OBJECTIVES:

Collecting social determinants data is challenging. We assigned patients a neighborhood-level social determinant measure, the area of deprivation index (ADI), by using census data. We then assessed the association between neighborhood deprivation and asthma hospitalization outcomes and tested the influence of insurance coverage.

METHODS:

A retrospective cohort study of children 2 to 17 years old admitted for asthma at 8 hospitals. An administrative database was used to collect patient data, including hospitalization outcomes and neighborhood deprivation status (ADI scores), which were grouped into quintiles (ADI 1, the least deprived neighborhoods; ADI 5, the most deprived neighborhoods). We used multivariable models, adjusting for covariates, to assess the associations and added a neighborhood deprivation status and insurance coverage interaction term.

RESULTS:

A total of 2270 children (median age 5 years; 40.6% girls) were admitted for asthma. We noted that higher ADI quintiles were associated with greater length of stay, higher cost, and more asthma readmissions (P < .05 for most quintiles). Having public insurance was independently associated with greater length of stay (β: 1.171; 95% confidence interval [CI]: 1.117–1.228; P < .001), higher cost (β: 1.147; 95% CI: 1.093–1.203; P < .001), and higher readmission odds (odds ratio: 1.81; 95% CI: 1.46–2.24; P < .001). There was a significant deprivation–insurance effect modification, with public insurance associated with worse outcomes and private insurance with better outcomes across ADI quintiles (P < .05 for most combinations).

CONCLUSIONS:

Neighborhood-level ADI measure is associated with asthma hospitalization outcomes. However, insurance coverage modifies this relationship and needs to be considered when using the ADI to identify and address health care disparities.

Despite advances in diagnosis and treatment, studies have reported persistent disparity gaps in asthma care and outcomes in both adults and children with lower socioeconomic status, particularly those with lower education, lower income, public insurance coverage, and of disadvantaged racial and/or ethnic groups.14  Disparities are associated with poorer asthma control, more frequent exacerbations, and higher emergency department and hospital admission and/or readmission rates.4,5 

Eliminating health disparities is 1 of the Healthy People 2010 improvement goals and is a key recommendation in the National Academy of Medicine (formerly the Institute of Medicine) 2001 report “Crossing the Quality Chasm.”6  Despite a decade of effort and some improvements, health disparities in general and particularly for children with asthma remain.7  A barrier to addressing health disparities is a difficulty collecting patient-level socioeconomic characteristics (eg, sex, race, ethnicity, education, income, poverty, housing, etc), which are known as social determinants of health.8,9 

The assessment of social determinants is time consuming and intrusive for frontline staff and patients.10  Besides basic demographics (age, sex, race, and ethnicity), most health care organizations do not routinely collect social determinant data, limiting their ability to compare performance and identify and address health care disparities. An improved collection of social determinants may guide effective interventions to reduce disparities in vulnerable populations.

Alternative measures at the neighborhood level that use readily available census variables11,12  offer an opportunity for health care organizations to obtain social determinant data. Previous studies have reported associations between neighborhood-level measures and health behaviors, mortality, and health disparities.1316  Yet, the use of neighborhood measures as a proxy for patient-level social determinants in the health care setting, particularly for children with asthma, is not widespread.

Intermountain Healthcare has implemented in the organization’s electronic medical record (EMR) system a measure of neighborhood-level social determinants, the area of deprivation index (ADI),17  by using Singh’s ADI approach.18  The ADI, also called a measure of neighborhood deprivation or neighborhood wealth,18,19  includes 17 US census block group variables related to education, employment, income and/or poverty, and housing and reflects the average socioeconomic characteristics of individuals within a neighborhood.20,21  Within the Intermountain Healthcare EMR, each patient is automatically assigned an ADI score on the basis of his or her geocoded address (the address is converted into geographic coordinates, including latitude and longitude) for use at the point of care and to support health care planning, research, and quality-improvement activities.

Previous studies have shown that private insurance patients have better clinical outcomes.22,23  Yet, the influence of insurance coverage on health care outcomes within deprived and advantaged neighborhoods has not been studied in children with asthma. The objective of this study was to investigate the value of the ADI by testing its association with the outcomes of children who are hospitalized with asthma. We also sought to assess the independent association of insurance coverage on outcomes and test if insurance coverage modifies the relationship between neighborhood deprivation and asthma outcomes.

Our study included 8 hospitals in Utah participating in a previous quality-improvement study24  involving all children age 2 to 17 years who were hospitalized for asthma. The intervention involved the implementation of a standardized asthma care process model and an assessment of its impact. These hospitals are operated by Intermountain Healthcare, a regional, not-for-profit, integrated health care delivery system with 22 hospitals and 185 clinics and urgent care facilities located in Utah and southeastern Idaho.25  Intermountain Healthcare provides health care services to ∼1 680 000 patients, including 60% of Utah’s residents and 95% of Utah’s children.26  Participating hospitals included Primary Children’s Hospital, a 289-bed, freestanding, academic, tertiary-care children’s hospital affiliated with the Department of Pediatrics at the University of Utah and located in Salt Lake City, and 7 community hospitals located throughout Utah, including the following: American Fork Hospital in American Fork, Dixie Regional Medical Center in St George, Logan Regional Hospital in Logan, McKay-Dee Hospital in Ogden, Riverton Hospital in Riverton, Utah Valley Hospital in Provo, and Cedar City Hospital in Cedar City. Overall, the 8 hospitals provide inpatient care to ∼700 children with asthma each year and serve urban (Primary Children’s Hospital, Riverton Hospital, McKay-Dee Hospital, and Utah Valley Hospital) and rural (American Fork Hospital, Dixie Regional Medical Center, Cedar City Hospital, and Logan Regional Hospital) populations. The Intermountain privacy board and the University of Utah institutional review board approved the study.

