Modelling recurrent events: a tutorial for analysis in epidemiology

Abstract

In many biomedical studies, the event of interest can occur more than once in a participant. These events are termed recurrent events. However, the majority of analyses focus only on time to the first event, ignoring the subsequent events. Several statistical models have been proposed for analysing multiple events. In this paper we explore and illustrate several modelling techniques for analysis of recurrent time-to-event data, including conditional models for multivariate survival data (AG, PWP-TT and PWP-GT), marginal means/rates models, frailty and multi-state models. We also provide a tutorial for analysing such type of data, with three widely used statistical software programmes. Different approaches and software are illustrated using data from a bladder cancer project and from a study on lower respiratory tract infection in children in Brazil. Finally, we make recommendations for modelling strategy selection for analysis of recurrent event data.

Job strain and informal caregiving as predictors of long-term sickness absence: A longitudinal multi-cohort study

“…To account for recurrent events of long-term sickness absence we used the Prentice, Williams and Peterson Total Time (PWP-TT) model (41), which is appropriate if the occurrence of the first event increases the likelihood of a recurrence. In PWP-TT, multiple events are ordered by stratification based on the prior number of events, such that all participants are at risk of an event in the first stratum, but only those with a prior event are at risk for a successive event (41).…”

“…In PWP-TT, multiple events are ordered by stratification based on the prior number of events, such that all participants are at risk of an event in the first stratum, but only those with a prior event are at risk for a successive event (41). We truncated analy-ses at two spells of long-term sickness absence during follow-up, due to small number of individuals with a greater number of spells.…”

Builiding Methods for a Proactive Prescription Drug Surveillance System

“…Since this is a unit-day measure with patients experiencing multiple daily high dose opioid days, the Prentice, William, and Peterson (PWP) recurrent event model was used to estimate the number of high-dose opioid days for Kansas patients by gender and age groups. 4,5 Start time was the first prescription date with a high-dose opioid and stop time was the next high-dose opioid date during a study period from January 1, 2014 to Feb 29, 2016. The PWP model is a statistical model that allows for the estimation of covariates on an event history (i.e.…”

Fluoroquinolone Impact on Nasal Methicillin-Resistant and Methicillin-Sensitive Staphylococcus aureus Colonization Durations in Neurologic Long-Term-Care Facilities

“…Associations between factors and the loss of MRSA, MSSA, or S. aureus colonization were investigated using Cox proportional hazards models (21). Because a patient could have several colonization episodes, a shared-frailty term was used to model a lack of independence between a given patient's events (22). Episodes with a single S. aureus-positive nasal swab collected during the last week of the patient's follow-up were not analyzed, because survival analysis requires that each episode be observed over some nonnull time interval.…”