Background: Sickle cell disease is the most frequent hereditary disease in Brazil, and people with the disease may be hospitalised several times in the course of their lives. The purpose of this study was to estimate the hazard ratios of factors associated with the time between hospital admissions.
Methods: The study sample comprised all patients admitted, from 2000 to 2004, to a university hospital in Rio de Janeiro State, south-east Brazil, as a result of acute complications from sickle cell disease (SCD). Considering the statistical problem of studying individuals with multiple events over time, the following extensions of Cox's proportional hazard ratio model were compared: the independent increment marginal model (Andersen-Gill) and the random effects model.
Results: The study considered 71 patients, who were admitted 223 times for acute events related to SCD. The hazard ratios for hospital readmission were statistically significant for the prior occurrence of vaso-occlusive crisis and development of renal failure. However, analysis of residuals of the marginal model revealed evidence of non-proportionality for some covariates.
Conclusion: the results from applying the two models were generally similar, indicating that the findings are not highly sensitive to different approaches. The better fit by the frailty model suggests that there are unmeasured individual factors with impact on hospital readmission.