causalCmprsk {causalCmprsk}

Estimation of Average Treatment Effects (ATE) of Point Intervention on Time-to-Event Outcomes with Competing Risks

Description

The package accompanies the paper of Charpignon et al. (2022). It can be applied to data with any number of competing events, including the case of only one type of event. The method uses propensity scores weighting for emulation of baseline randomization. The package implements different types of weights: ATE, stabilized ATE, ATT, ATC and overlap weights, as described in Li et al. (2018), and different treatment effect measures (hazard ratios, risk differences, risk ratios, and restricted mean time differences).

Details

The causalCmprsk package provides two main functions: fit.cox that assumes Cox proportional hazards structural models for cause-specific hazards, and fit.nonpar that does not assume any model for potential outcomes. The function get.weights returns estimated weights that are aimed for emulation of a baseline randomization in observational data where the treatment was not assigned randomly, and where conditional exchangeability is assumed. The function get.pointEst extracts a point estimate corresponding to a specific time point from the time-varying functionals returned by fit.cox and fit.nonpar. The function get.numAtRisk allows to obtain the number-at-risk statistic in the raw and weighted data.

References

M.-L. Charpignon, B. Vakulenko-Lagun, B. Zheng, C. Magdamo, B. Su, K.E. Evans, S. Rodriguez, et al. 2022. Causal inference in medical records and complementary systems pharmacology for metformin drug repurposing towards dementia. Nature Communications 13:7652.

F. Li, K.L. Morgan, and A.M. Zaslavsky. 2018. Balancing Covariates via Propensity Score Weighting. Journal of the American Statistical Association 113 (521): 390–400.