A0378
Title: A pseudo-value approach to causal deep learning of semi-competing risks
Authors: Yi Li - University of Michigan (United States)
Stephen Salerno - Fred Hutchinson Cancer Center (United States) [presenting]
Abstract: While mortality is often the main focus of cancer studies, non-fatal events (i.e., disease progression) can vitally impact patient outcomes. Recurrence after curative treatment is a crucial endpoint in lung cancer, affecting second-line treatment options. Estimating the de-confounded effect of an intervention on disease recurrence is a key aspect of assessing cancer treatments. However, semi-competing risks complicate causal inference when death prevents disease recurrence. Existing approaches for estimating causal quantities for semi-competing risks rely on complex objective functions with often strong assumptions. To address these challenges, a deep learning approach is proposed for estimating the causal effect of treatment on non-fatal outcomes in the presence of dependent censoring. The three-stage approach involves estimating the non-fatal survival function, constructing jackknife pseudo-survival probabilities at fixed time points, and fitting a deep neural network to estimate the effect of treatment. The pseudo-survival probabilities serve as target values for developing causal estimators that are consistent and do not rely on assumptions like proportional hazards, which enables estimating survival average causal effects through direct standardization. The approach is evaluated through numerical studies, and it is applied to the Boston Lung Cancer Study to estimate the effect of surgical tumor resection in patients with early-stage non-small cell lung cancer.
