Abstract
Objective or Purpose:
To evaluate the association of social determinants of health (SDoH) with eye care utilization among people with diabetes mellitus using the 2013–2017 National Health Interview Survey (NHIS).
Design:
Retrospective cross-sectional study
Subjects, Participants, and/or Controls:
Participants ≥18 years of age with self-reported diabetes
Methods, Intervention, and/or Testing:
SDoH in the following domains were used: 1) economic stability, 2) neighborhood, physical environment, and social cohesion, 3) community and social context, 4) food environment, 5) education, and 6) health care system. An aggregate SDoH score was calculated and divided into quartiles with Q4 representing those with the highest adverse SDoH burden. Survey-weighted multivariable logistic regression models evaluated the association of SDoH quartile with eye care utilization in the preceding 12 months. A linear trend test was conducted. Domain specific mean SDoH scores were calculated, and the performance of domain specific models were compared using area under the curve (AUC).
Main Outcome Measure:
Eye care utilization in the preceding 12 months
Results
Of 20,807 adults with diabetes, 43% had not utilized eye care. Greater adverse SDoH burden was associated with decrements in odds of eye care utilization (p<0.001 for trend). Participants in the highest quartile of adverse SDoH burden (Q4), had 59% lower odds (OR 0.41, 95% CI 0.37–0.46) of eye care utilization than those in Q1. The domain specific model using economic stability had the highest performing AUC (0.63, 95% CI 0.62 to 0.64).
Conclusions
Among a national sample of people with diabetes, adverse SDoH were associated with decreased eye care utilization. Evaluating and intervening upon the effects of adverse SDoH may be a means by which to improve eye care utilization and prevent vision loss.
Precis
Adverse social determinants of health across multiple domains were associated with decreased eye care utilization among a nationally representative sample of people with diabetes mellitus.
INTRODUCTION
Social determinants of health (SDoH) are the conditions in which people live, work, and play and the wider set of social structures and economic systems that shape the conditions of daily living.1 Several frameworks exist that organizes SDoH into various domains.1,4,5 The Kaiser Family Foundation is one of these frameworks and divides SDoH into domains of economic stability, neighborhood, physical environment, and social cohesion, community and social context, food environment, education, and health care system.5 Mounting evidence indicates that SDoH affect health outcomes, possibly to a greater degree than medical care.2,3 Diabetic retinopathy, the ophthalmic manifestation of diabetes mellitus, is no exception to the impacts of SDoH. Studies have demonstrated the role of employment status, educational attainment, and income in the prevalence of diabetic retinopathy.6,7 The same adverse SDoH that are associated with more prevalent diabetic retinopathy have also been associated with underutilization of eye care services.8–10 Underutilization of eye care is a major risk factor for vision loss from diabetic retinopathy as it can lead to missed opportunities for the early detection and prompt intervention of vision-threatening complications.10,11 The American Academy of Ophthalmology recommends, at a minimum, annual eye examinations for the screening and follow-up of diabetic retinopathy.12 Despite these national guidelines, eye care utilization among patients with diabetes ranges anywhere between 15% to 77%.13–15
There is a paucity of ophthalmic literature that evaluates a multidimensional set of SDoH in a nationally representative sample. Existing studies either only focus on a specific population or includes a limited set of SDoH.6,10,16 Understanding the most impactful SDoH can help guide further interventions to address SDoH on the national level. The purpose of this study was to evaluate the association of a comprehensive set of SDoH using the Kaiser Family Foundation framework with eye care utilization among a nationally representative sample of adults with diabetes mellitus.
METHODS
Study Design / Data Source
This was a retrospective study of pooled data from the 2013–2017 National Health Interview Survey (NHIS). NHIS is a cross-sectional household interview survey conducted by the National Center for Health Statistics, part of the Centers for Disease Control and Prevention, of the civilian noninstitutionalized people residing in the 50 states and in the District of Columbia.17 The multistage area probability sampling of this survey adjusts for nonresponse and allows for national representativeness. Adults ≥18 years of age with self-reported diabetes during the study period were included. (Supplemental Material) All methods adhered to the declaration of Helsinki. The study was considered exempt by the Johns Hopkins University Institutional Review Board, as the data are publicly available and de-identified.
Variables
Social Determinants of Health
We used the SDoH framework as outlined by the Kaiser Family Foundation.5 A list of 55 questions were used to represent 33 SDoH variables in the following 6 domains: 1) economic stability, 2) neighborhood, physical environment, and social cohesion, 3) community and social context, 4) food environment, 5) education, and 6) health care system.18 Each variable was dichotomized for analysis and the higher value assigned to represent the more adverse SDoH (e.g., uninsured as compared to insured). (Supplemental Material) The aggregate SDoH score across all years of the survey was summed and divided into quartiles with the lowest quartile (Q1) representing the participants with the least adverse SDoH burden and Q4 with the highest adverse SDoH burden. We also calculated a domain specific SDoH score as the mean SDoH score within each of the 6 domains.
Eye Care Utilization
The outcome was self-reported eye care utilization in the preceding 12 months. Participants were considered to have had eye care utilization if they responded “Yes” to the following question “During the past 12 months, have you seen or talked to an optometrist, ophthalmologist, or eye doctor (someone who prescribes eyeglasses) about your own health?”16 (Supplemental Material)
Statistical Analysis
Survey weights provided by NHIS were employed in all analyses to account for the complex design of the NHIS. Summary statistics were used to describe participant characteristics stratified by eye care utilization status. Pearson’s χ2 test was used to compare categorical variables. The analyses were divided into three parts focused on: 1) individual SDoH variables, 2) adverse SDoH quartile with sensitivity analyses, and 3) domain specific SDoH scores.