This was a retrospective cohort study of children 2 to 17 years old who were discharged from 1 of 8 Intermountain hospitals between January 1, 2010, and June 30, 2015, with a primary International Classification of Diseases, Ninth Revision diagnostic code 493 for asthma.

The enterprise data warehouse (EDW), an integrated database linking administrative and clinical data from Intermountain hospitals and clinics,27  was used to identify patients and collect data. Data collected included the following: patient demographics (age, sex, and race), ADI score, insurance (public, private, or self-pay), and severity of illness (SOI), which was divided into 4 levels (1, low severity; 4, high severity) by using All Patient Refined Diagnosis Related Groups.28  Within the EDW, ADIs are generated as raw scores based on 2013 US census block group data and categorized into quintiles (1, lower neighborhood deprivation; 5, greater neighborhood deprivation) of equal proportions (not at the patient level) and then assigned to patients given the most recent address on record, which is a method used in European census reporting statistics.1719 

Outcome variables, also collected by using the EDW, included hospital length of stay (LOS) in hours, variable hospitalization costs, and asthma-related 6-month readmission, which is defined as any emergency department revisit or hospital readmission for asthma as the primary diagnosis at any of 22 Intermountain hospitals within 6 months of a hospital discharge, as was used in a previous study.24  Variable cost data were derived from the Intermountain Healthcare internal costing system, which provides actual direct costs (not charges) associated with services (eg, costs of supplies and drugs, salaries of nurses and technicians, etc) rendered during each hospitalization.

Descriptive analyses were used to describe the study population, and regression analyses were used to assess the univariate relationship of the ADI or insurance status, independently and combined, with child asthma outcomes. Specifically, we used multiple linear regressions analysis to assess the impact on hospital LOS and costs after natural log transformation to account for skewing, and we used multiple logistic regression analysis to determine the impact on readmissions and then introduced the ADI quintile and/or insurance coverage interaction term. Multivariable analyses were adjusted for age, race and/or ethnicity, sex, and disease severity using All Patient Refined Diagnosis Related Groups SOI scores.

There were 2351 patients (3003 encounters) discharged from the 8 Intermountain hospitals between January 2010 and June 2015 with a primary International Classification of Diseases, Ninth Revision diagnostic code of 493 for asthma. One hundred twenty-three (4.1%) encounters without insurance coverage (self-pay) were excluded, leaving 2880 encounters (2270 patients) with either private (1761 encounters) or public insurance (1119 encounters) and ADI quintile assignments (Fig 1). Of the 1119 encounters with public insurance, 90.5% (n = 1013) were for patients with Medicaid, and 9.5% (n = 106) were for patients covered under the Children’s Health Insurance Program. The overall median age of the study population was 5 years (interquartile range [IQR] 3–8), and 40.0% (908 of 2270) of encounters were for female patients.

FIGURE 1

Consort diagram.

FIGURE 1

Consort diagram.

We found no difference in sex and SOI distributions (Table 1) based on neighborhood deprivation status (ADI quintiles) but significant differences in age (P = .026), distribution of race and/or ethnicity (P < .001), and insurance coverage (P < .001). White children (in ADI 1, 87.3% and ADI 2, 84.1%) were more likely to live in advantaged neighborhoods compared with minority children (ADI 1, 12.7% and ADI 2, 15.9%), whereas the disadvantaged neighborhoods had a mix of white (ADI 3, 62.7% and ADI 4, 53.9%) and minority (ADI 3, 37.3% and ADI 4, 46.1%) children. Patients of Hispanic ethnicity are the largest minority population in Utah, representing 16% of the population. In the advantaged neighborhoods (ADI 1and ADI 2), the majority (84.4% and 74.1%) of patients had private compared with public insurance (15.6% and 25.9%). Of patients living in disadvantaged neighborhoods (ADI 3 and ADI 4), 50.4% and 67.6% had public insurance compared with 49.6% and 32.4% with private insurance, respectively. On the basis of insurance coverage (private versus public), there were no differences in age, sex, and SOI distributions among groups, but there were significant differences (P < .001) in racial and/or ethnic distributions.

