The long-term objectives of this project are to develop methods for risk adjustment for surgical procedures which are performed in populations without significant medical comorbidities. These methods will be used as (1) prospective aids to assist patient and provider decision-making about the risks and benefits of surgery, and (2) as tools for quality assurance and quality improvement programs, cost containment, and health services research.
The specific aims are to (A) assess the reliability and validity of physician review of operative notes as a method of risk adjustment by comparing physician global ratings of surgical difficulty to outcomes, (B) develop multivariate models for the prediction of medical and surgical complications of hysterectomy based on preoperative findings for prospective risk prediction, and (C) develop multivariate models including preoperative and intraoperative findings not routinely used for risk adjustment which account for variations in morbidity, length of stay, and hospital costs. Hysterectomy is the most commonly performed major surgery in non-pregnant women. Although mortality is low, complication rates are approximately 10 percent. Wide variability in complication rates has been observed between hospitals and between different patient populations. In order to understand the degree to which variations in the quality of care are responsible for the observed variations in complication rates, further knowledge about the relationship between clinical factors affecting the performance of the surgery itself, such as the presence of adhesions, the size of the uterus, or the extent of diseases such as endometriosis or gynecologic cancer, and complications is needed. First, the technical difficulty of the procedure and the associated intraoperative findings documented in dictated operative notes will be rated by experienced gynecologic surgeons. Second, clinical findings associated with difficult surgery will be recorded from medical records by trained chart abstractors. Third, multivariate regression models will be constructed to determine which combination of variables best predict complication risk, length of stay, and charges. Future studies will assess the applicability of these methods to other settings.
