Data-driven Cluster Analysis Reveals Increased Risk for Severe Insulin-deficient Diabetes in Black/African Americans

Abstract

Context: Diabetes is a heterogenic disease and distinct clusters have emerged, but the implications for diverse populations have remained understudied. Objective: Apply cluster analysis to a diverse diabetes cohort in the US Deep South. Design: Retrospective hierarchical cluster analysis of electronic health records from 89 875 patients diagnosed with diabetes between January 1, 2010, and December 31, 2019, at the Kirklin Clinic of the University of Alabama at Birmingham, an ambulatory referral center. Patients: Adult patients with International Classification of Diseases diabetes codes were selected based on available data for 6 established clustering parameters (glutamic acid decarboxylase autoantibody; hemoglobin A1c; body mass index; diagnosis age; HOMA2-B; HOMA2-IR); ∼42% were Black/African American. Main Outcome Measure(s): Diabetes subtypes and their associated characteristics in a diverse adult population based on clustering analysis. We hypothesized that racial background would affect the distribution of subtypes. Outcome and hypothesis were formulated prior to data collection. Results: Diabetes cluster distribution was significantly different in Black/African Americans compared to Whites (P