Generalised estimating equations (GEEs) to handle missing data and time-dependent variables in longitudinal studies: an application to assess the evolution of health related quality of life in coronary patients (HRQoL)

OBJECTIVES: to analyse the evolution of health related quality of life (HRQoL) in coronary patients (CP) and to identify predictive factors influencing this evolution in a situation with missing data and time-dependent variables. DESIGN: a prospective, repeated measures study. SETTING AND PARTICIPANTS: a total of 175 CPs were included. General estimating equations (GEEs) models were used to assess the evolution of HRQoL in these patients. GEE models, which are not commonly used in this context, are applied here as an alternative to traditional techniques, which do not handle missing data and time-dependent covariates adequately. MAIN OUTCOME MEASURES: HRQoL assessed by SF-36v1 Questionnaire at baseline, 3 months and 6 months after discharge. RESULTS: physical role functioning, bodily pain, general health perception, vitality, and the physical component summary of the SF-36 improved over follow-up. Being a woman, being older, and having higher scores on the GHQ-28 were associated with a decrease in HRQoL over time. Previous history of coronary heart disease, comorbidities, revascularisation, rehospitalisation, and episodes of angina had a negative impact on HRQoL, especially between 3 and 6 months after discharge. CONCLUSION: analysis of the evolution of HRQoL with a longitudinal approach using GEE models shows the predictive effect of the variables analysed during the follow-up, including time itself and time-dependent covariates such as the evolution of mental health. In addition, it makes it possible to particularise the predictive effect of covariates at each period within the follow-up.