Authors :
Presenting Author: Adetoun Abisogun Musa, MD – Kingdom Healing Institute
Donald Wesson, MD – Dell Medical School, The University of Texas at Austin
Tiffany Jones-Smith, CEO – The State of Texas Kidney Foundation
Kevin Smith, Chairman of the Board – The State of Texas Kidney Foundation
Rationale: Cardiometabolic diseases (CMD) encompass a constellation of disorders, including chronic kidney disease (CKD), hypertension (HTN), diabetes (DM), and obesity. These disorders pose significant public health challenges due to their association with end-organ damage and neurological complications. The links between creatinine levels, estimated glomerular filtration rate (eGFR), and neurobehavioral disorders like epilepsy and dementia have been poorly described in the literature, warranting further investigation. We explored datamarkers and potential predictors of kidney disease (as measured by eGFR), epilepsy, and dementia by extracting data containing social determinants of health (SDOH) and biometric measurements, using advanced AI/ML techniques. The aim of this study is to develop predictive models for neurological sequelae such as epilepsy and dementia. Future aims include exploring causative relationships and developing personalized prevention and early intervention strategies.
Methods: Data from 14,561 individuals in the Texas Kidney Foundation (TKF) dataset (2010-2023) were analyzed, including variables such as family history, physical and laboratory measurements, race, and current health insurance status. AI/ML models included decision trees, ordinal logistic regression, and XGBoost. These models were then validated using 200,000 participants from the All of Us dataset. Cross-validation and statistical techniques ensured the model robustness.
Results: Creatinine levels, age, BMI, diastolic blood pressure, current health insurance status, and diabetes history were identified as key predictors for determining eGFR. Contrary to conventional expectations, race was found to be a non-significant predictor for eGFR. Validation using the All of Us dataset demonstrated a relatively high influence of both creatinine and eGFR in predicting epilepsy and dementia. The models demonstrated strong predictive performance, with AUC values of 0.74 for epilepsy and 0.75 for dementia.
Conclusions:
The non-significance of race in predicting eGFR and neurobehavioral disorders challenges existing guidelines, highlighting the need for precise datamarkers, including SDOH, particularly in marginalized communities. Current reliance on race as a predictor may overlook critical biomarkers like health insurance status, potentially underestimating kidney disease and neurobehavioral risks. Future research will explore causative relationships through Mendelian randomization studies and implement AI-driven early detection in communities to reduce health disparities. This study underscores AI/ML's potential in identifying key datamarkers for kidney disease and neurological sequelae, emphasizing the underexplored role of creatinine and eGFR in predicting epilepsy and dementia. The findings suggest shifting from race-based to biomarker-focused healthcare approaches, particularly for underserved populations. Subgroup analyses, bias investigation, and stakeholder engagement will refine models, integrate ethical considerations, and advance preventative neurology, precision medicine, and epilepsy research.
Funding: This research was funded by the National Institutes of Health (NIH) under Agreement NO. 1OT2OD032581-01.