Socioeconomic Disparities and Emergency Department Visits for Diabetic Retinopathy in California

Abstract

Purpose: To investigate the association between neighborhood-level socioeconomic factors, quantified by the Distressed Communities Index, and emergency department visits for diabetic retinopathy (DR). Methods: All patients who presented to the emergency department for DR in California were analyzed using the State Emergency Department Database (2018–2020). Patients were stratified by Distressed Communities Index score and DR severity. Logistic regression was applied to explore the independent correlation between Distressed Communities Index scores and proliferative DR (PDR). Results: Of 2 725 195 emergency department visits for diabetic patients, Distressed Communities Index data were available for 2 459 577 (90.3%); 39 693 were for DR, including 13 617 (34.3%) for PDR. Hispanics (44.2%) were the largest racial/ethnic group to present for PDR, followed by non-Hispanic Whites (19.6%) and non-Hispanic Blacks (19.3%). A significant association was observed between the Distressed Communities Index and emergency department visits for PDR, with distressed neighborhoods having the highest incidence (adjusted odds ratio [aOR], 1.63; 95% CI, 1.20-2.23; P = .001). Other predictors included Hispanic ethnicity (aOR, 2.21; 95% CI, 1.97-2.48; P < .001) and Black race (aOR, 1.46; 95% CI, 1.28-1.67; P < .001) compared with White race and having Medicaid (aOR, 1.37; 95% CI, 1.13-1.65; P = .001) compared with private insurance. Conclusions: The Distressed Communities Index identified patients residing in the most distressed neighborhoods as being at the highest risk for presenting to the emergency department for PDR based on 7 socioeconomic factors. Policymakers may consider the Distressed Communities Index as a tool for targeting DR prevention strategies and improving healthcare accessibility.

Introduction

The US Department of Health and Human Services developed the Healthy People 2030 program ¹ to promote the health of all people in the United States and eliminate health disparities. It identified social determinants of health, encompassing the environments in which people live, learn, work, and play, as critical factors affecting health and quality-of-life outcomes that contribute to health disparities and inequities.

Diabetic retinopathy (DR) is the leading cause of blindness in working-age people in the US. Still, not all people share the same risk for developing or losing vision from DR. Previous studies have described how individual-level factors including race^{2 –4} and proxies for socioeconomic position such as educational level^5,6 or insurance coverage ⁷ can affect the prevalence of DR. Lacking is how a comprehensive and systematic approach can be implemented to predict the influence of community-level environmental factors on disease presentation and outcomes.

In this study, we leveraged data from the US Healthcare Costs and Utilization Project’s State Emergency Department Databases ⁸ to describe the patient characteristics of those presenting to the emergency department as a result of DR and explored the association between neighborhood-level socioeconomic disparities and emergency department utilization. As Baxter et al ⁹ described, the type of databases often used contribute to the current gaps in data on social determinants of health in ophthalmology. Those derived from insurance-based claims exclude uninsured patients and often those with Medicaid and, therefore, cannot comment on groups of patients who frequently encounter barriers to healthcare access and are at risk for worse health outcomes. Databases derived from outpatient electronic health records often lack granular data on social determinants of health, such as education level and income. The US National Institutes of Health recently launched the All of Us database with the aim of enrolling diverse and underrepresented research participants; however, patient response rates to queries about socioeconomic variables have been especially low. ¹⁰

By pairing the State Emergency Department Database with the Distressed Communities Index, ¹¹ we were able to include all patients, regardless of insurance status, and simultaneously evaluate a broad range of socioeconomic factors affecting the severity of DR in patients presenting to the emergency department. Being better able to account for socioeconomic disparities may help better predict outcomes in DR. Policymakers can use these data to develop programs that help those at the highest risk for losing vision as a result of DR.

Methods

The Howard University Institutional Review Board ruled that approval was not required for this study. Informed consent was not plausible given the retrospective nature of this study. The study adhered to the tenets of the Declaration of Helsinki.

The California State Emergency Department Database was queried from January 2018 to December 2020 to identify patients discharged with a primary diagnosis of DR based on the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes. The State Emergency Department Database is part of a set of databases developed for the Healthcare Cost and Utilization Project to inform policymaking at the local and national levels. The State Emergency Department Database captures patient discharge information for all emergency department visits by state. The composition and completeness of data vary by state. California is populous and diverse, socioeconomically and culturally, providing a robust and rich dataset. In addition, the California State Emergency Department Database included data on race. Eligible patients were aggregated by ZIP codes to calculate their socioeconomic distress levels using the Distressed Communities Index.

