Introduction and Objective: While the Veterans Affairs (VA) MOVE! Program (DPP-like lifestyle intervention) demonstrated a 24% relative risk reduction in type 2 diabetes (T2D) over 15 years, individual responses varied substantially. We examined this heterogeneity of effects.Methods: Using 2005-2024 VA electronic health records data, we identified NDPP-eligible veterans who enrolled in MOVE! from 2005 to 2009 (n=28,668) and MOVE!-eligible nonparticipants using 1:1 high-dimensional propensity score matching. We applied a causal survival forest algorithm and Shapley Additive exPlanations values to identify key effect modifiers influencing the effectiveness of MOVE! over 15 years from demographics, disease history, medications, vitals, and biomarkers. We further developed a scoring algorithm to estimate personalized T2D risk reduction.Results: We identified ten key effect modifiers and corresponding effect scores. Participants with prediction scores (range: -9 to 18) above 0 demonstrated measurable benefits from MOVE!, with those in the highest quintile (scores >7) achieving a 36% reduction in T2D risk (95% CI: 33%-39%) over 15 years.Conclusion: Our personalized prediction algorithm effectively identifies individuals who receive the greatest benefit from MOVE! Program, providing insights to guide targeted T2D prevention efforts in clinical practice and health policy.
Q. Xue: None. P. Li: None. J. Lee: None. T. Moin: None. S. Raffa: None. M.K. Ali: Advisory Panel; Eli Lilly and Company. Y. Hong: Consultant; Weight Watchers International. M.K. Shah: None. L.S. Phillips: Research Support; Janssen Pharmaceuticals, Inc, Boehringer-Ingelheim. Other Relationship; AbbVie Inc, ForaCare, Inc., Diasyst, Inc. Research Support; Kowa Pharmaceuticals America, Inc, Pfizer Inc. H. Shao: None.
CDC (U18DP006747)
Source link
Leave a Reply