Part 1:
The association of each SDoH variable with eye care utilization was calculated using logistic regression in univariable models, and multivariable models adjusted for race/ethnicity (non-Hispanic White, non-Hispanic Black, non-Hispanic Asian, or Hispanic), age (18–44, 45–64, or 65+); sex (male or female), geographic region (Northeast, Midwest, South, or West), Charlson Comorbidity index (CCI)), vision impairment (yes or no), and year of the survey.19 (Supplemental Material) The CCI is a composite measure representing the overall health of the patient with a higher scores indicating sicker patients.19 Missing responses for most SDoH variables were minimal, ranging from <1% to 9% of unweighted respondents, with the exception of one variable on the inability to pay for medical bills that had a 81% missing response.
Part 2:
The association of the adverse SDoH quartile with eye care utilization was analyzed in a similar fashion in univariable and multivariable models. A linear trend test was conducted by adjusted estimates after estimation to evaluate the linear relationship between increasing adverse SDoH quartile and eye care utilization. An F-test was performed to evaluate the significance of an interaction term between race/ethnicity and adverse SDoH quartile. Sensitivity analyses were performed on the multivariable models to stratify rather than adjust by race/ethnicity. This was done to avoid the inappropriate operationalization of race/ethnicity as proxies for socioeconomic status. Additional sensitivity analyses were performed on the multivariable models by representing aggregate SDoH score as a continuous variable centered at the sample mean, rather than quartiles, and with an interaction term between race/ethnicity and aggregate SDoH score. Sub-analyses were performed to evaluate the association between adverse SDoH quartile and eye care utilization in multivariable models among patients with Type 1 diabetes and those with Type 2 diabetes. (Supplemental Material)
part 3:
Summary statistics were provided to describe the domain specific SDoH scores. A set of separate multivariable logistic regression models were constructed to evaluate the association between each of the 6 domain specific SDoH scores with eye care utilization. The performance of each model was evaluated using ten-fold crossvalidation to generate the mean area under the receiver operating curve (AUC) and its associated 95% Confidence Interval (CI). The AUC of each domain model was compared to the AUC of the base model with only covariates using a nonparametric approach that utilizes the theory developed for generalized U-statistics.20 The resulting test statistic follows an asymptotic chi-squared distribution (χ2 test).20 Sensitivity analyses were performed to evaluate the domain specific models stratified by race/ethnicity. Sub-analyses were also performed among patients with Type 1 diabetes and those with Type 2 diabetes. All analyses were performed using Stata (StataCorp, College Station, TX, USA, Stata/SE Version 16.1).
RESULTS
A total of 20,807 individuals representing 27.5 million adults with diabetes mellitus were included in the study. (Table 1) Most participants were older than 45 years, and there was a roughly equal distribution of males and females. By race/ethnicity, 64% were non-Hispanic White, 16% non-Hispanic Black, 5% non-Hispanic Asian, and 17% Hispanic. In the lowest SDoH quartile (representing the group with the least adverse SDoH burden), there were more non-Hispanic White (75%) participants as compared to the overall sample. In the highest quartile (representing the group with the highest adverse SDoH burden), there were more non-Hispanic Black (20%) and Hispanic (30%) participants as compared to the overall sample. (Supplemental Table 2) About half the participants did not utilize eye care in the preceding 12 months. (Table 1)
Table 1:
Demographic characteristics of participants in the 2013–2017 National Health Interview Survey (NHIS) with self-reported diabetes mellitus stratified by self-reported eye care utilization in the preceding 12 months.
| Overall | No eye care utilization | Eye care utilization | p value | |
|---|---|---|---|---|
| N (Weighted %) | N (Weighted %) | N (Weighted %) | ||
| Sample (N) | 20807 | 8762 | 12045 | |
| Weighted sample, (weighted %) | 27467169 (100) | 11755196 (43) | 15711973 (57) | |
| Age category | <0.001 | |||
| 18 to 44 | 2469 (14) | 1450 (20) | 1019 (10) | |
| 45 to 64 | 8466 (45) | 3910 (48) | 4556 (43) | |
| 65 and older | 9872 (41) | 3402 (32) | 6470 (47) | |
| Sex | 0.650 | |||
| Male | 9679 (51) | 4113 (51) | 5566 (50) | |
| Female | 11128 (49) | 4649 (49) | 6479 (50) | |
| Race/ethnicity | <0.001 | |||
| Non-Hispanic White | 12466 (64) | 4877 (60) | 7589 (67) | |
| Non-Hispanic Black | 3707 (16) | 1663 (16) | 2044 (15) | |
| Non-Hispanic Asian | 585 (5) | 217 (3) | 368 (4) | |
| Hispanic | 3181 (17) | 1612 (21) | 1569 (14) | |
| Region | <0.001 | |||
| Northeast | 3364 (17) | 1245 (15) | 2119 (19) | |
| Midwest | 4356 (22) | 1693 (21) | 2663 (23) | |
| South | 8148 (40) | 3658 (41) | 4490 (38) | |
| West | 4939 (21) | 2166 (23) | 2773 (20) | |
| Comorbidities | <0.001 | |||
| CCI = 1 | 11393 (58) | 5170 (62) | 6223 (54) | |
| CCI ≥ 2 | 9414 (42) | 3592 (38) | 5822 (46) | |
| Vision impairment | <0.001 | |||
| No | 16556 (81) | 7119 (83) | 9437 (79) | |
| Yes | 4245 (19) | 1639 (17) | 2606 (21) | |
| Survey Year | 0.004 | |||
| 2013 | 4143 (19) | 1851 (19) | 2292 (18) | |
| 2014 | 4458 (19) | 1995 (20) | 2463 (18) | |
| 2015 | 4196 (19) | 1797 (20) | 2399 (19) | |
| 2016 | 4407 (21) | 1699 (20) | 2708 (22) | |
| 2017 | 3603 (22) | 1420 (21) | 2183 (23) |
Abbreviations: CCI = Charlson Comorbidity Index
Adverse SDoH in all six domains were associated with lack of eye care utilization in the preceding 12 months. (Table 3) Greater adverse SDoH burden, as represented by the adverse SDoH quartile, was associated with decrements in odds of eye care utilization (p<0.001 for trend). (Table 4) In multivariable analysis, participants in the highest quartile of adverse SDoH burden (Q4), had an estimated 58% decreased odds (OR 0.42, 95% CI 0.37 to 0.47) of eye care utilization as compared to those with the least adverse SDoH burden (Q1). In stratified analysis, results were consistent across race/ethnic groups. Among non-Hispanic White persons, the highest quartile of adverse SDoH burden was associated with an estimated 57% decrease odds (OR 0.43, 95% CI 0.37 to 0.50) of eye care utilization, 52% (OR 0.48, 95% CI 0.37 to 0.62) among non-Hispanic Black persons, 90% (OR 0.10, 95% CI 0.04 to 0.24) among non-Hispanic Asian persons, and 65% (OR 0.35, 95% CI 0.26 to 0.49) among Hispanic persons. (Table 4) In the sub-analysis, greater adverse SDoH burden was associated with decreased odds of eye care utilization among participants with Type 1 diabetes (representing 7% of the weighted sample) as well as Type 2 diabetes (83%). (Supplemental Table 5)
Table 3:
The association of each social determinants of health (SDoH) variable with eye care utilization in the past 12 months.
