For individual health examination data, the traditional mathematical model based on large samples can not meet the modeling requirements of physical examination data. Based on the analysis of the characteristics of individual physical examination data, an improved discrete grey model of approximately non-homogeneous index series suitable for individual physical examination indicator health warning was first constructed. Secondly, in order to reduce the limitation of the prediction accuracy of a single model, the inverse variance method was used to assign weights to the discrete grey model and the differential autoregressive moving average model, and the best weight value was obtained when the sum of squares of the model errors reached the minimum. Thus, the prediction results of the two models were combined to achieve the modeling and trend analysis of health indicators, timely grasp the changes of individual health indicators and discover potential disease hazards. The relative simulation error of the prediction model on the experimental data set decreases in comparison with the optimal benchmark model, which indicates that the grey time series combination model has higher simulation accuracy. The shortcomings of traditional static analysis based on single physical examination indicators and the limitations of single model prediction results are solved. Individual differences are emphasized, and the effect of health warning can be effectively improved.