For women with gestational diabetes mellitus (GDM), identifying those at higher risk of requiring insulin therapy remains challenging when lifestyle measures do not achieve glycaemic control. In Diabetes Research and Clinical Practice, a survival forest analysis assessed a predictive model to estimate time to insulin initiation using data from the EMERGE trial.

The analysis included 413 women with GDM and applied a random survival forest (RSF) model to predict time to insulin initiation. Models were developed separately for placebo and metformin groups. Predictors included maternal characteristics and early glucose measurements collected during the first two weeks after randomisation. Model performance was evaluated using the concordance index, time-dependent area under the curve (AUC), and Brier score. Decision curve analysis compared the net benefit of the RSF model with default clinical strategies.

In the placebo group, the RSF model achieved a concordance index of 0.71 (95% CI, 0.64–0.77), time-dependent AUC values of at least 0.70, and Brier scores of 0.2 or lower. In the metformin group, the concordance index was 0.72 (95% CI, 0.64–0.80), with time-dependent AUC values of at least 0.75 and Brier scores of 0.2 or lower. Decision curve analysis showed higher net benefit for the RSF model across clinically relevant threshold probabilities.

These results show that the RSF model identified women with GDM at higher risk of requiring insulin, although prospective validation is needed to confirm generalisability.