| No eye care utilization | Eye care utilization | Unadjusted | Adjusted* | |||
|---|---|---|---|---|---|---|
| N (Weighted %) | N (Weighted %) | OR (%95 CI) | p value | OR (95% CI) | p value | |
| Economic Stability | ||||||
| Never or previously employed | 5491 (58) | 8199 (64) | 1.25 (1.16, 1.35) | <0.001 | 0.85 (0.78, 0.93) | 0.001 |
| No sick leave | 3895 (48) | 4563 (40) | 0.71 (0.65, 0.77) | <0.001 | 0.75 (0.69, 0.81) | <0.001 |
| Low family income | 5548 (62) | 6039 (48) | 0.55 (0.51, 0.6) | <0.001 | 0.54 (0.49, 0.59) | <0.001 |
| Difficulty paying medical bills | 2034 (24) | 1955 (17) | 0.64 (0.58, 0.7) | <0.001 | 0.71 (0.64, 0.79) | <0.001 |
| Unable to pay medical bills | 1214 (59) | 1062 (54) | 0.83 (0.69, 0.99) | 0.035 | 0.83 (0.68, 1.00) | 0.047 |
| High Financial Distress Composite Score | 2551 (32) | 2573 (23) | 0.63 (0.58, 0.69) | <0.001 | 0.74 (0.67, 0.81) | <0.001 |
| Cost-related medication nonadherence | 1353 (19) | 1262 (11) | 0.54 (0.48, 0.61) | <0.001 | 0.61 (0.54, 0.69) | <0.001 |
| Foregone/delayed care due to cost | 1938 (22) | 1669 (15) | 0.60 (0.55, 0.66) | <0.001 | 0.67 (0.61, 0.75) | <0.001 |
| Neighborhood, physical environment, and social cohesion | ||||||
| Renting (vs. owned home) | 3667 (37) | 3858 (27) | 0.62 (0.57, 0.67) | <0.001 | 0.72 (0.66, 0.79) | <0.001 |
| Disagree: people in neighborhood help each other | 1918 (24) | 2186 (18) | 0.73 (0.67, 0.79) | <0.001 | 0.79 (0.72, 0.87) | <0.001 |
| Disagree: people in neighborhood can be trusted | 1867 (23) | 2007 (16) | 0.66 (0.60, 0.72) | <0.001 | 0.74 (0.67, 0.82) | <0.001 |
| Disagree: this is a close-knit neighborhood | 3208 (40) | 4212 (36) | 0.85 (0.79, 0.91) | <0.001 | 0.90 (0.84, 0.97) | 0.008 |
| Disagree: people in neighborhood I can count on | 1820 (23) | 2054 (17) | 0.69 (0.63, 0.76) | <0.001 | 0.78 (0.71, 0.86) | <0.001 |
| Community and social context | ||||||
| Psychological distress (Kessler K6 Scale) | 671 (8) | 693 (6) | 0.70 (0.60, 0.80) | <0.001 | 0.76 (0.64, 0.88) | 0.001 |
| Food Insecurity | ||||||
| Food Insecurity | 1614 (17) | 1502 (11) | 0.58 (0.52, 0.64) | <0.001 | 0.65 (0.58, 0.73) | <0.001 |
| Education | ||||||
| English language proficiency (not well/not at all) | 722 (10) | 672 (6) | 0.60 (0.52, 0.7) | <0.001 | 0.66 (0.55, 0.81) | <0.001 |
| <=High school | 4042 (41) | 4431 (32) | 0.69 (0.64, 0.74) | <0.001 | 0.66 (0.61, 0.72) | <0.001 |
| Health Information Technology use: Did not look up health info on internet | 5780 (64) | 7349 (58) | 0.76 (0.70, .83) | <0.001 | 0.67 (0.61, 0.73) | <0.001 |
| Health Information Technology use: Filled a prescription online | 7879 (91) | 10598 (87) | 0.63 (0.55, 0.71) | <0.001 | 0.60 (0.52, 0.70) | <0.001 |
| Health Information Technology use: Scheduled a healthcare appointment online | 8053 (93) | 11012 (90) | 0.65 (0.55, 0.76) | <0.001 | 0.60 (0.50, 0.71) | <0.001 |
| Health Information Technology use: Communicated with healthcare provider online | 7912 (91) | 10638 (87) | 0.62 (0.54, 0.71) | <0.001 | 0.58 (0.50, 0.67) | <0.001 |
| Health Information Technology use: Used internet chat rooms to learn about health | 8280 (97) | 11644 (96) | 0.79 (0.63, 1.00) | 0.047 | 0.73 (0.57, 0.93) | 0.01 |
| Health care system | ||||||
| Uninsured | 984 (12) | 371 (3) | 0.25 (0.21, 0.29) | <0.001 | 0.33 (0.28, 0.39) | <0.001 |
| No usual source of care | 561 (7) | 221 (2) | 0.27 (0.22, 0.33) | <0.001 | 0.35 (0.28, 0.44) | <0.001 |
| Trouble finding a doctor/provider in past 12 months | 344 (4) | 350 (3) | 0.72 (0.59, 0.88) | 0.002 | 0.77 (0.62, 0.95) | 0.015 |
| MD’s office does not accept you as new patient in past 12 months | 280 (3) | 349 (3) | 0.97 (0.80, 1.17) | 0.726 | 1.06 (0.87, 1.30) | 0.58 |
