The performances of the Cox, Andersen-Gill, Prentice- Williams-Peterson-total time and Wei-Lin-Weissfeld- total-time models in Identifying risk factors in patients with recurrent diseases : a kidney infections example

Abstract

In many longitudinal studies, when subjects are followed over a period of time, recurrent event frequently occur. However, some analysis focusses only on time to the first event, ignoring the subsequent events. The main objective of this paper was to compare the extended standard Cox models, such as Andersen-Gill (AG), Prentice-Williams-Peterson total time (PWP-TT), PWP-Gap time model, Wei Lin-Weissfeld total time (WLW-TT), and Cox frailty model, to identify risk factors associated with kidney re-infection. Empirical evaluation and comparison of these different models were performed. The better model was assessed based on the goodness of fit criteria (AIC, BIC and likelihood ratio test). Kidney data that was downloaded from the R statistical software using the command data(“kidney”) was used to perform analyses in this study. The PWP-TT model had lower standard errors, AIC and BIC values compared to other models, therefore fitted data better and was used to interpret results. The results showed that 81% (HR = 0.19; 95% CI: 0.09-0.39) of the female patients were less likely to experience kidney reinfection than male patients. The risk of recurrent kidney infection was significantly high (HR = 2.32; 95% CI: 1.25-4.29) to patients having an Acute Neptiritis (AN) disease compared to patients with other diseases. While the prevalence of kidney infection remains the public health problem, intervention strategies and awareness campaigned are needed to in order to minimize risk factors behind the recurrent of the disease.