Geographic Modeling and Social Isolation in ADPKD

Background: Social and environmental factors may impact the development of ESRD in patients with autosomal dominant polycystic kidney disease (ADPKD). We investigated social and place-based determinants of health in ADPKD patients with ESRD. Methods: Patients completed the PKU-impact scale and Brics Social Determinants of Health surveys assessing social isolation, the impact of PKD on emotional and physical well-being, and fatigue. The impact of these factors on ESRD development is modelled with univariate and multivariate logistic regression. Non-linear relationships are modelled with Random Forest (RF) and variable importance analysis is performed to rank the importance of factors on ESRD outcomes (Fig. 1). Geographical clustering on impact factors is performed to elucidate location-based disparities in ESRD outcome. Results: 89 ADPKD patients included (51.9 mean age, 47 males, 42 females; 29 non-European ancestry, 60 European), 49 had ESRD. Social isolation score (adjusted OR = 1.58, 95% CI: 1.03–2.51, p = 0.041) and BMI (adjusted OR = 0.91, 95% CI: 0.84–0.98, p = 0.016) were significant in univariate and multivariable analysis. Pseudo R2=0.134. RF captures 92% of the variance in the ESRD outcome, indicating non-linear relationships. Age to ESRD significantly differ among clusters (Fig. 2), as quantified with a two-sample Kolmogorov-Smirnov test (p=0.07). Conclusion: Patients with ESRD had lower BMIs and reported higher social isolation scores than patients without. Tree-based non-linear regression captures variability in ADPKD outcomes more accurately compared to linear models. Geographic clustering of ADPKD drivers point to location-based heterogeneity.