A decision support system on surgical treatments for rotator cuff tears

Abstract

Treatment of patients with rotator cuff tears usually starts with physical therapy, but some patients will still eventually need surgery. Ineffective physical therapy increases the time and cost of treatment and pain for patients. The quality of treatment can be improved if patients who will not respond to physical therapy are identified at an early stage. However, there is little research available to systematically help physicians make a timely decision on whether a surgical treatment is eventually needed or not. In this research, we developed a decision support system that can predict the probability of eventually needing a surgical treatment by effectively analyzing the available patients’ information at an early stage. Missing value imputation, variable selection, and classification methods are integrated in developing such a decision support system. The probability given by our model will either confirm physician's expert decision, or remind physician if there is any information ignored. This research has the potential to improve patient safety, reduce cost of unnecessary treatment, and help physicians prevent treatment errors.

Keywords:

• Decision support
• probabilistic prediction
• data imputation
• feature selection
• logistic regression
• rotator cuff tears

Acknowledgments

This research is supported by the Laboratory for Optimization and Computation in Orthopaedic Surgery (LOCOS) at the Department of Orthopaedic, University of Michigan Medical School. The work of K. Paynabar was partially supported by George Family Foundation. The data in this study are provided by the University of Michigan MedSport Clinic. The authors would like to thank Dr. Richard Hughes and Mr. Jaimee Gauthier for their help with background understanding and data collection.