Left-censored recurrent event analysis in epidemiological studies : a proposal for when the number of previous episodes is unknown