Development and Validation of a Model Using Routine Clinical Features to Improve Diabetes Diagnosis

Abstract Context The hemoglobin A1c (A1c) and fasting plasma glucose (FPG) miss diabetes diagnosis in many individuals, but the alternative, oral glucose tolerance testing (OGTT), is difficult to obtain. Objective To develop and validate a model to identify diabetes missed by A1c and FPG but detected by OGTT and associate its outcomes with mortality. Design We trained machine learning models using National Health and Nutrition Examination Survey (NHANES) data, applied them to corresponding National Center for Health Statistics mortality data, and externally validated the models with Diabetes Prevention Program Outcomes Study (DPPOS) data. We tested models including the A1c, vital signs, complete blood count, and comprehensive metabolic panel with and without the FPG. Setting Retrospective analysis Patients or Other Participants NHANES study subjects from 1999-2016 and DPPOS study subjects from 1996-2020. Intervention None Main Outcome Measures Area under the receiver operating curve, sensitivity, and specificity of models to detect diabetes Results The A1c/FPG+ model had better sensitivity for DPPOS subjects with A1c < 6.5% (48mmol/mol) and FPG < 126 mg/dL compared to a logistic regression of A1c and FPG and re-calibrated FPG for subjects both on (0.69 vs. 0.53 vs. 0.57, respectively) and off (0.75 vs. 0.65 vs. 0.68, respectively) medication. NHANES subjects without diabetes but with A1c/FPG+-predicted diabetes had elevated risk of mortality compared to those without A1c/FPG+-predicted diabetes (HR=1.4, p<2x10-16). Conclusions Machine learning models can improve detection of individuals who would need OGTT assessment to diagnose diabetes using standard non-fasting labs and vital signs with and without the FPG.