TABLE 1

Demographics by Study Grouping

Neighborhood Deprivation StatusInsurance Coverage Type
ADI 1ADI 2ADI 3ADI 4ADI 5PPublicPrivateP
No. encounters 557 652 621 482 568  1119 1761  
Age, y, median (IQR) 4 (3–7) 5 (3–8) 5 (3–8) 5 (3–8) 5 (3–8) .026 5 (3–8) 5 (3–8) .639 
Female sex, n (%) 241 (43.3) 280 (42.9) 235 (37.8) 191 (39.6) 222 (39.1) .213 441 (39.4) 728 (41.3) .304 
Race, n (%)      <.001   <.001 
 White 486 (87.3) 548 (84.1) 453 (72.9) 302 (62.7) 306 (53.9)  643 (57.5) 1452 (82.5)  
 Hispanic 9 (1.6) 10 (1.5) 36 (5.8) 24 (4.9) 63 (11.1)  106 (9.5) 36 (2.0)  
 Other 28 (5.0) 36 (5.5) 59 (9.5) 73 (15.2) 80 (14.1)  136 (12.1) 140 (7.9)  
 Unknown 34 (6.1) 58 (8.9) 73 (11.8) 83 (17.2) 119 (20.9)  234 (20.9) 133 (7.6)  
SOIa, median (IQR)      .620   .453 
 1 366 (65.7) 405 (62.1) 376 (60.5) 293 (60.8) 366 (64.4)  684 (61.1) 1122 (63.7)  
 2 144 (25.9) 190 (29.1) 182 (29.3) 143 (29.7) 150 (26.4)  324 (28.9) 485 (27.5)  
 3 39 (7.0) 48 (7.4) 55 (8.9) 44 (9.1) 44 (7.8)  98 (8.8) 132 (7.5)  
 4 8 (1.4) 9 (1.4) 8 (1.3) 2 (0.4) 8 (1.4)  13 (1.2) 22 (1.3)  
Public insurance, n (%) 87 (15.6) 169 (25.9) 236 (38.0) 243 (50.4) 384 (67.6) <.001 — — — 
Neighborhood Deprivation StatusInsurance Coverage Type
ADI 1ADI 2ADI 3ADI 4ADI 5PPublicPrivateP
No. encounters 557 652 621 482 568  1119 1761  
Age, y, median (IQR) 4 (3–7) 5 (3–8) 5 (3–8) 5 (3–8) 5 (3–8) .026 5 (3–8) 5 (3–8) .639 
Female sex, n (%) 241 (43.3) 280 (42.9) 235 (37.8) 191 (39.6) 222 (39.1) .213 441 (39.4) 728 (41.3) .304 
Race, n (%)      <.001   <.001 
 White 486 (87.3) 548 (84.1) 453 (72.9) 302 (62.7) 306 (53.9)  643 (57.5) 1452 (82.5)  
 Hispanic 9 (1.6) 10 (1.5) 36 (5.8) 24 (4.9) 63 (11.1)  106 (9.5) 36 (2.0)  
 Other 28 (5.0) 36 (5.5) 59 (9.5) 73 (15.2) 80 (14.1)  136 (12.1) 140 (7.9)  
 Unknown 34 (6.1) 58 (8.9) 73 (11.8) 83 (17.2) 119 (20.9)  234 (20.9) 133 (7.6)  
SOIa, median (IQR)      .620   .453 
 1 366 (65.7) 405 (62.1) 376 (60.5) 293 (60.8) 366 (64.4)  684 (61.1) 1122 (63.7)  
 2 144 (25.9) 190 (29.1) 182 (29.3) 143 (29.7) 150 (26.4)  324 (28.9) 485 (27.5)  
 3 39 (7.0) 48 (7.4) 55 (8.9) 44 (9.1) 44 (7.8)  98 (8.8) 132 (7.5)  
 4 8 (1.4) 9 (1.4) 8 (1.3) 2 (0.4) 8 (1.4)  13 (1.2) 22 (1.3)  
Public insurance, n (%) 87 (15.6) 169 (25.9) 236 (38.0) 243 (50.4) 384 (67.6) <.001 — — — 

—, not applicable.

a

SOI 1, minor; 2, moderate; 3, major; 4, extreme.

In univariate analysis (Table 2, top 2 sections), we noted that LOS, costs, and readmission rates were significantly different (P < .001) across ADI quintiles. Specifically, as neighborhood deprivation increases from low- to high-ADI quintiles, LOS, costs, and readmission rates also tend to increase, with patients living in the least deprived neighborhoods (ADI1) having the best outcomes and those living in the most deprived neighborhoods (ADI5) having worse outcomes. Patients with public (compared with private) insurance also had greater LOS, higher costs, and higher readmission rates (P < .001). In the combined groupings of neighborhood deprivation and insurance coverage (Table 2, bottom section), patients with public (compared with private) insurance had consistently greater LOS, higher costs, and more readmissions within each ADI quintile–insurance combination; however, across quintiles, the relationship was less consistent (eg, LOS, cost, and readmissions did not progressively increase with increasing ADI quintile in patients with public insurance).

TABLE 2

Outcomes by Study Grouping (Univariate, Unadjusted Analysis)

