Modeling Liver Cirrhosis Progression Using Hidden Markov Model

Objective:The aim of this study was to introduce hidden Markov model to liver cirrhosis using medical records data in order to estimate the transition probabilities through the different degenerative phases of liver cirrhosis, find out related factors and provide theory basis for liver cirrhosis prevention. We explore the matters needing attention in using hidden Markov model and provide methodological guidance for other similar chronic disease. Methods:We employed a hidden Markov model to determine the transition probabilities between two states of liver cirrhosis(compensatory phase and decompensatory phase), and of misclassification. The covariates inserted in the model were sex, age, the presence of smoking and drinking, the presence of complication of diabetes and hypertension, the presence of hepatitis C virus infection and hepatitis B virus infection. The analysis was conducted in patients with medical diagnosis of liver cirrhosis. The course of starting point record in the medical records is presumed as the observed onset of cirrhosis.Results:At the level of α=0.10, univariate analysis showed that age, HCV, drinking, smoking, diabetes and hypertension were statistically significant for transition from compensatory phase to decompensatory phase; At the level of α=0.05, multivariate analysis showed that age, drinking and hypertension were statistically significant for transition from compensatory phase to decompensatory phase. we can conclude that drinking(HR=9.299), older(HR=1.04), hypertension(HR=1.815) are risk factors for development of liver cirrhosis. According to the fitted hidden Markov model, compensatory phase have a average sojourn time of 0.937 years. The transition Intensities from compensatory phase to decompensatory phase is 1.067. Conclusion:The hidden Markov model allowing for misclassification is well suited to analyses of medical records data or similar data, since it is able to capture bias due to the fact that the quality and accuracy of the available information are not always optimal.