Because of increased obesity and population aging, the global burden of diabetes is rising. Postoperative infection not only increase the complexity of healthcare and prolong hospitalization, but also increase medical cost. Although preventive and treatment measures are documented for complications in diabetes, little guidance exists for postoperative infectious complications in surgical patients with diabetes, despite evidence suggesting greater susceptibility to infections.The contributing factors of postoperative infectious complications are complicated and difficult to define. Big data analysis using Taiwan National Health Insurance Research Database (NHIRD) combined with data mining techniques could be an ideal approach. Data mining techniques including artificial neural networks (ANNs), classification regression tree, and support vector machine are applicable to analysis of large database. ANNs may be superior to statistical methods in some aspects, such as the development of non-linear models, tolerance for missing data, and high adaptation. Prediction models based on data mining techniques may have better predictive performance than those based on logistic regression analysis. We focused on NHIRD analysis and development of ANN models and have published several papers about “risk of fracture in patients with diabetes”, “postoperative adverse outcomes in diabetes patients”, and “neural network modeling to predict hypotensive risk in patients undergoing anesthesia”.This is a two-year plan. In the first year, using Taiwan NHIRD, we will evaluate the risk of postoperative infectious complications in surgical patients with diabetes and identify the risk factors to develop risk assessment models for infectious complications. In the second year, we use clinical database, which include more parameters such as laboratory data, drug prescriptions, comorbid conditions, and severity of diseases, to develop risk assessment models for infectious complications. Blood sugar data including fasting plasma glucose and glycated hemoglobin are recorded in clinical database. These data can be used to indicate the status of glycemic control and severity of the disease. Finally, we evaluate the predictive performance of the models and assess the practical value in clinical practice.With the help of the risk assessment models, physician can explain surgical risk to patients and their families by evidence; medical staff can be alerted to prevent or reduce the occurrence of complications; for patients, they can know their surgical risk in advance. The models can provide a lot of help to healthcare and doctor-patient relationship.
|Effective start/end date||8/1/18 → 10/1/19|
- data mining
- artificial neural network
- risk analysis