LOS, h, Median (IQR)Hospitalization Costs, Median (IQR)6-mo Readmission, % (SD)
Neighborhood deprivation status    
 ADI 1 38 (25–55) 1499 (1008–2264) 11.1 (31.5) 
 ADI 2 40 (27–55) 1585 (1099–2326) 12.0 (32.5) 
 ADI 3 42 (30–66) 1649 (1094–2770) 15.8 (36.5) 
 ADI 4 41 (28–63) 1631 (1051–2609) 12.7 (33.3) 
 ADI 5 43 (31–64) 1684 (1149–2789) 18.7 (39.0) 
 P <.001 <.001 <.001 
Insurance coverage type    
 Public 43 (31–67) 1736 (1163–2829) 18.4 (38.8) 
 Private 39 (26–57) 1542 (1037–2311) 11.3 (31.7) 
 P <.001 <.001 <.001 
Combined groupings    
 ADI 1, private 37 (25–55) 1441 (982–2238) 11.1 (31.4) 
 ADI 1, public 41 (27–54) 1832 (1231–2487) 11.5 (32.1) 
 ADI 2, private 38 (26–50) 1534 (1061–2192) 10.5 (30.7) 
 ADI 2, public 46 (32–73) 1867 (1198–3022) 15.9 (36.7) 
 ADI 3, private 41 (29–66) 1609 (1084–2691) 11.9 (32.5) 
 ADI 3, public 44 (31–66) 1734 (1128–2845) 22.0 (41.5) 
 ADI 4, private 41 (24–62) 1603 (999–2546) 10.5 (30.7) 
 ADI 4, public 42 (30–66) 1640 (1128–2727) 14.8 (35.6) 
 ADI 5, private 42 (28–58) 1567 (1084–2441) 13.6 (34.3) 
 ADI 5, public 43 (33–66) 1735 (1183–2950) 21.1 (40.8) 
LOS, h, Median (IQR)Hospitalization Costs, Median (IQR)6-mo Readmission, % (SD)
Neighborhood deprivation status    
 ADI 1 38 (25–55) 1499 (1008–2264) 11.1 (31.5) 
 ADI 2 40 (27–55) 1585 (1099–2326) 12.0 (32.5) 
 ADI 3 42 (30–66) 1649 (1094–2770) 15.8 (36.5) 
 ADI 4 41 (28–63) 1631 (1051–2609) 12.7 (33.3) 
 ADI 5 43 (31–64) 1684 (1149–2789) 18.7 (39.0) 
 P <.001 <.001 <.001 
Insurance coverage type    
 Public 43 (31–67) 1736 (1163–2829) 18.4 (38.8) 
 Private 39 (26–57) 1542 (1037–2311) 11.3 (31.7) 
 P <.001 <.001 <.001 
Combined groupings    
 ADI 1, private 37 (25–55) 1441 (982–2238) 11.1 (31.4) 
 ADI 1, public 41 (27–54) 1832 (1231–2487) 11.5 (32.1) 
 ADI 2, private 38 (26–50) 1534 (1061–2192) 10.5 (30.7) 
 ADI 2, public 46 (32–73) 1867 (1198–3022) 15.9 (36.7) 
 ADI 3, private 41 (29–66) 1609 (1084–2691) 11.9 (32.5) 
 ADI 3, public 44 (31–66) 1734 (1128–2845) 22.0 (41.5) 
 ADI 4, private 41 (24–62) 1603 (999–2546) 10.5 (30.7) 
 ADI 4, public 42 (30–66) 1640 (1128–2727) 14.8 (35.6) 
 ADI 5, private 42 (28–58) 1567 (1084–2441) 13.6 (34.3) 
 ADI 5, public 43 (33–66) 1735 (1183–2950) 21.1 (40.8) 

LOS and hospitalization costs were log transformed before analysis. ADI 1, least deprived; ADI 5, most deprived.

After controlling for age, race and/or ethnicity, sex, and disease severity, we found that overall ADI scores were positively associated with increased LOS (β: 1.04; P < .001), costs (β: 1.03, P < .001), and readmissions (odds ratio [OR]: 1.15; 95% confidence interval [CI]: 1.06–1.24). Also, we found that higher neighborhood deprivation quintiles (ADIs 3–5) were independently associated (P < .05 for most quintile) with greater LOS, higher costs, and greater 6-month odds of readmission by using ADI quintile 1 (the most advantaged) as the reference group (Table 3, top 2 sections). Furthermore, we found that public insurance recipients had significantly longer LOS (β: 1.171; P < .001), higher costs (β: 1.147; P < .001), and higher odds of readmission (OR: 1.81; 95% CI: 1.46–2.24; P < .001) by using private insurance as the reference group (Table 3, middle section).

TABLE 3

Comparisons by Study Grouping (Adjusted Analysis for Age, Sex, Race, and SOI)

