Retrospective Study of Recurrence and Associated Factors of Type 2 Diabetes Treated at Adama General Hospital, Oromia, Ethiopia: A Comparison of Cox-PH and Shared Lognormal Frailty Models

Abstract

Background. Recovery from type 2 diabetes is frequently recurrent, as a single patient may recover from more than one over time. The goal of this study was to know the recurrent event (time to recovery) and associated factors of type 2 diabetes in Adama General Hospital, Ethiopia, by comparing shared lognormal frailty and Cox-PH models. Methods. A retrospective analysis of 302 type 2 diabetic patients (01, 2011–01, and 2016) was considered. Descriptive statistics were used to summarize the study variables. The standard Cox-proportional hazards model and a shared lognormal frailty model have been compared. The latter model with a 95% significance level was fitted, variables with $P\text{\hspace{0.17em}value}<\text{0.}05$ were considered significant, and the adjusted hazard ratio has been used to measure the strength of the risk. Results. About 56.6% of the patients recovered. The average recovery time was 33.53 (standard deviation, 20.404 ) weeks. Gender (adjusted HR = 1.168, 95% CI = (0.93, 1.46), $P<0.05$ ), family history (adjusted HR = 0.765, 95% CI = (0.59, 0.99), $P<0.05$ ), cholesterol level (adjusted HR = 0.738, 95% CI = (0.57, 0.96), $P<0.05$ ), alcohol use (adjusted HR = 0.698, 95% CI = (0.53, 0.92), $P<0.05$ ), and smoking cigarette (adjusted HR = 0.674, 95% CI = (0.51, 0.89), $P<0.05$ ) were statistically significant. The estimated frailty term’s variance was 0.426 ( $P\text{\hspace{0.17em}value}=\text{0.}028$ ). Also, the author presents a comparison study for the same data by using a model selection criterion and suggests a better model (shared lognormal frailty model). Conclusion. Finally, the median recovery time was 30 weeks. Female patients had a better chance of recovery than male patients. A shared lognormal frailty model outperformed the Cox-PH model in fitting the data and controlling event interdependence. There was risk heterogeneity among patients. Positive family history, high cholesterol level, alcohol use, and smoking have an inverse relationship with the overall likelihood of the patients’ recovery time. Therefore, future improvement measures against type 2 DM recovery should take all events (for example, the first, second, and third recovery in this study) and these identified factors into account.