Extended cox proportional hazard model to analyze and predict conversion from mild cognitive impairment to alzheimer's disease

According to Alzheimer's Association, 70% of dementia cases are caused by Alzheimer's Disease (AD). Typically, AD progresses in three stages: mild (early stage), moderate (middle stage), and severe (late stage). Recently, predicting the conversion of mild to moderate and to severe is essential. Mini-Mental State Examination (MMSE) is a multi-domain cognitive test that is used by physicians to evaluate patients with cognitive impairment. Number of researchers pointed out that MMSE is a poor predictor to identify AD. In this work, we propose to improve MMSE predictability by using a combination of predictors with MMSE. Cox proportional Hazard model (CPHM) used to analyze the clinical NACC Uniform Data Set (UDS). Two extensions of CPHM were used; Anderson-Gill (AG) and frailty. Explanatory predictors' effects on time to failure were assessed at 5% level of significance using CPHM. Suggested models assessed using Akaike's Information Criterion (AIC) and Bayesian information criterion (BIC), and Cox-Snell residuals. Results showed that AG model more appropriate to model the NACC dataset. According to AG model the predication accuracy improved when MMSE combined with other predictors. Moreover, some predictors effect of the power of MMSE as a predictor.