nLOS, h, β-Coefficient (95% CI)PHospitalization Costs, β-Coefficient (95% CI)P6-mo Readmission, OR (95% CI)P
Neighborhood deprivation        
 ADI 1 557 0.0 (reference)  0.0 (reference)  1.0 (reference)  
 ADI 2 652 1.060 (0.989–1.135) .102 1.043 (0.971–1.120) .246 1.08 (0.76–1.54) .673 
 ADI 3 621 1.179 (1.100–1.265) <.001a 1.134 (1.055–1.219) .001a 1.50 (1.07–2.11) .020a 
 ADI 4 482 1.141 (1.059–1.230) .001a 1.081 (1.001–1.168) .046a 1.16 (0.79–1.69) .435 
 ADI 5 568 1.181 (1.099–1.269) <.001a 1.141 (1.060–1.230) <.001a 1.84 (1.30–2.58) .001a 
Insurance coverage type        
 Private 1761 0.0 (reference)  0.0 (reference)  1.0 (reference)  
 Public 1119 1.171 (1.117–1.228) <.001a 1.147 (1.093–1.203) <.001a 1.81 (1.46–2.24) <.001a 
Combined groupings        
 ADI 1, private 470 0.0 (reference)  0.0 (reference)  1.0 (reference)  
 ADI 1, public 87 1.063 (0.925–1.220) .388 1.161 (1.007–1.339) .040a 1.03 (0.50–2.12) .929 
 ADI 2, private 483 1.001 (0.928–1.081) .970 1.011 (0.934–2.465) .787 0.94 (0.62–1.41) .759 
 ADI 2, public 169 1.294 (1.163–1.441) <.001a 1.250 (1.120–1.395) <.001a 1.55 (0.94–2.57) .088 
 ADI 3, private 385 1.156 (1.065–1.255) <.001a 1.141 (1.049–1.241) .002a 1.09 (0.71–1.66) .702 
 ADI 3, public 236 1.256 (1.142–1.383) <.001a 1.200 (1.088–1.323) <.001a 2.35 (1.54–3.59) <.001a 
 ADI 4, private 239 1.085 (0.988–1.194) .088 1.069 (0.970–1.178) .176 0.95 (0.57–1.57) .831 
 ADI 4, public 243 1.230 (1.120–1.353) <.001a 1.151 (1.045–1.269) .004a 1.43 (0.91–2.27) .123 
 ADI 5, private 184 1.100 (0.992–1.220) .071 1.064 (0.956–1.183) .255 1.27 (0.76–2.12) .360 
 ADI 5, public 384 1.247 (1.148–1.355) <.001a 1.230 (1.129–1.339) <.001a 2.21 (1.50–3.24) <.001a 
nLOS, h, β-Coefficient (95% CI)PHospitalization Costs, β-Coefficient (95% CI)P6-mo Readmission, OR (95% CI)P
Neighborhood deprivation        
 ADI 1 557 0.0 (reference)  0.0 (reference)  1.0 (reference)  
 ADI 2 652 1.060 (0.989–1.135) .102 1.043 (0.971–1.120) .246 1.08 (0.76–1.54) .673 
 ADI 3 621 1.179 (1.100–1.265) <.001a 1.134 (1.055–1.219) .001a 1.50 (1.07–2.11) .020a 
 ADI 4 482 1.141 (1.059–1.230) .001a 1.081 (1.001–1.168) .046a 1.16 (0.79–1.69) .435 
 ADI 5 568 1.181 (1.099–1.269) <.001a 1.141 (1.060–1.230) <.001a 1.84 (1.30–2.58) .001a 
Insurance coverage type        
 Private 1761 0.0 (reference)  0.0 (reference)  1.0 (reference)  
 Public 1119 1.171 (1.117–1.228) <.001a 1.147 (1.093–1.203) <.001a 1.81 (1.46–2.24) <.001a 
Combined groupings        
 ADI 1, private 470 0.0 (reference)  0.0 (reference)  1.0 (reference)  
 ADI 1, public 87 1.063 (0.925–1.220) .388 1.161 (1.007–1.339) .040a 1.03 (0.50–2.12) .929 
 ADI 2, private 483 1.001 (0.928–1.081) .970 1.011 (0.934–2.465) .787 0.94 (0.62–1.41) .759 
 ADI 2, public 169 1.294 (1.163–1.441) <.001a 1.250 (1.120–1.395) <.001a 1.55 (0.94–2.57) .088 
 ADI 3, private 385 1.156 (1.065–1.255) <.001a 1.141 (1.049–1.241) .002a 1.09 (0.71–1.66) .702 
 ADI 3, public 236 1.256 (1.142–1.383) <.001a 1.200 (1.088–1.323) <.001a 2.35 (1.54–3.59) <.001a 
 ADI 4, private 239 1.085 (0.988–1.194) .088 1.069 (0.970–1.178) .176 0.95 (0.57–1.57) .831 
 ADI 4, public 243 1.230 (1.120–1.353) <.001a 1.151 (1.045–1.269) .004a 1.43 (0.91–2.27) .123 
 ADI 5, private 184 1.100 (0.992–1.220) .071 1.064 (0.956–1.183) .255 1.27 (0.76–2.12) .360 
 ADI 5, public 384 1.247 (1.148–1.355) <.001a 1.230 (1.129–1.339) <.001a 2.21 (1.50–3.24) <.001a 

LOS and costs were log transformed before analysis. ADI 1, least deprived; ADI 5, most deprived.

a

Statistically significant.

After adding an ADI quintile–insurance coverage interaction term, we found that insurance coverage modifies the relationship between neighborhood deprivation and asthma hospitalization outcomes (Table 3, bottom section). Within quintiles, patients with public insurance had consistently greater (P < .05 for most combinations) LOS, higher costs, and higher odds of readmission, whereas patients with private insurance had shorter LOS, lower costs, and smaller readmission odds by using patients living in the least deprived neighborhoods with private insurance (ADI quintile 1, private) as the reference group. However, across quintiles, the relationship was more complex, without consistent improvement or worsening outcomes as ADI increased in either public or privately insured patients.

Overall, patients who were living in the most deprived neighborhoods showed an 18% increase in LOS, 14% increase in costs, and 1.84 times–increased readmission odds compared with those in quintile 1 (Table 3, top section). Patients with public insurance had a 17% increase in LOS, ∼15% increase in costs, and 1.81 times–increased readmission odds compared with those with private insurance (Table 3, middle section). However, when patients living in most deprived neighborhoods had public insurance, they had an ∼25% increase in both LOS and costs, and 2.2-fold increased readmission odds compared with those living in least deprived neighborhoods with private insurance (Table 3, bottom section).

We found that the ADI was associated independently with asthma hospitalization outcomes. We specifically noted an association of higher ADI quintiles with increased LOS, costs, and readmission rates and lower ADI quintiles (lower neighborhood deprivation) with reduced LOS, costs, and readmission rates. Overall, the ADI may have value as a proxy for patient-level social determinants for children who are hospitalized with asthma, with potential for use in other health conditions for both children and adults.

Barriers to addressing health care disparities in health care settings include a difficulty collecting patient-level social determinant data because collection is time consuming and intrusive.10  With increasing movement toward value-based purchasing and pay for performance in the United States, health care researchers and policy makers need easy ways of incorporating social determinant data. Easily computed and validated, neighborhood-level measures derived from readily available data offer an easy way to include social determinant data in health care systems if they can be shown to be comparable to patient-level social determinants. The ADI, which is generated from low-cost census data,18  can be stored in an electronic format and assigned automatically to patients in the EMR on the basis of the patient address for use in point-of-care interventions and comparative analyses to support health care planning, research, and quality improvement.