| MD’s office does not accept your insurance in past 12 months | 341 (4) | 446 (4) | 0.99 (0.82, 1.18) | 0.897 | 1.11 (0.92, 1.35) | 0.266 |
| Delayed medical care: couldn’t get through on phone | 293 (3) | 415 (4) | 1.07 (0.87, 1.32) | 0.513 | 1.15 (0.93, 1.44) | 0.198 |
| Delayed medical care: couldn’t get appointment soon enough | 680 (8) | 989 (9) | 1.03 (0.90, 1.18) | 0.687 | 1.10 (0.95, 1.26) | 0.206 |
| Delayed medical care: wait too long at MD’s office | 582 (7) | 694 (6) | 0.89 (0.76, 1.04) | 0.137 | 0.96 (0.81, 1.13) | 0.632 |
| Delayed medical care: not open when you could go | 293 (4) | 402 (4) | 0.98 (0.78, 1.22) | 0.855 | 1.05 (0.83, 1.33) | 0.665 |
| Delayed medical care: no transportation | 427 (4) | 480 (3) | 0.80 (0.67, 0.95) | 0.011 | 0.80 (0.66, 0.96) | 0.019 |
| Somewhat/very dissatisfied (quality) or no healthcare in past year | 1065 (14) | 734 (6) | 0.42 (0.37, 0.48) | <0.001 | 0.48 (0.42, 0.56) | <0.001 |
Abbreviations: OR = odds ratio; CI = confidence interval
Adjusted by age, sex, race/ethnicity, geographic region, Charlson Comorbidity Index, year of the survey, and vision impairment.
Table 4.
The association of adverse social determinants of health (SDoH) quartile with eye care utilization in the overall model and models stratified by race and ethnicity.
| Race | SDoH quartile | No eye care utilization | Eye care utilization | Unadjusted | Adjusted * | ||
|---|---|---|---|---|---|---|---|
| N (weighted %) | N (weighted %) | OR (95% CI) | p value | OR (95% CI) | p value | ||
| Overall | Ql | 2478 (32) | 4839 (45) | Reference | Reference | ||
| Q2 | 1883 (20) | 2832 (22) | 0.78 (0.70, 0.85) | <0.001 | 0.70 (0.63, 0.77) | <0.001 | |
| Q3 | 2261 (24) | 2574 (20) | 0.57 (0.52, 0.63) | <0.001 | 0.57 (0.51, 0.63) | <0.001 | |
| Q4 | 2140 (24) | 1800 (14) | 0.40 (0.36, 0.45) | <0.001 | 0.42 (0.37, 0.47) | <0.001 | |
| p for trend | <0.001 | <0.001 | |||||
| Non-Hispanic White | Q1 | 1713 (39) | 3572 (51) | Reference | Reference | ||
| Q2 | 1124 (22) | 1844 (23) | 0.81 (0.72, 0.91) | 0.001 | 0.73 (0.64, 0.83) | <0.001 | |
| Q3 | 1108 (21) | 1361 (16) | 0.61 (0.53, 0.69) | <0.001 | 0.59 (0.51, 0.67) | <0.001 | |
| Q4 | 932 (19) | 812 (10) | 0.43 (0.37, 0.5) | <0.001 | 0.44 (0.38, 0.51) | <0.001 | |
| p for trend | <0.001 | <0.001 | |||||
| Non-Hispanic Black | Q1 | 340 (23) | 579 (33) | Reference | Reference | ||
| Q2 | 360 (19) | 458 (20) | 0.72 (0.55, 0.96) | 0.023 | 0.65 (0.49, 0.87) | 0.004 | |
| Q3 | 507 (31) | 549 (27) | 0.60 (0.46, 0.78) | <0.001 | 0.54 (0.41, 0.71) | <0.001 | |
| Q4 | 456 (27) | 458 (20) | 0.52 (0.39, 0.68) | <0.001 | 0.47 (0.35, 0.62) | <0.001 | |
| p for trend | <0.001 | <0.001 | |||||
| Non-Hispanic Asian | Q1 | 94 (42) | 175 (60) | Reference | Reference | ||
| Q2 | 46 (22) | 97 (22) | 0.69 (0.31, 1.56) | 0.376 | 0.46 (0.21, 1.03) | 0.06 | |
| Q3 | 50 (23) | 69 (14) | 0.44 (0.21, 0.94) | 0.034 | 0.37 (0.16, 0.85) | 0.02 | |
| Q4 | 27 (13) | 27 (3) | 0.18 (0.07, 0.51) | 0.001 | 0.12 (0.04, 0.39) | <0.001 | |
| p for trend | 0.001 | 0.001 | |||||
| Hispanic | Q1 | 227 (16) | 325 (25) | Reference | Reference | ||
| Q2 | 286 (17) | 333 (21) | 0.79 (0.53, 1.16) | 0.227 | 0.67 (0.45, 1.01) | 0.055 | |
| Q3 | 483 (29) | 487 (30) | 0.64 (0.45, 0.91) | 0.012 | 0.57 (0.40, 0.82) | 0.003 | |
| Q4 | 616 (38) | 424 (25) | 0.41 (0.29, 0.57) | <0.001 | 0.37 (0.26, 0.53) | <0.001 | |
| p for trend | <0.001 | <0.001 | |||||
Overall model is adjusted by age, sex, geographic region, race/ethnicity, Charlson Comorbidity Index, year of the survey, and vision impairment; stratified models are adjusted by age, sex, geographic region, Charlson Comorbidity Index, year of the survey, and vision impairment.
