Estimating equations for causal survival analysis with pooled logistic regression

Background: Pooled logistic regression models are commonly applied in survival analysis. However, the standard implementation can be computationally demanding, which is further exacerbated when using the nonparametric bootstrap for inference. To ease these computational burdens, investigators often coarsen time intervals or assume a parametric models for time. These approaches impose restrictive assumptions, which may not always have a well-motivated substantive justification. Methods: Here, the pooled logistic regression model is re-framed using estimating equations to simplify computations and allow for inference via the empirical sandwich variance estimator, thus avoiding the more computationally demanding bootstrap. The proposed implementation is demonstrated using two examples with publicly available data. The performance of the empirical sandwich variance estimator is illustrated using a Monte Carlo simulation study. Results: As shown in the applied examples, the proposed implementation substantially reduced run-times and could be applied without needing to coarsen the data. In the simulation study, the empirical sandwich variance estimator results in nominal confidence interval coverage. Conclusions: The implementation proposed here offers an improved alternative to the standard implementation of pooled logistic regression without needing to impose restrictive constraints on time.