The Economic Innovation Group constructed the Distressed Communities Index to serve as a tool for measuring the comparative economic well-being of US communities. The Distressed Communities Index is derived from US Census data and the American Community Survey 5-Y Estimates for 2016 through 2020 and captures more than 99% of the US population.

The Distressed Communities Index creates a composite ranking by ZIP code based on the following 7 socioeconomic metrics: (1) the proportion of the population aged 25 years or older without a high school diploma or equivalent, (2) the ratio of housing units that are vacant after adjustment for recreational, seasonal, or occasional-use vacancies, (3) the proportion of the population aged 25 to 54 years who are not working (unemployed or not in the labor force), (4) the proportion of residents living below the federal poverty rate, (5) the median household income as a percentage of a metro area or the state median household income, (6) changes in the number of employees working in the area, and (7) the number of business establishments in the ZIP code. These evenly weighted variables are used to calculate a ZIP code’s rank compared with that of its geographic peers and then are normalized to obtain distress scores from 0 (no distress) to 100 (severe distress). These are stratified into quintiles of well-being as follows: prosperous, comfortable, mid-tier, at-risk, and distressed. Patients who did not have complete information to derive quintiles were excluded.

The primary outcome was emergency department visits for DR. Patients were stratified into those with nonproliferative DR and those with proliferative DR (PDR). Covariates included age, sex, race, insurance type, median household income (Quartile I: $1–$49 999; Quartile II: $50 000–$64 999; Quartile III: $65 000–$85 999; Quartile IV: $86 000+), and preexisting medical conditions (obesity, systemic hypertension).

Data were analyzed using Stata software (version 14, StataCorp). Descriptive statistics, such as frequencies and percentages, were used to describe patient demographics and socioeconomic variables. The χ² test was used to assess the association between categorical variables with the Distressed Communities Index and the relationship between Distressed Communities Index quintiles and emergency department visits for DR. Multivariate logistic regression analyses were used to estimate the relationship between the Distressed Communities Index and DR, controlling for covariates. Results are reported using an adjusted odd ratio (aOR), a 95% CI, and a statistically significant value of P < .05.

Results

There were 2 725 195 emergency department visits for patients with diabetes in the California State Emergency Department Database, of which 2 459 577 (90.3%) were available in the Distressed Communities Index during the study period. Stratification by economic poverty level (measured as Distressed Communities Index quintiles) showed significant socioeconomic disparities and variance in medical comorbidities (Supplemental Table 1). Notably, age, race, sex, and insurance coverage all varied by Distressed Communities Index quintile (P ≤ .001). Patients from prosperous neighborhoods had the oldest median age and were the most likely to be male, to be White, have Medicare, and earn the highest income. Patients from distressed neighborhoods had the lowest median age and were the most likely to be female, to be Hispanic, have Medicaid, and earn the lowest income.

A total of 39 693 patients were discharged from the emergency department with a primary diagnosis of DR, representing 1.47% of all emergency department visits for patients with diabetes. Patients’ age, sex, race, insurance status, income, and the Distressed Communities Index quintile varied by DR severity on bivariate analysis (P ≤ .001) (Table 1). Those with PDR were more likely to be younger, male, and Hispanic or Black; have Medicaid; earn the lowest income; and be in the distressed (ie, lowest) Distressed Communities Index quintile.

Table 1.

Sociodemographic Characteristic Stratified by Type of DR.