Our findings are consistent with the few previous studies that used neighborhood-level social determinant measures to evaluate their impact on asthma outcomes in children. Using limited census variables, Liu and Pearlman29  reported an association between disadvantaged neighborhoods and higher readmission rates in children with asthma. A nationwide study from Sweden also reported an association between disadvantaged neighborhoods and an increased risk of childhood asthma hospitalization independent of maternal-level sociodemographic characteristics.30  Beck et al31  used neighborhood-level measures in multiple studies and identified that children with asthma are likely to return to the hospital after discharge and that neighborhood toxic stressors affect asthma control.32  Juhn et al developed and validated a housing-based socioeconomic status index called HOUSES to overcome the lack of socioeconomic measures and demonstrated its value in multiple studies.33  More recently, Auger et al34  reported a correlation between individual family characteristics and neighborhood factors in children who are hospitalized with asthma. Other non-asthma–related studies in adults also found an association between disadvantaged neighborhood variables and poor health and between increased hospitalization and mortality.12,35  Yet, previous studies have been inconsistent in their selection of census-based variables, often including a limited number (<5) and varying combinations of variables and omitting many key variables previously reported to be associated with outcomes.20,21,36  In our study, we used the ADI, which includes a comprehensive and standardized list of 17 key US census block group variables.31  Despite being validated 2 decades ago,18  ADI use in the US health care setting remains uncommon.19  Intermountain Healthcare is currently using the ADI at the point of care to identify high-risk, first-time expectant mothers for referral to the Salt Lake County–based Nurse Family-Partnership program and to identify children with asthma from more deprived neighborhoods with previous asthma admissions for referral to home environmental assessment and/or remediation. We are not aware of its previous integration into an EMR or its use at the point of care.

As previously reported in the literature, we also found that insurance coverage was an independent predictor of asthma outcomes. We specifically noted that having private compared with public insurance was independently associated with better asthma outcomes. Our findings are consistent with previous studies in both children and adults that have demonstrated this relationship,23,37,38  although a recent study has raised some uncertainty.39  However, in our study, we explore this further by looking at the modifying effect of insurance coverage on the relationship between the ADI and asthma outcomes. Specifically, we found that the relationship between the ADI and asthma outcomes is modified by insurance coverage, although for odds of readmission, results were significant only for ADI 3, public and ADI 5, public, which is possibly due to limited power to detect differences because overall readmission rate was low at 14%. This suggests that having private insurance may have the potential to mitigate the negative effects of living in disadvantaged neighborhoods. The basis of this is likely multifactorial, including higher household income (despite neighborhoods) and better access to outpatient services and asthma medications, leading to improved asthma control and reduced asthma exacerbation frequency and severity.

Our study has several limitations. First, our study includes data from 2010 to 2015, but ADI calculation was based on 2013 census data (because the ADI is not updated yearly). The actual deprivation status of individual neighborhoods may have changed during the study period, and such change may affect our results. However, we believe changes, if any, are minimal because individual neighborhood deprivation status is unlikely to change in this time frame, and if patients move, it is likely they will move to neighborhoods of comparable deprivation status.40 

Second, our study was not designed to uncover neighborhood or insurance coverage contextual factors associated with poorer asthma outcomes. Although we did not collect neighborhood contextual factors, we can speculate about factors explaining differences in outcomes based on insurance coverage, such as lower access to continuous care, because lower public insurance reimbursement to providers may contribute to outpatient clinics limiting or refusing publicly insured patients, thus hampering a child’s ability to receive early and comprehensive care and increasing the risk of having poor asthma control and exacerbations.41,42  Public insurance recipients are also often financially challenged and have limited education, both of which contribute to low adherence to treatment, and are more likely to live in environments with increased exposure to asthma triggers, making asthma exacerbations more common.43  Overall, poor outcomes among patients with public insurance may highlight the unmet health care needs of recipients, including limited access to outpatient services and asthma medications.

A third problem with our study is a limitation of the generalizability of our findings beyond our study populations for the following reasons: (1) our study was conducted in Utah, a less diverse state with a high proportion of white children. Results could be different if the study was conducted in states with more racially and socioeconomically diverse populations; (2) because of the unique characteristics of US health care financing, our findings may not be applicable to other countries.

Fourth, we only measured readmissions to the Intermountain Healthcare system and may have missed readmissions at non-Intermountain facilities. The impact of this is unknown but is unlikely to change our results because Intermountain Healthcare covers the 95% of pediatric inpatient care occurring in Utah. Finally, we included SOI as a confounding factor in our regression analyses because we did not collect data on asthma chronic severity. SOI is an imperfect marker of illness severity at presentation because it reflects diagnosis codes and resource use after admission. If we had been able to include chronic severity data, results may have been different. Our results showing poor outcomes of patients living in the most deprived neighborhoods could be due to a disproportionate number of sicker children living in those neighborhoods.

Our future plans include a direct comparison of ADI and patient-level social determinants, defining other ADI-sensitive diseases, assessing the underlying mechanisms and neighborhood contextual factors responsible for the associations with health care outcomes, and defining potential point-of-care and community-based interventions to mitigate the negative effects of neighborhood deprivation and improve the outcomes of vulnerable patients.

The ADI, a neighborhood-level measure, exhibits associations with asthma hospitalization outcomes and may be a proxy of patient-level social determinants that can be used to support health care planning, research, quality improvement, and point-of-care interventions to reduce disparities. However, in childhood asthma, insurance coverage modifies the relationship between neighborhood characteristics and outcomes and needs to be considered when assessing the impact of neighborhood deprivation or implementing targeted interventions to address disparities in vulnerable patients.

FUNDING: No external funding.

Drs Knighton, Stone, and Nkoy conceptualized and designed the study, participated in data collection, analysis, and interpretation of results, and drafted the initial manuscript; Drs Fassl, Johnson, Maloney, and Savitz participated in the study concept, design, and interpretation of the data, and critically review of the manuscript; and all authors approved the final manuscript as submitted.