Abbreviations: OR = odds ratio; CI = confidence interval
The F-test did not show a significant interaction between race/ethnicity and SDoH quartile (p=0.31). However, in sensitivity analyses using aggregate SDoH score as a continuous variable in the multivariable model, there was a statistically significant interaction between race/ethnicity and aggregate SDoH score but only for the non-Hispanic Asian group. (Supplemental Table 6)
The average domain specific SDoH score was 0.35 (standard deviation (SD) 0.23) for economic stability, 0.28 (SD 0.32) for neighborhood, physical environment, and social cohesion, 0.07 (SD 0.25) for community and social context, 0.15 (SD 0.36) for food environment, 0.69 (SD 0.20) for education, and 0.05 (SD 0.10) for health care system. In analyses comparing the domain specific SDoH models, the model with the domain of economic stability had the highest AUC (0.63, 95% CI 0.62 to 0.64) as compared to other domains. (Table 7) In stratified analysis of each of the domain specific models, results were similar across race/ethnic groups. (Table 7) Results were also similar in the sub-analyses among participants with Type 1 diabetes as well as Type 2 diabetes. (Supplemental Table 8)
Table 7:
Comparing the performance of multivariate models using area under the curve (AUC) with the domain specific adverse social determinants of health (SDoH) score, in the overall and stratified by race and ethnicity.
| Race | Model | AUC (95% CI) | p value |
|---|---|---|---|
| Overall | Base model only* | 0.612 (0.604, 0.620) | |
| Base model with economic stability | 0.637 (0.630, 0.645) | <0.001 | |
| Base model with neighborhood, physical environment, and social cohesion | 0.617 (0.609, 0.625) | 0.001 | |
| Base model with community and social context | 0.616 (0.608, 0.624) | 0.005 | |
| Base model with food Insecurity | 0.616 (0.608, 0.624) | 0.002 | |
| Base model with education | 0.626 (0.618, 0.633) | <0.001 | |
| Base model with health care system | 0.617 (0.609, 0.625) | <0.001 | |
| Non-Hispanic White | Base model only* | 0.600 (0.590, 0.610) | |
| Base model with economic stability | 0.632 (0.622, 0.642) | <0.001 | |
| Base model with neighborhood, physical environment, and social cohesion | 0.607 (0.597, 0.617) | 0.001 | |
| Base model with community and social context | 0.604 (0.594, 0.614) | 0.006 | |
| Base model with food Insecurity | 0.608 (0.598, 0.618) | <0.001 | |
| Base model with education | 0.615 (0.605, 0.625) | <0.001 | |
| Base model with health care system | 0.604 (0.594, 0.615) | 0.005 | |
| Non-Hispanic Black | Base model only* | 0.591 (0.573, 0.610) | |
| Base model with economic stability | 0.609 (0.591, 0.627) | 0.003 | |
| Base model with neighborhood, physical environment, and social cohesion | 0.593 (0.575, 0.612) | 0.576 | |
| Base model with community and social context | 0.596 (0.578, 0.615) | 0.811 | |
| Base model with food Insecurity | 0.588 (0.570, 0.607) | 0.508 | |
| Base model with education | 0.607 (0.589, 0.625) | 0.001 | |
| Base model with health care system | 0.595 (0.577, 0.613) | 0.125 | |
| Non-Hispanic Asian | Base model only* | 0.607 (0.559, 0.655) | |
| Base model with economic stability | 0.628 (0.581, 0.675) | 0.203 | |
| Base model with neighborhood, physical environment, and social cohesion | 0.626 (0.579, 0.674) | 0.196 | |
| Base model with community and social context | 0.615 (0.567, 0.663) | 0.290 | |
| Base model with food Insecurity | 0.607 (0.559, 0.655) | 0.941 | |
| Base model with education | 0.630 (0.583, 0.677) | 0.154 | |
| Base model with health care system | 0.614 (0.566, 0.662) | 0.580 | |
| Hispanic | Base model only* | 0.640 (0.621, 0.659) | |
| Base model with economic stability | 0.663 (0.644, 0.682) | <0.001 | |
| Base model with neighborhood, physical environment, and social cohesion | 0.642 (0.623, 0.661) | 0.499 | |
| Base model with community and social context | 0.642 (0.623, 0.661) | 0.912 | |
| Base model with food Insecurity | 0.642 (0.623, 0.661) | 0.320 | |
| Base model with education | 0.650 (0.631, 0.669) | 0.025 | |
| Base model with health care system | 0.654 (0.636, 0.673) | <0.001 |
Abbreviations: AUC = area under the curve; CI = confidence interval; SDoH = social determinants of health
Overall model includes age, sex, geographic region, race/ethnicity, Charlson Comorbidity Index, year of the survey, and vision impairment; stratified models include age, sex, geographic region, Charlson Comorbidity Index, year of the survey, and vision impairment.
DISCUSSION
This study of multidimensional SDoH in a nationally representative sample of people with diabetes demonstrates the substantial impact of SDoH on eye care utilization. Increasing burden of adverse SDoH was associated with lower eye care utilization, a finding that was consistent across race and ethnic groups. Comparing between the SDoH domains, economic stability had a stronger association with eye care utilization as compared to the domains of neighborhood, physical environment, and social cohesion; community and social context; food environment; education; and health care system. These associations suggest possible areas of focus for screening or interventions to promote eye care utilization among people with diabetes.
