Association of Neighborhood Resources and Race and Ethnicity With Readmissions for Diabetic Ketoacidosis at US Children’s Hospitals

Key Points Question Is neighborhood opportunity differentially associated with health outcomes by race and ethnicity among US children with diabetic ketoacidosis? Findings In this cross-sectional study including 72 726 pediatric encounters for diabetic ketoacidosis, the probability of readmission within 365 days was significantly higher among non-Hispanic Black children compared with Hispanic children and non-Hispanic White children at the same level of opportunity. Within racial and ethnic groups, children with very low opportunity had significantly greater probability of readmission compared with those with very high opportunity. Meaning These findings suggest that despite having similar measures of neighborhood opportunity, non-Hispanic Black children with type 1 diabetes experience disparities in health outcomes compared with children in other racial and ethnic groups.


Introduction
Diabetic ketoacidosis (DKA) is the leading cause of hospitalization, morbidity, and mortality in children with type 1 diabetes (T1D). 1,2Previous studies have identified racial and ethnic disparities among children with T1D, particularly regarding risk of DKA.A study from the Centers for Disease Control and Prevention 3 found that non-Hispanic Black (hereinafter Black) children had a nearly 2-fold increased risk of diabetes-related mortality, which included deaths due to DKA, compared with non-Hispanic White (hereinafter White) children.Further, White children are less likely to require hospitalization for DKA compared with children of other races and ethnicities. 2,4Although socioeconomic factors-including lack of health insurance or public insurance, 2,4-12 lower income or living in areas of poverty, 2,4,9,10,12 and lower parental educational achievement 10 -have been associated with higher risk and severity of DKA, other community attributes may influence health outcomes.
The Child Opportunity Index 2.0 (COI) was developed in 2020 and is a novel instrument that provides a comprehensive assessment of neighborhood attributes that may elucidate structural inequities and provide a more robust and practical assessment of social factors that influence health outcomes. 13Studies have shown that greater neighborhood opportunity is associated with fewer acute care visits 14 and hospitalizations for ambulatory care conditions. 15,16It is unclear whether neighborhood opportunity influences diabetes-related outcomes.To address this gap in knowledge, we evaluated whether COI categorical scores are associated with diabetes-related outcomes by race and ethnicity with regard to (1) readmissions for DKA and (2) complications of DKA, including acute kidney injury (AKI) and cerebral edema (CE).

Study Design, Participants, and Setting
We conducted a cross-sectional study of children and adolescents younger than 21 years who were discharged from inpatient, observation, or emergency department care with a primary diagnosis of T1D with ketoacidosis (International Classification of Diseases, Ninth Revision [ICD-9], codes 250.11 and 250.13

or International Statistical Classification of Diseases and Related Health Problems, Tenth
Revision [ICD-10], codes E10.10 and E10.11) from January 1, 2009, to December 31, 2019.We excluded encounters from 2019 because these were used to determine readmission for 2018, duplicate encounters (ie, 2 encounters with the same admission date), and encounters with missing COI categorical scores or demographic information.This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline and was deemed exempt from review and the need for informed consent by the institutional review board of Childrens' Minnesota.

Data Sources
Clinical data were obtained from the Pediatric Health Information System (PHIS), which is managed by the Children's Hospital Association, Lenexa, Kansas, and includes data from the largest US children's hospitals.Data quality and reliability are assured through a joint effort between the Children's Hospital Association and participating hospitals.Participating hospitals provide encounterlevel data, including demographics, diagnoses, procedures, and measures of resource use.Complex chronic conditions were identified using diagnoses, procedures, measures of resource use, and associated complex chronic conditions. 17Race and ethnicity were determined by site-specific practices at each participating PHIS hospital, including self-report by a child's guardian.We categorized race and ethnicity into 5 groups: Asian, Black, Hispanic, White, and other (including American Indian, multiracial, other race, and missing).