Characteristic	Total Population (N = 39 693)	Nonproliferative DR (n = 26 076)	Proliferative DR (n = 13 617)	P Value
Age (y)				.001
Median	62	73	65
IQR	51-73	64-80	56-74
Female	19 860 (50.0)	13 283 (50.9)	6577 (48.3)	<.001
Race/ethnicity, n (%)				<.001
White	11 394 (28.8)	8722 (33.5)	2672 (19.6)
Black	7494 (18.9)	4874 (18.7)	2620 (19.3)
Hispanic	13 297 (33.6)	7288 (28.0)	6009 (44.2)
AAPI	6748 (17.0)	4656 (17.9)	2092 (15.4)
Native American	202 (0.5)	133 (0.5)	69 (0.5)
Other	491 (1.2)	350 (1.3)	141 (1.0)
Insurance, n (%)				<.001
Medicare	28 791 (72.5)	19 806 (76.0)	8985 (66.0)
Medicaid	1938 (4.9)	982 (3.8)	956 (7.0)
Private	8612 (21.7)	5095 (19.5)	3517 (25.8)
Uninsured	240 (0.6)	129 (0.5)	111 (0.8)
Other	109 (0.3)	63 (0.2)	46 (0.3)
Income, n (%)				<.001
Quartile I	8882 (22.7)	5393 (21.0)	3489 (25.9)
Quartile II	9302 (23.8)	5882 (22.9)	3420 (25.4)
Quartile III	11 930 (30.5)	7957 (31.0)	3973 (29.5)
Quartile IV	9047 (23.1)	6477 (25.2)	2570 (19.1)
DCI, n (%)				<.001
Prosperous	6318 (17.3)	4534 (18.9)	1784 (14.2)
Comfortable	10 491 (28.6)	7070 (29.4)	3421 (27.1)
Mid-tier	9456 (25.8)	5944 (24.8)	3512 (27.9)
At risk	9273 (25.3)	5844 (24.3)	3429 (27.2)
Distressed	1089 (3.0)	628 (2.6)	461 (3.7)
BMI, n (%)				.27
Underweight	160 (1.8)	115 (1.9)	45 (1.5)
Normal BMI	228 (2.6)	157 (2.6)	71 (2.4)
Overweight	433 (94.9)	281 (4.7)	152 (5.1)
Obese Class I	2091 (23.4)	1356 (22.8)	735 (24.6)
Obese Class II	2844 (31.8)	1901 (32.0)	943 (31.6)
Obese Class III	3175 (35.6)	2132 (35.9)	1043 (34.9)
Hypertension, n (%)	9624 (77.8)	6644 (76.9)	2980 (79.8)	<.001

Abbreviations: AAPI, Asian American/Pacific Islander; BMI, body mass index; DCI, Distressed Communities Index; DR, diabetic retinopathy.

Table 2 shows the multivariate logistic regression results for the association of various social determinants of health with PDR. Age was inversely associated with PDR, with a minimal decrease of 0.04% in odds for each year’s increase (aOR, 0.958; 95% CI, 0.955-0.963; P < .001). Racial/ethnic minorities had higher odds of PDR than Whites: Hispanics (aOR, 2.21; 95% CI, 1.97-2.48; P < .001); Blacks (aOR, 1.46; 95% CI, 1.28-1.67; P < .001). Public insurance holders had higher odds of having PDR than private insurance holders: Medicaid (aOR, 1.37; 95% CI, 1.13-1.65; P = .001); Medicare (aOR,1.34; 95% CI, 1.19-1.51; P < .001). Uninsured (or self-pay) patients made up a small percentage of the study population (3%) and were not found to be more likely to present with PDR. Patients with preexisting hypertension were also at significant risk, with a 48% higher risk for PDR compared with normotensive patients (aOR, 1.48; 95% CI, 1.33-1.65; P < .001). There was no significant association with weekend visits compared with weekday visits (aOR, 1.06; 95% CI, 0.96-1.17; P = .243).

Table 2.

Association Between Neighborhood Economic Poverty Level and Proliferative Diabetic Retinopathy.

Variable	aOR	Lower CI	Upper CI	P Value
Age	0.958	0.955	0.963	<.001
Female	0.94	0.88	1.04	.306
Race/ethnicity
White	Ref	—	—	—
Black	1.46	1.28	1.67	<.001
Hispanic	2.21	1.97	2.48	<.001
AAPI	1.36	1.20	1.56	<.001
Native American	1.85	1.01	3.40	.048
Other	1.27	0.88	1.84	.199
Insurance
Private	Ref	—
Medicare	1.34	1.19	—	—
Medicaid	1.37	1.13	1.51	<.001
Uninsured	1.35	0.80	1.65	.001
Other	0.38	0.14	2.25	.259
DCI
Prosperous	Ref	—	—	—
Comfortable	1.01	0.89	1.15	.863
Mid-tier	1.12	0.98	1.28	.111
At risk	0.96	0.84	1.10	.583
Distressed	1.63	1.20	2.23	.002
Weekend Visit	1.06	0.96	1.17	.243
Hypertension	1.48	1.33	1.65	<.001

Abbreviations: AAPI, Asian American/Pacific Islander; aOR, adjusted odds ratio; DCI, Distressed Communities Index; Ref, reference.