1
Beck
AF
,
Huang
B
,
Auger
KA
,
Ryan
PH
,
Chen
C
,
Kahn
RS
.
Explaining racial disparities in child asthma readmission using a causal inference approach
.
JAMA Pediatr
.
2016
;
170
(
7
):
695
703
2
Bloomberg
GR
,
Banister
C
,
Sterkel
R
, et al
.
Socioeconomic, family, and pediatric practice factors that affect level of asthma control
.
Pediatrics
.
2009
;
123
(
3
):
829
835
3
Crocker
D
,
Brown
C
,
Moolenaar
R
, et al
.
Racial and ethnic disparities in asthma medication usage and health-care utilization: data from the National Asthma Survey
.
Chest
.
2009
;
136
(
4
):
1063
1071
4
Malhotra
K
,
Baltrus
P
,
Zhang
S
,
McRoy
L
,
Immergluck
LC
,
Rust
G
.
Geographic and racial variation in asthma prevalence and emergency department use among Medicaid-enrolled children in 14 southern states
.
J Asthma
.
2014
;
51
(
9
):
913
921
5
Bacon
SL
,
Bouchard
A
,
Loucks
EB
,
Lavoie
KL
.
Individual-level socioeconomic status is associated with worse asthma morbidity in patients with asthma
.
Respir Res
.
2009
;
10
:
125
6
Committee on Quality Health Care in America; Institute of Medicine
.
Crossing the Quality Chasm: A New Health System for the 21st Century
.
Washington, DC
:
National Academy Press
,
2001
7
Hunt
B
,
Whitman
S
.
Black: white health disparities in the United States and Chicago: 1990-2010
.
J Racial Ethn Health Disparities
.
2015
;
2
(
1
):
93
100
8
Baker
DW
,
Hasnain-Wynia
R
,
Kandula
NR
,
Thompson
JA
,
Brown
ER
.
Attitudes toward health care providers, collecting information about patients’ race, ethnicity, and language
.
Med Care
.
2007
;
45
(
11
):
1034
1042
9
Hasnain-Wynia
R
,
Baker
DW
.
Obtaining data on patient race, ethnicity, and primary language in health care organizations: current challenges and proposed solutions
.
Health Serv Res
.
2006
;
41
(
4, pt 1
):
1501
1518
10
Hill
J
,
Nielsen
M
,
Fox
MH
.
Understanding the social factors that contribute to diabetes: a means to informing health care and social policies for the chronically ill
.
Perm J
.
2013
;
17
(
2
):
67
72
11
Brennan
SL
,
Turrell
G
.
Neighborhood disadvantage, individual-level socioeconomic position, and self-reported chronic arthritis: a cross-sectional multilevel study
.
Arthritis Care Res (Hoboken)
.
2012
;
64
(
5
):
721
728
12
Brown
AF
,
Liang
LJ
,
Vassar
SD
, et al
.
Neighborhood disadvantage and ischemic stroke: the Cardiovascular Health Study (CHS)
.
Stroke
.
2011
;
42
(
12
):
3363
3368
13
Poyser
MA
,
Nelson
H
,
Ehrlich
RI
, et al
.
Socioeconomic deprivation and asthma prevalence and severity in young adolescents
.
Eur Respir J
.
2002
;
19
(
5
):
892
898
14
Roberts
ME
,
Lowndes
L
,
Milne
DG
,
Wong
CA
.
Socioeconomic deprivation, readmissions, mortality and acute exacerbations of bronchiectasis
.
Intern Med J
.
2012
;
42
(
6
):
e129
e136
15
Krieger
N
,
Feldman
JM
,
Waterman
PD
,
Chen
JT
,
Coull
BA
,
Hemenway
D
.
Local residential segregation matters: stronger association of census tract compared to conventional city-level measures with fatal and non-fatal assaults (total and firearm related), using the Index of Concentration at the Extremes (ICE) for racial, economic, and racialized economic segregation, Massachusetts (US), 1995-2010
.
J Urban Health
.
2017
;
94
(
2
):
244
258
16
Krieger
N
,
Waterman
PD
,
Spasojevic
J
,
Li
W
,
Maduro
G
,
Van Wye
G
.
Public health monitoring of privilege and deprivation with the index of concentration at the extremes
.
Am J Public Health
.
2016
;
106
(
2
):
256
263
17
Knighton
AJ
,
Savitz
L
,
Belnap
T
,
Stephenson
B
,
VanDerslice
J
.
Introduction of an area deprivation index measuring patient socioeconomic status in an integrated health system: implications for population health
.
EGEMS (Wash DC)
.
2016
;
4
(
3
):
1238
18
Singh
GK
.
Area deprivation and widening inequalities in US mortality, 1969-1998
.
Am J Public Health
.
2003
;
93
(
7
):
1137
1143
19
Kind
AJ
,
Jencks
S
,
Brock
J
, et al
.
Neighborhood socioeconomic disadvantage and 30-day rehospitalization: a retrospective cohort study
.
Ann Intern Med
.
2014
;
161
(
11
):
765
774
20
Begaj
I
,
Khosla
S
,
Ray
D
,
Sharif
A
.
Socioeconomic deprivation is independently associated with mortality post kidney transplantation
.
Kidney Int
.
2013
;
84
(
4
):
803
809
21
Grimaud
O
,
Leray
E
,
Lalloué
B
, et al
.
Mortality following stroke during and after acute care according to neighbourhood deprivation: a disease registry study
.
J Neurol Neurosurg Psychiatry
.
2014
;
85
(
12
):
