Background and purpose
Diabetic retinopathy (DR) is a leading cause of vision impairment among adults with diabetes. Socioeconomic factors, including income, influence health outcomes by shaping access to care, adherence to treatment, and engagement in preventive services. This study uses NHANES data from 1999 to 2020 to examine whether family income to poverty ratio (PIR) is associated with DR prevalence in U.S. adults aged 20 years and older, and whether PIR could serve as a practical predictor for DR risk.
Methods
Researchers analyzed NHANES data spanning 1999–2020, comprising 39,210 adults after excluding those under 20 or with missing PIR or DR data. PIR is calculated as family income divided by the federal poverty threshold, adjusted for family size and survey year. PIR values of 5.0 or more were capped at 5.0. Participants were categorized into Group 1 (PIR < 5) and Group 2 (PIR ≥ 5).
DR status was self-reported based on a physician diagnosis of diabetic eye involvement. The study applied univariate and multivariate logistic regression to estimate adjusted odds ratios (ORs), with two models controlling for different sets of covariates. Model 1 included age, sex, race, and education; Model 2 added BMI, waist circumference, smoking, diabetes duration, insulin use, HbA1c, albumin, systolic blood pressure, total cholesterol, and HDL-C. Restricted cubic spline (RCS) analyses explored potential nonlinearities, while ROC curves assessed predictive performance, with AUC reported.
Key findings
Overall, DR prevalence was 2.5% in the study cohort, higher in the PIR < 5 group (2.7%) than in the PIR ≥ 5 group (1.5%). After full adjustment, PIR remained negatively associated with DR: OR 0.91 (95% CI 0.86–0.97; p = 0.002). In subgroup analyses, Non-Hispanic Whites showed a statistically significant protective association (OR 0.87; 95% CI 0.80–0.96; p = 0.005), while other race groups did not show a statistically significant interaction (P for interaction = 0.017).
Curve fitting indicated a linear inverse relationship between PIR and DR prevalence within the PIR < 5 stratum, supporting a monotonic reduction in risk with higher PIR. The model’s discrimination was acceptable, with an AUC of 0.747 after adjusting for age (95% CI 0.74–0.76). Sensitivity reached about 82%, but specificity hovered around 59%, suggesting PIR can flag higher-risk individuals for screening but should not replace clinical evaluation.
Interpretation and implications
The study demonstrates an inverse association between PIR and DR prevalence among U.S. adults, implying that higher household income relative to poverty reduces DR risk. PIR likely captures multiple mediators of health, including access to eye care, ability to afford treatments, and adherence to diabetes management. This finding supports policy and public health efforts to heighten DR screening and diabetes management in low-income populations to reduce vision loss and broader socioeconomic burdens.
Limitations include the cross-sectional design, reliance on self-reported DR data, potential unmeasured confounders, and the inability to distinguish type 1 from type 2 diabetes in DR analysis. Nonetheless, PIR provides a simple, population-level proxy for socioeconomic risk that can inform targeted screening strategies and resource allocation.
Public health relevance
Given the rising global burden of DR, integrating PIR into risk stratification may help identify communities in need of intensified screening programs. Clinicians and health systems should consider socioeconomic context when planning outreach, education, and ophthalmologic services to curb preventable blindness related to diabetes.
Conclusion
Higher PIR is associated with a lower likelihood of DR among U.S. adults, with the association most evident among Non-Hispanic Whites. PIR shows potential as a screening adjunct for DR risk, reinforcing the call for equity-focused screening and diabetes care to mitigate social and health burdens.
