Prediction mortality among laboratory-confirmed avian influenza h7h9 patients: risk classification model based on variable selection from multiple imputation.

The clinical risk points system makes complex statistical models practical and convenient for clinical use. This risk points system helps clinicians make their decisions for the treatment process quickly with its characteristic as a scientific tool for predicting risks of diseases or incorporating effective evidence-based approaches. To develop the clinical risk points system for data with missing observations, variable selection arises as one of the statistical problems with multiple imputation (MI). Also, we are confronted with the challenge of developing a simultaneous risk points system with multiply-imputed datasets. In our study, we suggest a multiple imputation-stepwise method (MI-Stepwise) across multiply-imputed data to yield a consistent variable selection. Simulations are conducted and we apply the methods to the Asian lineage avian influenza Asian H7N9 virus (A/H7N9) study in the China Centers for Disease Control and Prevention (China CDC) to predict death.