Time series monitors of outcomes. A new dimension for measuring quality of care

Despite the popularity of risk-adjusted outcomes as quality of health care indicators, their instability with time and their inability to provide reliable comparisons of small volume providers have raised questions about the feasibility and credibility of using these measures. In this article the authors describe a new analytic strategy to address these problems by examining risk-adjusted mortality with time, "Time Series Monitors of Outcome" (TSMO), and its application to cardiac surgery performed throughout the Department of Veterans Affairs between April 1987 and September 1992.

Expected operative mortality for 24,029 patients undergoing coronary artery bypass surgery at all 43 centers performing this procedure was estimated using a logistic regression model to adjust for patient-specific risk factors. The ratio of observed-to-expected operative mortality was calculated for each hospital for each of the 11 6-month periods. Poisson regression models were used to identify high and low outlier hospitals based on significant deviation from the 5.5 year overall mean and/or the individual hospital's trend of observed-to-expected ratios with time.

This method identified four high and one low outlier hospitals based on significant deviations from the overall mean and three upward and seven downward trending outlier hospitals based on significant deviations in trend with time. A significant downward trend in observed-to-expected ratios of 4% per year also was observed for all coronary artery bypass graft procedures performed throughout the Department of Veterans Affairs during the last 5.5 year period.

Time Series Monitors of Outcome should help reduce misclassification of outliers due to random variation in outcomes as well as provide more reliable comparative information from which to evaluate provider performance.