We found an association between underutilization of eye care and adverse SDoH in all domains of the Kaiser Family Foundation framework. These adverse determinants have been previously linked either underutilization of health care or poor health outcomes. In our study, the neighborhood, physical environment, social support, and social cohesion domain included questions on housing, and social cohesion within the neighborhood. Poor housing conditions has been associated with lower odds of eye care utilization among a sample of Washington, D.C. Medicaid beneficiaries with diabetes.10 Studies in the broader diabetes literature suggest that less social support can be associated with worse glycemic control.4 The domain of community and social context included questions representing psychological distress. Psychological distress has been linked to poor outcomes including all-cause mortality, cardiovascular disease, and cancer mortality among adults with and without diabetes.21 The domain of food environment included questions on hunger and food insecurity. Food insecurity is a risk factor for poor diabetes management and adverse health outcomes as people might resort to more affordable energy-dense foods that directly raise serum glucose or divert necessary financial resources to purchase food.4,22 The domain of education included questions on language, educational attainment, and use of health information technology. Lower educational attainment has been associated with poor vision outcomes and visual difficulties across multiple studies.23 Lastly, the domain of health care system included questions on health coverage, provider availability, and quality of care. Health coverage, in the form of insurance, is a major risk factor for vision outcomes and eye care utilization.23,24 Interestingly, the health care system domain included the greatest number of questions that were not significantly associated with eye care utilization, particularly some of the ones centered around provider availability. It could be that these questions reflected the availability of the individual’s primary care or other providers and were not specific to their eye care provider.
In comparing AUCs from different SDoH domains, we found that the model with the domain specific SDoH score for economic stability outperformed models that included other domains. The domain of economic stability included survey questions on employment status, family income, medical bills, overall financial distress, and cost-related medication underuse. Numerous studies have demonstrated an association between lower family income, difficulty paying medical bills, and cost-related medication underuse with worse health outcomes and underutilization of health care, including eye care.23,25,26 It could be that economic stability, as represented by this specific set of questions, operates slightly upstream of the other SDoH domains. One’s family income and employment status certainly impacts quality of housing and physical environment (represented in the domains of neighborhood, physical environment, and social cohesion), type of neighborhood that one lives in (community and social context), food security (food environment), use of health information technology (education), and health coverage (health care system).
The negative impact of aggregate adverse SDoH burden on eye care utilization was similar across race and ethnic groups. We did not find a differential association of adverse SDoH by race and ethnicity in our primary analysis but did detect one in the sensitivity analysis.27 We chose to represent aggregate adverse SDoH burden using quartiles in the primary analysis because it is a way to assess the general impact of SDoH while accounting for the interactions between factors.18,28,29 There are some limitations to representing SDoH in this fashion. It assumes that each SDoH included in the score has an equal impact which may not be the case. However, despite the limitations, this metric has been used previously and has been associated with other negative health outcomes including stroke, vaccination, and hospitalization among others.28–31 There could be several reasons for not finding a difference in the interaction of race and ethnicity with adverse SDoH quartile. Since the power needed to detect statistically significant effect modification is higher than what is needed for a pairwise comparison, our sample could be underpowered to detect such a difference.32 It could be that specific adverse SDoH have a differential impact by race and ethnicity, for example annual family income, while others do not.27 By grouping SDoH together and representing it in quartiles for analysis, we could be reducing our ability to detect specific differences. The significant finding of aggregate adverse SDoH score, represented as a continuous variable, with race and ethnicity in the sensitivity analysis was entirely driven by the non-Hispanic Asian group. This group had very few respondents, particularly in the higher quartiles (with only 54 in quartile 4) which could be contributing to measurement error. This analysis calls attention to the need to further investigate the impact of adverse SDoH on eye care utilization among the non-Hispanic Asian community.
The underutilization of eye care with greater adverse SDoH burden impacted both patients with Type 1 and Type 2 diabetes. Although Type 1 and Type 2 diabetes have distinct pathophysiology and affect different patient populations, having more adverse SDoH appears to be a barrier to eye care in both groups. This is perhaps not surprising as adverse SDoH has been shown to negatively impact care across multiple diseases.40 SDoH also impacts many aspects of diabetes care and not just vision health.4,9 It is possible that specific patient populations and types of health care are more sensitive to certain SDoH than others, for example having someone accompany patients to office visits appears to be critical among those who have vision impairment.41 However, because SDoH are interconnected, much more work is needed to isolate the effects of a given adverse SDoH independent of other SDoH.
This work emphasizes the interconnectedness of SDoH and their impact on health outcomes. Screening and addressing the impacts of adverse SDoH should consider the multidimensionality of these determinants.10,33 Effective screening should be a balance between feasibility of the survey but simultaneously assessing multiple domains. Our work suggests that questions in each of the domains evaluated would be appropriate for such a screening tool. Although many standardized multidimensional SDoH collection tools exist, some customizations and additions are likely needed to tailor the screening to the population and health outcome of interest.34,35 For example, quality of care was associated with eye care utilization but is often not represented in standard multidimensional evaluations of SDoH. Other studies have shown the association of diabetic retinopathy education and knowledge with eye care utilization, but again, most standard SDoH questionnaires do not contain ophthalmic-specific data.26,36,37 More work should be done to construct the ideal screening SDoH tool that is appropriate for ophthalmic use. Ultimately, addressing adverse SDoH and connecting patients to needed services could be a means by which to improve eye care utilization among people with diabetes and prevent vision loss.
