Simplifications that don’t work: When ignoring competing and recurrent events leads down the wrong causal path.
Epidemiologic analyses require independent, identically distributed outcomes, conditional on covariates. This assumption may be violated when mediating variables cause two different events and one event prevents the other from occurring, or when repeated events are more frequent in high-risk individuals (eg. injury, pregnancy). Analyses that ignore these correlations may lead to inappropriate conclusions and ineffective interventions. This symposium will illustrate the concepts/underlying assumptions, and demonstrate methods that better address particular research questions.
We will address important challenges and solutions for injury epidemiology (Shrier), perinatal epidemiology (Naimi), and highlight the commonality for all recurrent and/or competing event outcomes as well as discuss where methodological research is required to address ongoing limitations (Lau).