Charles C.H. Liu, Simon J.H. Wu, I-Jen Chiang*, Yu-Chuan Li* , Lu, S.-Y. Haw-Yen Chiu** Division of Plastic Surgery, Cathay General Hospital, Taipei, Taiwan *Graduate Institute of Medical Informatics, Taipei Medical Univeristy, Taipei, Taiwan **Division of Plastic Surgery, National Cheng-Kung University Hospital, Tainan, Taiwan*

PURPOSE:

Data mining, also referred to as knowledge discovery in database (KDD), is the automatic extraction of patterns representating knowledge implicitly stored in large database. Our work is based on the experience on a large series in our center (Liu, 2001), and now shifts to more complicated amputation trauma in the population-based, multiinstitutional database for 5 years.

MATERIALS AND METHODS:

National Health Insurance claim database in Taiwan from 1996 to 2000 is released for clinical research since 2001. With its wide coverage of 96% of 24 million population, it is promising for study of surgical operation and follow-up. Our preliminary study in 1/20 sampled dataset yielded 368 cases of limb replantations with 14274 detailed medical orders. After the first hospitalization, totally 4202 outpatient visits and another 174 related hospitalizations, mostly for other operations, are analyzed. The outcome is evaluated by other procedures in the follow-up database.

We are working on the complete trauma datasets for limb replantation with total case number of nearly 7000. The results would be compared with the 368 cases with more clinical details.

METHODS AND RESULTS:

The overall profile of replantation surgery in this regions will be presented. Digital replantation are our focus. 289 digital and 63 thumb replantations were present, and 181, 34, 9, 3, and 1 cases underwent replantations for one to five digits respectively. The operation time wer grouped in two-hour interval, with 230, 181 and 91 cases for <2, 2-4, and >4 groups. There were 20 combined flap and 13 combined vascular procedures. Anitcoagulants were used in 264 cases, including aspirin, dextran, and prostaglandin.

The above clinical factors will be evaluated on outcome factors in acute and late hospitalization and outpatient visits, including associated procedures, reexplorations, lengths of hospital stay, various types of medical expense, and subsequent amputations and other reconstruction surgeries, especially microsurgery such as toe transfer. The medications, and policy of reoperations in different characteristics of hospitals or surgeons will also be discussed.

Various data mining algorithms would be applied to disclose implicit patterns, including association rules, classification trees, regression analysis, neural networks, and cluster analysis. The results could be compared with previous findings in the more clinically informative series of our centers.

REFERENCE: Charles C.H.Liu, Simon J.H. Wu, Yu-Chuan Li, I-Jen Chiang, K.S.Liu, S.Y. Lu, and H.Y. Chiu. Prognostic prediction of digital replantation --- data mining the CGH database of 800 cases. Proceedings of the Inaural Congress of the World Society for Reconstructive Microsurgery. Bologna, Italy, 2001, 93-95.