This study has several limitations. Due to the nature of the NHIS, the study relies on self-reported data. Although self-reported eye care utilization has been used in previous studies, the degree to which self-reported eye care utilization corresponds to actual utilization has not been directly studied. It is also unknown whether self-reported eye care utilization corresponds to receipt of a diabetic eye examination as the NHIS survey question includes glasses prescription as an indication for eye care utilization. Additionally, the recommended frequency of diabetic eye examinations depends on the severity of eye disease.12,38 We do not know if the self-reported utilization of eye care within the past 12 months is appropriate for that individual’s level of eye disease, since that level of granularity is not included in this dataset. The definition of appropriate utilization used for this study reflects the most general recommendation of annual eye examinations.12 Lastly, we were constrained by the survey questions used by the NHIS. Some of the questions used to represent specific SDoH variables might not be the most commonly used across studies. For example, the National Academy of Medicine, formerly known as the Institute of Medicine, recommends a single question to assess for food insecurity that asks respondents the amount of food the household has to eat.39 For this study, a series of 10 questions were used to assess for food insecurity that also included details such as the number of days the respondent had to skip meals or not eaten for a whole day. Although both these sets of questions represent the concept of food insecurity, the use of distinct questions might generate artificial differences that make it challenging to compare between studies. Despite these limitations, this is among the few studies to examine the association of multidimensional SDoH and eye care utilization among a national sample, thus maximizing generalizability.
In conclusion, increasing burden of adverse SDoH has a progressively negative association with eye care utilization in a nationally representative sample of adults with diabetes mellitus. Evaluating and addressing multidimensional adverse SDoH could be a means by which to improve eye care utilization and prevent vision loss in this population.
Supplementary Material
Financial Support
This project was supported by a Career Development Award from the Research to Prevent Blindness (CXC), National Center for Advancing Translational Sciences funded Clinical and Translational Science Award (Grant KL2TR003099 CXC, and UL1-TR003098 JW), K23 award from the NIH/NEI (award number K23EY033440) (CXC), K24 award from the NIH/NIDDK (award number K24HL148181) (DCC), and an unrestricted grant from Research to Prevent Blindness (Wilmer Eye Institute). Dr. Cai is the Jonathan and Marcia Javitt Rising Professor of Ophthalmology.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Conflicts of Interest
None of the authors
REFERENCES
- 1.World Health Organization. A conceptual framework for action on the social determinants of health. Published 2010. Accessed January 1, 2022. https://www.who.int/sdhconference/resources/ConceptualframeworkforactiononSDH_eng.pdf
- 2.Nwanyanwu KH, Newman-Casey PA, Gardner TW, Lim JI. Beyond HbA1c: Environmental Risk Factors for Diabetic Retinopathy. Journal of Clinical & Experimental Ophthalmology. 2015;6(2):1–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Marmot M Social determinants of health inequalities. The Lancet. 2005;365(9464):1099–1104. [DOI] [PubMed] [Google Scholar]
- 4.Hill-Briggs F, Adler NE, Berkowitz SA, et al. Social Determinants of Health and Diabetes: A Scientific Review. Diabetes Care. 2020;44(1):258–279. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Artiga S, Hinton E. Beyond Health Care: The Role of Social Determinants in Promoting Health and Health Equity. Published 2005. Accessed March 25, 2021. https://www.kff.org/racial-equity-and-health-policy/issue-brief/beyond-health-care-the-role-of-social-determinants-in-promoting-health-and-health-equity/
- 6.Silverberg EL, Sterling TW, Williams TH, Castro G, de la Vega PR, Barengo NC. The Association between Social Determinants of Health and Self-Reported Diabetic Retinopathy: An Exploratory Analysis. International Journal of Environmental Research and Public Health. 2021;18(2):792. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Nwanyanwu KH, Andoh J, Chen E, Xu Y, Deng Y. Social Determinants of Health in Diabetic Retinopathy in the US: Evidence from the National Health and Nutrition Examination Survey (2005–2008). Investigative Ophthalmology & Visual Science. 62:1135. [Google Scholar]
- 8.Murchison AP, Hark L, Pizzi LT, et al. Non-adherence to eye care in people with diabetes. BMJ Open Diabetes Research & Care. 2017;5(1):e000333–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Chen R, Cheadle A, Johnson D, Duran B. US Trends in Receipt of Appropriate Diabetes Clinical and Self-care From 2001 to 2010 and Racial/Ethnic Disparities in Care. The Diabetes Educator. 2014;40(6):756–766. [DOI] [PubMed] [Google Scholar]
- 10.Cai CX, Li Y, Zeger SL, McCarthy ML. Social determinants of health impacting adherence to diabetic retinopathy examinations. BMJ Open Diabetes Research & Care. 2021;9(1):e002374. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Ferris FL. Results of 20 Years of Research on the Treatment of Diabetic Retinopathy. Preventive Medicine. 1994;23(5):740–742. [DOI] [PubMed] [Google Scholar]
- 12.Flaxel CJ, Adelman RA, Bailey ST, et al. Diabetic Retinopathy Preferred Practice Pattern®. Ophthalmology. 2020;127(1):P66–P145. [DOI] [PubMed] [Google Scholar]
- 13.Sloan FA, Yashkin AP, Chen Y. Gaps in Receipt of Regular Eye Examinations among Medicare Beneficiaries Diagnosed with Diabetes or Chronic Eye Diseases. Ophthalmology. 2014;121(12):2452–2460. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Lee PP, Feldman ZW, Ostermann J, Brown DS, Sloan FA. Longitudinal rates of annual eye examinations of persons with diabetes and chronic eye diseases. Ophthalmology. 2003;110(10):1952–1959. [DOI] [PubMed] [Google Scholar]
