Aims: This study aims to investigate challenges associated with diabetes prevalence estimates in stroke survivors, focusing on the issue of misclassification bias in diagnostic tests, and to propose measures for improving the accuracy of these estimates. Methods: The study examines the inherent misclassification biases associated with the diagnostic tests, including Fasting Blood Glucose (FBG), Oral Glucose Tolerance Test (OGTT), and Hemoglobin A1c (HbA1c), commonly used to identify diabetes in stroke survivors. To address misclassification biases, three parameter Bayesian latent class models are applied to delineate true prevalence from the apparent prevalence reported in studies, using FBG, OGTT, HbA1c as standard diagnostic tests for diabetes. Results: The results revealed discrepancies between apparent and true prevalence of diabetes in stroke patients, highlighting the influence of the sensitivity and specificity of each diagnostic test on prevalence estimates. Conclusions: Correcting misclassification biases in diabetes diagnostic tests is crucial for accurate prevalence estimates in stroke survivors, which is necessary for proper diagnosis and patient care. The study underscores the need for future research to address data biases and uncertainties in diagnostic test measures, which will optimize the accuracy of diabetes diagnosis in this vulnerable population.



