Early management of DFIs produced divergent outcomes in a causal machine-learning analysis published in Diabetology. Using de-identified electronic health record data from 1,434 adults treated at the University of California, San Francisco between 2015 and 2024, the study compared outcomes of early (<3 days) versus delayed or absent treatment.

Early treatment increased hospitalization risk by 29 percent (Targeted Maximum Likelihood Estimation risk difference 0.293; 95% CI, 0.220–0.367) but was associated with a protective trend against lower-extremity amputation (risk difference −0.040; 95% CI, −0.098 to 0.066). The paradox reflects prioritization of more severely ill patients for rapid intervention and potential exposure misclassification due to fragmented care records.

Results highlight the complexity of real-world causal inference in diabetic complications. Even advanced analytic methods require integrated data systems and clinically informed algorithms to yield accurate insights for patient management and decision support.