- 15.Benoit SR, Swenor B, Geiss LS, Gregg EW, Saaddine JB. Eye Care Utilization Among Insured People With Diabetes in the U.S., 2010–2014. Diabetes Care. 2019;42(3):427–433. [DOI] [PubMed] [Google Scholar]
- 16.Zhang X, Beckles GL, Chou CF, et al. Socioeconomic Disparity in Use of Eye Care Services Among US Adults With Age-Related Eye Diseases: National Health Interview Survey, 2002 and 2008. JAMA Ophthalmology. 2013;131(9):1198–1206. [DOI] [PubMed] [Google Scholar]
- 17.About the National Health Interview Survey. Accessed December 12, 2022. https://www.cdc.gov/nchs/nhis/about_nhis.htm
- 18.Javed Z, Valero-Elizondo J, Maqsood MH, et al. Social determinants of health and obesity: Findings from a national study of US adults. Obesity. 2022;30(2):491–502. [DOI] [PubMed] [Google Scholar]
- 19.Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. Journal of Chronic Diseases. 1987;40(5):373–383. [DOI] [PubMed] [Google Scholar]
- 20.DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the Areas under Two or More Correlated Receiver Operating Characteristic Curves: A Nonparametric Approach. Biometrics. 1988;44(3):837. [PubMed] [Google Scholar]
- 21.Huang W, Aune D, Ferrari G, et al. Psychological Distress and All-Cause, Cardiovascular Disease, Cancer Mortality Among Adults with and without Diabetes. Clinical Epidemiology. 2021;13:555–565. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Berkowitz SA, Seligman HK, Choudhry NK. Treat or Eat: Food Insecurity, Cost-related Medication Underuse, and Unmet Needs. The American Journal of Medicine. 2014;127(4):303–310.e3. [DOI] [PubMed] [Google Scholar]
- 23.Su NH, Moxon NR, Wang A, French DD. Associations of Social Determinants of Health and Self-Reported Visual Difficulty: Analysis of the 2016 National Health Interview Survey. Ophthalmic Epidemiology. 2019;27(2):1–5. [DOI] [PubMed] [Google Scholar]
- 24.Gibson DM. Eye Care Availability and Access Among Individuals with Diabetes, Diabetic Retinopathy, or Age-Related Macular Degeneration. JAMA Ophthalmology. 2014;132(4):471–477. [DOI] [PubMed] [Google Scholar]
- 25.Hom GL, Cwalina TB, Jella TK, Singh RP. Assessing financial insecurity among common eye conditions: a 2016–2017 national health survey study. Eye. Published online 2021:1–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Brechner RJ, Cowie CC, Howie LJ, Herman WH, Will JC, Harris MI. Ophthalmic Examination Among Adults with Diagnosed Diabetes Mellitus. JAMA. 1993;270(14):1714–1718. [PubMed] [Google Scholar]
- 27.Crews DC, Charles RF, Evans MK, Zonderman AB, Powe NR. Poverty, Race, and CKD in a Racially and Socioeconomically Diverse Urban Population. American Journal of Kidney Diseases. 2010;55(6):992–1000. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Khan SU, Acquah I, Javed Z, et al. Social Determinants of Health Among Non-Elderly Adults with Stroke in the United States. Mayo Clinic Proceedings. 2022;97(2):238–249. [DOI] [PubMed] [Google Scholar]
- 29.Valero-Elizondo J, Javed Z, Khera R, et al. Unfavorable social determinants of health are associated with higher burden of financial toxicity among patients with atherosclerotic cardiovascular disease in the US: findings from the National Health Interview Survey. Archives of Public Health. 2022;80(1):248. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Thompson EL, Rosen BL, Maness SB. Social Determinants of Health and Human Papillomavirus Vaccination Among Young Adults, National Health Interview Survey 2016. Journal of Community Health. 2019;44(1):149–158. [DOI] [PubMed] [Google Scholar]
- 31.Wray CM, Tang J, López L, Hoggatt K, Keyhani S. Association of Social Determinants of Health and Their Cumulative Impact on Hospitalization Among a National Sample of Community-Dwelling US Adults. Journal of General Internal Medicine. Published online 2021:1–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Gelman A You need 16 times the sample size to estimate an interaction than to estimate a main effect. Published March 15, 2018. Accessed March 14, 2023. https://statmodeling.stat.columbia.edu/2018/03/15/need-16-times-sample-size-estimate-interaction-estimate-main-effect/
- 33.Williams AM, Sahel JA. Addressing Social Determinants of Vision Health. Ophthalmology and Therapy. 2022;11(4):1371–1382. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Gold R, Bunce A, Cowburn S, et al. Adoption of Social Determinants of Health EHR Tools by Community Health Centers. Annals of Family Medicine. 2018;16(5):399–407. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Nitkin K A New Way to Document Social Determinants of Health.pdf Published 2011. Accessed March 25, 2021. https://www.hopkinsmedicine.org/office-of-johns-hopkins-physicians/best-practice-news/a-new-way-to-document-social-determinants-of-health
- 36.Srinivasan NK. Diabetes and Diabetic Retinopathy: Knowledge, Attitude, Practice (KAP) among Diabetic Patients in A Tertiary Eye Care Centre. Journal of Clinical and Diagnostics Research. Published online 2017:1–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Zhang X, Williams DE, Beckles GL, et al. Diabetic Retinopathy, Dilated Eye Examination, and Eye Care Education Among African Americans, 1997 and 2004. Journal of the National Medical Association. 2009;101(10):1015–1021. [DOI] [PubMed] [Google Scholar]
- 38.Solomon SD, Chew E, Duh EJ, et al. Diabetic Retinopathy: A Position Statement by the American Diabetes Association. Diabetes Care. 2017;40(3):412–418. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.IOM (Institute of Medicine). Capturing Social and Behavioral Domains in Electronic Health Records: Phase 1. Washington, DC: The National Academies Press. Published online June 23, 2014. [PubMed] [Google Scholar]
- 40.Braveman P, Gottlieb L. The Social Determinants of Health: It’s Time to Consider the Causes of the Causes. Public Health Reports. 2014;129(1_suppl2):19–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Reed NS, Assi L, Pedersen E, Alshabasy Y, Deemer A, Deal JA, Willink A, Swenor BK. Accompaniment to healthcare visits: the impact of sensory impairment. BMC Health Services Research. 2020;20(1):990–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.