Online Program

Risk-Adjusted Incidence Monitoring on Hierarchical Survival Data with Recurrent Events (308455)

Keywords: Survival Analysis, Recurrent Events, Hierarchical Clusters, Cystic Fibrosis, Infection Incidence

We are interested in monitoring survival events while taking into account multi-level hierarchical structure and recurrent events in survival data. We aim to extend the mixed-effect Andersen-Gill model with risk adjustment from important risk factors and provide variability estimates of the survival events, which is packaged into a confidence interval. We propose that the estimate variability has three sources, one of which is obtained with the block jackknife method. A pipeline is developed which includes a chain of analysis steps ranging from preprocessing, multiple imputation, variable selection, model performance, and result interpretation. Simulations are conducted to evaluate our methodology of the risk-adjusted CIs. Monitoring cystic fibrosis (CF) incidence is used as clinical application to evaluate the practicality of our pipeline. Our results show promising usage of our risk-adjusted models when we train with larger periods of simulated data and shorter periods of CF data. Our method is useful for infection prevention and control, where health programs or hospitals want to gain knowledge on their infection rates in advance and be able to take proactive action.