914-P: The Impact of Racial and Ethnic Health Disparities in Diabetes Management on Clinical Outcomes: A Reinforcement Learning Analysis of Health Inequity

It is not known how racial/ethnic disparities in pediatric type 1 diabetes (T1D) treatment regimens contribute to the disparate clinical outcomes across racial/ethnic subgroups. Data were analyzed from the SEARCH for Diabetes in Youth study (diagnosed 2002-2005, 47.5% female, 77.5% non-Hispanic white (NHW), mean age at diagnosis=12.8±2.4 years). Youth were classified as nonwhite race/Hispanic ethnicity (nonwhite youth) or NHW race (white youth). Multivariate propensity scores estimated T1D treatment regimens (i.e., insulin delivery and glucose monitoring modality) in white versus nonwhite subgroups. Reinforcement learning, a machine learning technique, estimated the effect of each treatment regimen on HbA1c over time for nonwhite youth. Observed mean HbA1c over available visits for each individual was 9.0% and 8.2% for the nonwhite and white subgroups, respectively, difference=0.8%. Estimated mean HbA1c decreased by 0.33% (95%CI: -0.45%, -0.21%; p<0.001) if nonwhite youth received the white subgroup treatment regimen (Table), explaining ~40% of the observed HbA1c disparities. These data underscore the complexity of race-based T1D health inequity and the urgent need to identify and address other contributing factors.

View largeDownload slide

View largeDownload slide