JAMA Network Open | Emergency Medicine
The COI measures neighborhood resources and conditions across more than 72 000 US Census tracts (ie, neighborhoods) from all 50 states and Washington, DC, using data from numerous sources, including the National Center for Education Statistics and the US Department of Education, and is publicly available. 13,18The COI includes

Outcomes
The primary outcomes were readmissions for DKA within 30 days and within 365 days of an index visit resulting in hospitalization.For children with multiple admissions, each hospitalization was considered an index admission regardless of the number of days since the previous DKA admission.
Readmissions were coded as binary (yes or no) variables based on the number of days between the date of discharge and the date of the next admission with a primary diagnosis of DKA.Patients readmitted within 30 days were included among those readmitted within 365 days.Secondary outcomes included the proportion of encounters with codes for AKI (ICD-9: 584.5-584.9 and 586; ICD-10: N17.0-N17.2,N17.8, N17.9, and N19) or CE (ICD-9: 348.5;ICD-10: G93.6) at the index visit.
Acute kidney injury and CE were selected as secondary outcomes given the commonality of AKI [19][20][21][22] and the mortality associated with CE.Race and ethnicity Asian 1.7 (−1.6 to 5.0) 0.9 (−0.9 to 2.7) 0.9 (−0.9 to 2.7) 1.0 (−1.0 to 3.0) 0.5 (−0.5 to 1.For each of the 3 COI domains, Black children had a greater probability of 365-day readmission at all COI categories compared with Hispanic children and White children (Figure 3).Among children  Hospital readmissions are commonly used as a quality indicator, in part because they may be avoidable with appropriate care during the index hospitalization and sufficient ambulatory care after discharge.This concept is the motivation for financial penalties associated with pay-forperformance strategies, such as the Centers for Medicare & Medicaid Services Hospital Readmission Reduction Program 27 for adults and the Potentially Preventable Readmission measure used for pediatric readmissions. 28,29Although the use of readmissions as a quality indicator is controversial, 30,31 the burden to families and health systems is undeniable.It has been estimated that 30-day pediatric readmissions account for $678 million dollars of all annual health care costs. 32For children with T1D, the median charge for a DKA readmission is greater than $12 000, 1 and pediatric intensive care unit charges are even higher. 33Children who are readmitted also experience length of stays more than double those of patients who are not readmitted, 34 making the burden on families even greater as they spend more time away from work and incur personal expenses.[37] Investigation has shown that there is widespread variation in readmissions for DKA across US children's hospitals. 1A 2017 study 38 showed that 365-day readmissions were significantly higher for Black children with T1D (21.7%) compared with children of other races (13.4%).In another study using PHIS data from 2004 to 2012, 7 365-day readmissions for DKA occurred in 28% of children, and factors associated with readmission included public insurance and Black race.Both findings are consistent with our results.Further, the authors of one of these studies 7 showed that Black race was associated with higher odds of 365-day readmission in 59% of hospitals, suggesting variation in DKA admission frequency across hospitals.It is unclear what drives this variation, although a number of factors likely contribute, including implicit biases 39,40 and possibly neighborhood factors.More recently, Maxwell et al 9 conducted a retrospective population-based cohort study using data from Cincinnati Children's Hospital from 2011 to 2017 and examined whether Census tract poverty, race, and insurance status were associated with hospitalization for DKA in children.The authors conducted a subanalysis among children who were readmitted and found that Black race, public insurance status, and greater Census tract poverty were significantly associated with readmission. 9wever, this study was limited in that the number of readmissions was low, with only 42 children (10%) being readmitted during the study period. 9veral studies [14][15][16] have begun to explore the association among neighborhood opportunity, acute care use, and hospitalizations.One study found that residing in an area in the very low COI category was associated with significantly greater odds of having multiple acute care visits for conditions amenable to outpatient care compared with children who live in areas in the very high COI category. 14A similar study found that the hospitalization rate for asthma was significantly higher at 9.1 per 1000 children living in areas in the very low COI category compared with 1.8 per 1000 for those living in areas in the very high COI category. 15Area deprivation, a measure of neighborhood opportunity derived from American Community Survey data, has recently been associated with greater odds of hospital admission for recurrent DKA. 41However, that study was limited to a single state, which reduces generalizability.Only 1 previous study 16 has evaluated use of diabetes resources in association with COI categories.Krager et al 16 16 In the present study, we expand on these findings, using data from the largest children's hospitals merged with COI data.
Our results provide an equity-focused evaluation of disparities as they are associated with diabetes-related outcomes.The COI includes measures such as reading and math proficiency, adult educational attainment, and access to healthy food, all of which are particularly important for children living with T1D because they must calculate insulin dosing, manage diabetes technology (eg, insulin pumps), and find healthy food choices.The importance is further supported by previous work 4][45][46] In addition, household food insecurity has been associated with higher hemoglobin A 1c values and hospitalization rates in children with T1D, 47 which is notable because children with lower COI categories have more limited access to healthy food choices compared with those with higher COI categories. 13,48,49Less frequent use of diabetes technology, particularly continuous glucose monitoring devices, has also been associated with worse health outcomes in children with T1D, including higher rates of DKA 50 and time spent in hyperglycemia. 51In a recent study of 1500 children, Lai et al 52 showed that 54% of White children with T1D started continuous glucose monitoring in the outpatient setting compared with 31% of Black and 33% of Hispanic children.It is unclear why individuals from racial and ethnic minority groups less frequently start continuous glucose monitoring, but it is notable that this disparity persisted even after adjusting for insurance status, 52 suggesting that lack of coverage may not be a driving factor.The COI includes many socioeconomic measures; however, there are other unmeasured factors that may also contribute to disparities.5][56] Previous studies have shown that children with at least 2 missed clinic visits have significantly higher hemoglobin A 1c levels compared with children with 1 or no missed visits, 57,58 and missing endocrinology visits has been shown to be associated with higher odds of hospitalization for DKA. 59With these factors in mind, it is not surprising that children with lower COI levels, and particularly Black children, are at risk for DKA readmissions.
We also found that diabetes-related AKI was associated with race and ethnicity but not COI category.Although studies among adults with T1D have demonstrated racial differences in AKI [60][61][62] and recent work among children has highlighted the frequency of AKI in children with DKA, [19][20][21][22] no studies to our knowledge have explicitly studied the association between race and ethnicity and diabetes-related AKI risk in the pediatric population.Thus, our finding that race and ethnicity constituted a significant factor associated with AKI across all levels of COI is novel, and future research should focus on racial and ethnic disparities in AKI among children with DKA.In terms of CE risk, we found no association between COI category or race and ethnicity and risk of CE.To our knowledge, no studies have explicitly examined the association between race and ethnicity and CE risk in children with DKA, likely because CE is a rare event, and large sample sizes would be needed to detect any differences.