The Distressed Communities Index showed a positive association with the occurrence of PDR (Figure 1). Distressed neighborhoods had a 63% increased risk compared with the richest neighborhoods (aOR, 1.63; 95% CI, 1.20-2.23; P = .002).

Figure 1.

Forest plot showing the independent association between the Distressed Communities Index and presenting to the emergency department for proliferative diabetic retinopathy. The vertical line represents the reference group (residents of prosperous neighborhoods). “Overall, IV” shows the combined effect of all groups, calculated using the inverse variance method. The I² statistic measures the proportion of variability between different Distressed Communities Index quintiles. I ²  = 61.4% and a P value of 0.051 suggest no statistically significant heterogeneity among the quintiles. As the Distressed Communities Index category worsens from prosperous to distressed, the odds increase, with the distressed group having the highest risk.

Abbreviation: OR, odds ratio.

Conclusions

In this study, we explored the Distressed Communities Index as a tool for understanding the influence of 7 socioeconomic factors on DR. Using the California State Emergency Department Database, the largest state emergency department database available in the US, we found that the Distressed Communities Index predicted the risk for presenting to the emergency department for PDR, independent of demographic factors and insurance status. Those from distressed or the most disadvantaged neighborhoods had the highest risk. This study confirms the utility of the Distressed Communities Index in predicting important health outcomes because it has successfully predicted mortality after cardiac surgery, ¹² COVID-19, ¹³ limb amputation rates, ¹⁴ emergency department visits for firearm injuries, ¹⁵ and HIV prevalence among hospital admissions. ¹⁶

Although other ophthalmic studies have evaluated state and national emergency department databases, ¹⁷ to our knowledge this is the first study to present detailed data on emergency department presentation primarily for DR. As seen in outpatient and survey-based studies,^{2 –4,18} we found race to be predictive of a more advanced presentation of DR, with minority groups being more likely to present to the emergency department for PDR than Whites. As a corollary to previous studies that identified patients with public insurance having lower follow-up rates for PDR after emergency department visits, ¹⁹ we found that patients with public insurance were more likely to present to the emergency department for PDR than those with private insurance. This study builds on previous literature that identified hypertension with the progression of DR^20,21 by finding hypertension to be a risk factor for emergency department presentation for more advanced DR and in a more diverse patient population. In this study, younger age was a risk factor for emergency department presentation for PDR and is reflected in the lower proportion of Medicare patients and the higher proportion of Medicaid patients in the PDR group. This finding could, in part, be explained by the higher risk for PDR when DM is diagnosed at a younger age.^22,23 The number of uninsured patients in this study was relatively small. This may have limited the ability to detect differences in presentation based on DR severity among uninsured patients.

Large database studies have inherent strengths and limitations. The State Emergency Department Database provides the most extensive and representative data on emergency department visits by state and is rich in patient-level demographic data. It depends on the accuracy of ICD-10-CM coding and does not record ophthalmic examination data. It does not provide data on the duration of DM or glycemic control. Therefore, controlling for many medical comorbidities was not possible in this study, as has been the case in other large, cross-sectional, and population-based studies of socioeconomic factors and eye diseases.^24,25 Future State Emergency Department Database studies may be expanded to include other states or national data to confirm the generalizability of our findings. The Distressed Communities Index is a comprehensive socioeconomic metric based on ZIP code. This limits its ability to accurately estimate each patient’s specific risk for a given outcome. In exchange, it allows for the inclusion of neighborhood-level factors that often cannot be captured when data are recorded on the individual level only. Although the Distressed Communities Index includes a broad range of metrics, there may be unaccounted-for socioeconomic factors that might also influence health outcomes.

Effective healthcare requires more than meeting the clinical needs of patients. It requires addressing patients’ social and economic conditions and understanding the environmental determinants of health. This study improves our understanding of how socioeconomic deprivation, measured by the Distressed Communities Index, can increase the chances of presenting to the emergency department for more advanced forms of DR. In addition to raising awareness of this neighborhood-level inequity, this study deepens our understanding of risk factors for DR to aid policymakers in developing actionable strategies to improve vision-related outcomes for DR and work toward eliminating vision health disparities. Neighborhood-level data, as analyzed in our study, can better inform decisions on where best to invest resources and develop programs to promote preventive measures for diabetes, DR screening, and interventions to improve access to diabetic care. Further studies are needed to determine how tools such as the Distressed Communities Index can be incorporated into risk calculators to better predict advanced forms of DR and how improving socioeconomic disparities can yield better outcomes in DR.