-ranging Online Data for Epidemiologic Research database.²¹  One potential mechanism underlying these disparities is differential access to specialist care such as complex surgical care.¹¹  In its recent “Right Child/Right Surgeon” initiative, the American Pediatric Surgical Association underscored the importance of ensuring that all children have access to high-quality pediatric surgical care, including those living in rural and underserved areas.¹  Regionalization of pediatric surgical care, especially for complicated congenital conditions, has become widely accepted despite the longer travel distances that may be required.¹⁰  However, consideration of access to care is critically important, especially for underserved populations who live in rural settings and may experience unique medical, economic, and social disadvantage compared with urban populations.^22–24  Accessibility includes consideration of travel costs, distance, and time, in addition to transportation resources.^25,26  Lack of family resources and unaffordable expenses may make traveling great distances to a regionalized, urban center especially challenging.^12,14 

These challenges could lead to less optimal surgical outcomes among rural children, prompting the current study which evaluated outcomes after repair of congenital anomalies that require continued close follow-up in the postoperative period. Contrary to the hypothesis that rural infants would have worse outcomes, however, this study instead found generally similar outcomes, even among those living in the lowest income and most rural areas and the greatest distances from a children’s hospital. Among rural infants in the main analysis, the only adjusted outcome found to be worse was hospital readmissions within 30 days among those with EA/TEF. Although this difference could represent a clinically meaningful difference in outcomes, it could also reflect a lower threshold for admission for those that live more remotely. Secondary analyses of rurality also showed generally similar outcomes. Only GS infants living in the most rural area had greater hospital days during the index hospitalization and over 1 year compared with infants from a large urban area, which could be because of a higher threshold for discharge for infants living in the most remote areas. However, the overall findings of this study are reassuring that children’s hospitals are providing optimal surgical care to infants with complex congenital conditions living in rural locations.

The minimal differences in outcomes among rural infants in this study may highlight the importance of receiving surgical care for these complex conditions at major children’s hospitals, perhaps negating what would otherwise be disparate outcomes among rural children. Multiple studies have shown improved outcomes when children’s surgery is performed at high-volume, regionalized centers.¹⁴  For example, lower postoperative complications and costs were identified for infants undergoing laparoscopic pyloromyotomy at freestanding children’s hospitals.^14,27  Improved outcomes and decreased rates of bowel resection have also been seen among children who receive care for intussusception at children’s hospitals.⁶  In addition, neonates treated for necrotizing enterocolitis have better outcomes when treated at regional centers.⁸ 

This study’s findings should not be taken to mean that rural disparities in children’s surgical care do not exist. It is important to note that 2 types of disparity in surgical care among rural children could exist: First, differential outcomes among rural children who receive complex surgical care at a recommended center, and second, differential capacity for rural children to receive complex surgical care at an adequately resourced center that will provide optimal outcomes. Although this study may argue against the former, the latter cannot be evaluated with this particular study because the data source for this study includes only children cared for at major children’s hospitals. Those infants not cared for at major children’s hospitals could still be at risk for worse postoperative outcomes. Indeed, a considerable volume of children’s surgical care is not performed at children’s hospitals.²⁸  Rural children are likely to be in this group receiving care outside a major children’s hospital given that they generally live greater distances from children’s hospitals.^23,29  As a result, further work is necessary to evaluate if rural children treated at nonchildren’s hospitals receive optimal surgical care for these complex conditions and experience outcomes similar to those identified in this study. Finally, it is important to note that, although surgical care at children’s hospitals may achieve similar clinical outcomes, the impact of accessing and receiving that care for rural families, in particular those with limited resources, may still be distinct and particularly consequential.

In addition to the inherent selection bias in examining only children treated at major children’s hospitals, this study may also be limited by surveillance bias in the postoperative courses of rural and urban infants. This limitation is highlighted by the lower total hospital days and readmissions that were identified among rural infants with certain diagnoses in this study. If infants presented to local community hospitals rather than returning to their children’s hospitals for postoperative concerns, those encounters would not be captured in the PHIS. Those uncaptured encounters are likely to be higher among rural infants, potentially explaining the higher risk of those outcomes among urban infants. These findings may also highlight rural infants’ limited access to children’s hospitals after discharge, resulting in subsequent care at a local facility rather than at the children’s hospital where surgery was performed, which could also impact outcomes. Furthermore, although there were no differences in mortality at 1 year between rural and urban infants, we recognize that infants could have presented to and subsequently died at a local community hospital rather than at the index children’s hospital where surgical care was received.

This study is limited by the variables that are available in the PHIS. Clinically relevant details regarding pregnancy, prenatal care, and delivery timing could not be evaluated using the PHIS, nor could infants’ location of birth in relation to the location of the hospital providing surgical care. In addition, many clinically relevant outcomes beyond mortality, readmissions, and hospital length of stay could not be evaluated in this study. Although one would expect any major postoperative complications to result in a transfer back to the children’s hospital where surgical care was received, it is still possible that inferior outcomes in other clinically relevant measures among rural infants may have gone undetected in this study. This study may also be limited in its use of the H-RISK score to adjust for varying degrees of severity and comorbidity across a heterogeneous group of conditions. Nevertheless, disease severity is unlikely to vary on the basis of geographic residence, and therefore this limitation is unlikely to have biased the results of this study. Finally, although this study leveraged a large, national database to examine outcomes for these relatively rare congenital surgical conditions, it is still possible that the study was limited in its power to detect differences between rural and urban infants.

Despite longer travel distances and lower financial resources, outcomes for rural infants with congenital surgical conditions were similar to urban infants among those cared for at major children’s hospitals. Although these findings are encouraging, some rural children may not receive care at children’s hospitals, which could potentially lead to differential outcomes. Future studies are needed to investigate potential disparities in other clinically relevant outcomes among rural children and to determine if rural children cared for at nonchildren’s hospitals can achieve optimal surgical outcomes similar to those at major children’s hospitals.