1313
1318
22
Franklin
JA
,
Anderson
EJ
,
Wu
X
,
Ambrose
CS
,
Simões
EA
.
Insurance status and the risk of severe respiratory syncytial virus disease in United States preterm infants born at 32-35 weeks gestational age
.
Open Forum Infect Dis
.
2016
;
3
(
3
):
ofw163
23
Inverso
G
,
Mahal
BA
,
Aizer
AA
,
Donoff
RB
,
Chuang
SK
.
Health insurance affects head and neck cancer treatment patterns and outcomes
.
J Oral Maxillofac Surg
.
2016
;
74
(
6
):
1241
1247
24
Nkoy
F
,
Fassl
B
,
Stone
B
, et al
.
Improving pediatric asthma care and outcomes across multiple hospitals
.
Pediatrics
.
2015
;
136
(
6
).
25
James
BC
,
Savitz
LA
.
How intermountain trimmed health care costs through robust quality improvement efforts
.
Health Aff (Millwood)
.
2011
;
30
(
6
):
1185
1191
26
Byington
CL
,
Reynolds
CC
,
Korgenski
K
, et al
.
Costs and infant outcomes after implementation of a care process model for febrile infants
.
Pediatrics
.
2012
;
130
(
1
).
27
Evans
RS
,
Lloyd
JF
,
Pierce
LA
.
Clinical use of an enterprise data warehouse
. In:
AMIA Annual Symposium Proceedings 2012; November 3–7, 2012; Chicago, IL. 189–198
28
Baram
D
,
Daroowalla
F
,
Garcia
R
, et al
.
Use of the all patient refined-diagnosis related group (APR-DRG) risk of mortality score as a severity adjustor in the medical ICU
.
Clin Med Circ Respirat Pulm Med
.
2008
;
2
:
19
25
29
Liu
SY
,
Pearlman
DN
.
Hospital readmissions for childhood asthma: the role of individual and neighborhood factors
.
Public Health Rep
.
2009
;
124
(
1
):
65
78
30
Li
X
,
Sundquist
J
,
Calling
S
,
Zoller
B
,
Sundquist
K
.
Mothers, places and risk of hospitalization for childhood asthma: a nationwide study from Sweden: epidemiology of allergic disease
.
Clin Exp Allergy
.
2013
;
43
(
6
):
652
658
31
Beck
AF
,
Huang
B
,
Ryan
PH
,
Sandel
MT
,
Chen
C
,
Kahn
RS
.
Areas with high rates of police-reported violent crime have higher rates of childhood asthma morbidity
.
J Pediatr
.
2016
;
173
:
175.e1
182.e1
32
Beck
AF
,
Simmons
JM
,
Huang
B
,
Kahn
RS
.
Geomedicine: area-based socioeconomic measures for assessing risk of hospital reutilization among children admitted for asthma
.
Am J Public Health
.
2012
;
102
(
12
):
2308
2314
33
Juhn
YJ
,
Beebe
TJ
,
Finnie
DM
, et al
.
Development and initial testing of a new socioeconomic status measure based on housing data
.
J Urban Health
.
2011
;
88
(
5
):
933
944
34
Auger
KA
,
Kahn
RS
,
Simmons
JM
, et al
.
Using address information to identify hardships reported by families of children hospitalized with asthma
.
Acad Pediatr
.
2017
;
17
(
1
):
79
87
35
Stahler
GJ
,
Mennis
J
,
Cotlar
R
,
Baron
DA
.
The influence of neighborhood environment on treatment continuity and rehospitalization in dually diagnosed patients discharged from acute inpatient care
.
Am J Psychiatry
.
2009
;
166
(
11
):
1258
1268
36
Aljuburi
G
,
Laverty
AA
,
Green
SA
,
Phekoo
KJ
,
Bell
D
,
Majeed
A
.
Socioeconomic deprivation and risk of emergency readmission and inpatient mortality in people with sickle cell disease in England: observational study
.
J Public Health (Oxf)
.
2013
;
35
(
4
):
510
517
37
Bratton
SL
,
Roberts
JS
,
Watson
RS
,
Cabana
MD
.
Acute severe asthma: outcome and Medicaid insurance
.
Pediatr Crit Care Med
.
2002
;
3
(
3
):
234
238
38
Rosenberg
AR
,
Kroon
L
,
Chen
L
,
Li
CI
,
Jones
B
.
Insurance status and risk of cancer mortality among adolescents and young adults
.
Cancer
.
2015
;
121
(
8
):
1279
1286
39
Silber
JH
,
Rosenbaum
PR
,
Wang
W
, et al
.
Practice patterns in Medicaid and non-Medicaid asthma admissions
.
Pediatrics
.
2016
;
138
(
2
):
e20160371
40
Kunz
J
,
Pare
M
,
Solon
G
.
Are single-year measures of neighborhood characteristics useful proxies for children’s long-run neighborhood environment?
Econ Lett
.
2001
;
79
(
2
):
231
237
41
Bhandari
N
,
Shi
Y
,
Jung
K
.
Patient experience of provider refusal of Medicaid coverage and its implications
.
J Health Care Poor Underserved
.
2016
;
27
(
2
):
479
494
42
Pierce
TR
,
Mehlman
CT
,
Tamai
J
,
Skaggs
DL
.
Access to care for the adolescent anterior cruciate ligament patient with Medicaid versus private insurance
.
J Pediatr Orthop
.
2012
;
32
(
3
):
245
248
43
Rosser
FJ
,
Forno
E
,
Cooper
PJ
,
Celedón
JC
.
Asthma in Hispanics. An 8-year update
.
Am J Respir Crit Care Med
.
2014
;
189
(
11
):
1316
1327

Competing Interests

POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.

FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.