Limitations
Our study has several limitations.First, we could not account for readmissions to non-PHIS hospitals, which precluded our ability to evaluate patient-level data.Second, factors not included in the COI may contribute to readmissions.For instance, exposure to adverse childhood experiences has been shown to have an association with hospitalization rates. 63It is also possible that other patient or clinician factors contribute to readmissions.We were unable to account for the presence of coexisting mental health conditions, which have been shown to be associated with readmissions for DKA in children. 7,64Implicit bias by clinicians may further influence treatment decisions, including hospitalization. 39,65Third, we were unable to reliably account for use of diabetes technology, which has been associated with lower rates of DKA. 50,66,67For example, the prevalence of the ICD-10 code for insulin pump use (Z96.41) in PHIS was only 3% and varied significantly by hospital from less than 1% to 18%.Documentation of insulin pump use has also been an issue in other administrative data sets. 67Fourth, the COI uses Census tract data, which may not align with how individuals perceive their neighborhood boundaries.This factor is important to recognize as interventions aimed at addressing health disparities are designed and implemented.1][22] This lower rate is likely due to ascertainment of AKI by ICD-9 and ICD-10 coding because we did not have access to laboratory values, whereas other studies were able to use creatinine data to define AKI.

Conclusions
Our study revealed significant disparities in DKA health outcomes associated with neighborhood opportunity.Readmissions were higher in patients with low COI categories, adding strain to already disadvantaged populations.The results of our study may be useful both for clinicians and health care systems as they seek ways to reduce health disparities and advocate for patients and families as well as for policy makers and community leaders who seek to enact change on a population level.

Figure 2 .
Figure 2. Probabilities of the Main Outcomes of Readmission and Diabetes-Related Complications by Race and Ethnicity and Overall Neighborhood Child Opportunity Racial and Ethnic Disparities in Diabetic Ketoacidosis Readmissions at US Children's Hospitals JAMA Network Open.2022;5(5):e2210456.doi:10.1001/jamanetworkopen.2022.10456( Racial and Ethnic Disparities in Diabetic Ketoacidosis Readmissions at US Children's Hospitals JAMA Network Open.2022;5(5):e2210456.doi:10.1001/jamanetworkopen.2022.10456(Reprinted) May 5, 2022 2/14 Downloaded From: https://jamanetwork.com/ on 09/16/2023 -100) and quintile categorical scores (very low, low, moderate, high, and very high, ) for each domain and a composite for all Census tracts and zip codes in the US.Because the PHIS only includes zip codes, we used the COI at the zip code level.Notably, zip codes were assigned at the encounter level and could change if a patient moved between encounters.

