Chronic Disease Monitoring: Methodology for Classification Error and Self-Selection Bias Correction in Clinical...

Background/Objectives: Chronic diseases are among the leading causes of morbidity and healthcare costs worldwide. Diabetes mellitus is one of the most prevalent and costly chronic conditions in the United States, with a disproportionate burden in Puerto Rico. Surveillance of diabetes relies mainly on infrequent cohort studies and self-report surveys, which are limited in accuracy, segmentation, and timeliness. This study aimed to develop a generalizable methodology for monitoring chronic disease prevalence using routinely collected laboratory data, while correcting for systematic biases and diagnostic errors. Methods: We analyzed more than five years of de-identified laboratory test results (2020-2024) from a large, island-wide network of clinical laboratories in Puerto Rico. To produce unbiased prevalence estimates, we applied a mathematical correction framework that accounted for two main sources of distortion: (1) classification errors from treatment effects and test limitations, quantified through confusion matrices derived from longitudinal records; and (2) self-selection bias from differential testing rates, estimated empirically by demographic segment. Demographic reweighting ensured representativeness with respect to census data. Results: Using diabetes as a test case, corrected estimates for 2024 showed an adult prevalence of 18.0%, compared to 14.1% based on raw laboratory frequencies. The large amount of data provided high-resolution estimates by age, sex, and location, enabling fine-grained detection of demographic and geographic disparities. Conclusions: Bias-corrected laboratory surveillance provides accurate, timely, and demographically representative estimates of chronic disease prevalence. The methodology is scalable, cost-effective, and broadly applicable to other multi-stage chronic conditions, offering a foundation for next-generation public health monitoring and targeted interventions.