Recurrence and complication prevention is crucial in caring chronic patients. This paper presents a chronic disease recurrence prediction model in long-term caring for diabetes patients. A nursing information system is developed and applied in an institution-based health care center. Values of blood sugar, vital signs, BMI and nursing records of diabetes patients are collected during the daily nursing procedure. Since data are recorded in different time with variable sampling period, a flexible time series analysis method is presented for exploiting seasonal variation and cyclical fluctuation between diabetes and daily-recorded signals. Expert knowledge for chronic disease is built based on domain specific language to extract chronic disease care knowledge from nursing records. Finally, a health surveillance system is developed by integrating multi-modal daily caring information for longterm caring of chronic patients. Several experiments with statistical testing are conducted and the results compare the performance of the proposed method. The expected outcome will build a prediction model to early warning before diabetes will be recurrence.