JAMA Network Open | Emergency Medicine
The probability of readmission for DKA was associated with COI category (eFigure in the Supplement).Readmission for DKA within 365 days was highest for children living in very low-opportunity neighborhoods (19.2% [95% CI, 17.8%-20.5%])(Table2).Comparatively, the Table 1.Demographic and Encounter Characteristics by COI Category for children in very high-opportunity neighborhoods was 5.4 percentage points lower (13.8%[95% CI, 12.6%-15.0%]).The interaction between racial and ethnic groups and COI category was statistically significant (P = .04).

Table 2 .
Probability of Readmission at 30 and 365 Days, Acute Kidney Injury, and Cerebral Edema by Race and Ethnicity and COI Category Racial and Ethnic Disparities in Diabetic Ketoacidosis Readmissions at US Children's Hospitals At all levels of overall COI, Black children had a greater probability of readmission compared with Hispanic children and White children (Figure2and Table2).Across racial groups, the probability of readmission within 365 days of an index encounter was significantly higher among Black children with an overall very low COI category compared with White children (risk difference, 7.5 [95% CI, 5.9-9.1]percentagepoints) and Hispanic children (risk difference, 7.8 [95% CI, 6.0-9.6]percentagepoints) at the same level of COI (Table2).Similar racial differences were seen for children with an Abbreviation: COI, Child Opportunity Index 2.0.JAMA Network Open | Emergency MedicineJAMA Network Open.2022;5(5):e2210456.doi:10.1001/jamanetworkopen.2022.10456(Reprinted) May 5, 2022 5/14 Downloaded From: https://jamanetwork.com/ on 09/16/2023 socioeconomic COI domains, the probability of readmission within 365 days was significantly higher compared with that of White children (risk difference, 6.2 [95% CI, 4.0-8.4]percentagepoints) and Hispanic children (risk difference, 6.7 [95% CI, 4.1-9.3]percentagepoints) with very low socioeconomic COI domains (eTable in the Supplement).The interaction between racial and ethnic group and COI domain was statistically significant for the socioeconomic domain (Wald χ 2 = 30.22;P= .02).Similar results were seen for the health and environment and Figure 3. Probability of Readmissions Within 365 Days by Race and Ethnicity and Neighborhood Child Opportunity DomainsOur findings suggest that Black children experience disparities in T1D care despite having similar measures of neighborhood opportunity when compared with other racial groups.Using data from 49 US children's hospitals, we observed that Black children had greater probability of readmission for DKA compared with Hispanic children and White children at all COI categories.Within racial and ethnic groups, children with an overall very low COI category had significantly greater probability of readmission compared with those with an overall very high COI category.Our results have implications for health systems seeking to reduce disparities.
Child opportunity index 2.0 (COI) categorical scores at or below the 20th percentile were categorized as very low; greater than 20th to at or below the 40th percentile, as low; greater than 40th to at or below the 60th percentile, as moderate; greater than 60th to at or below the 80th percentile, as high; and greater than 80th percentile, as very high.Whiskers represent 95% CIs.JAMA Network Open | Emergency MedicineRacial and Ethnic Disparities in Diabetic Ketoacidosis Readmissions at US Children's Hospitals JAMA Network Open.2022;5(5):e2210456.doi:10.1001/jamanetworkopen.2022.10456(Reprinted)May 5, 2022 6/14 Downloaded From: https://jamanetwork.com/ on 09/16/2023 with very low Child opportunity index 2.0 (COI) categorical scores at or below the 20th percentile were categorized as very low; greater than 20th to at or below the 40th percentile, as low; greater than 40th to at or below the 60th percentile, as moderate; greater than 60th to at or below the 80th percentile, as high; and greater than 80th percentile, as very high.Whiskers represent 95% CIs.Discussion

on 09/16/2023 living
found that the hospitalization rate for children with diabetes and associated complications was significantly higher at 2.8 per 1000 children for those in areas with a very low COI category compared with 1.5 per 1000 children in areas with a very high COI category.