Enhancing Longitudinal Data Analysis with Unstructured EHRs: A Case Study of Renal Function Evaluation in Rare Disease

Electronic Health Records (EHRs) provide valuable longitudinal data for tracking disease progression, especially in rare diseases like ciliopathies which often involve chronic renal decline. While important biomarkers are available in structured databases, crucial information such as external lab tests and detailed disease history may only be found in clinical narratives. This study aims to enrich structured datasets with unstructured clinical text and assess its impact on estimating chronic kidney disease progression in ciliopathy patients. Our results demonstrate that data enrichment increased the number of eligible patients for longitudinal analysis by 73.5%, expanded available measurements by 189%, and significantly extended the median follow-up duration from 3.2 to 6.6 years. Using linear mixed regression to model individual estimated glomerular filtration (eGFR) rate trajectories over age, we found that data enrichment reduced standard errors by 30%, indicating a substantial increase in precision and reliability. These findings underscore the value of EHR data enrichment for longitudinal analysis in rare disease research.