A comparison of methods to obtain a composite performance indicator for evaluating clinical processes in trauma care.

BACKGROUND

Process performance indicators that evaluate trauma centers in clinical case management provide information essential to the improvement of trauma care. However, multiple indicators are needed to adequately evaluate process performance, which renders comparisons cumbersome. Several methods are available for generating composite indicators that measure global performance. The goal of this study was to compare three composite methods that are widely used in other health care domains to identify the most appropriate for trauma care process performance evaluation.

METHODS

In this retrospective, multicenter cohort study, 15 process performance indicators were implemented using data from a Canadian provincial trauma registry (19,853 patients; 59 centers) on patients with an Injury Severity Score (ISS) greater than 15. Composite scores were derived using three methods as follows: the indicator average, the opportunity model, and a latent variable model. Composite scores were evaluated in terms of discrimination, construct validity (association with an indicator of trauma center structural performance), criterion predictive validity (association with clinical outcomes), and forecasting (correlation over time).

RESULTS

All composite scores discriminated well between trauma centers. Only the average indicator score was correlated with improved structure (r = 0.29; 95% confidence interval [CI], 0.07-0.53), lower risk-adjusted mortality (r = -0.22; 95% CI, -0.46 to 0.04), and lower risk-adjusted complication rate (r = -0.48; 95% CI, -0.65 to -0.25). Composite scores calculated with 1999 to 2002 data all correlated with those calculated with 2003 to 2006 data (r = 0.49, 0.87, and 0.84 for the indicator average, the opportunity model, and the latent variable model, respectively).

CONCLUSION

Results suggest that of the three composite scores evaluated, only the indicator average demonstrates content and predictive criterion validity, discriminates between centers, and has good forecasting properties. In addition, this score is simple and intuitive and not subject to variation in weights over trauma systems and time. The observed association between higher indicator average scores and lower risk-adjusted mortality and complication rates suggests that improving process performance may improve patient outcome.