Predicting outcomes: Is there utility in risk scores?

This review shows how accurate predictions of perioperative risk can help inform patients and clinicians with respect to decision-making around surgery; identify patients who warrant further specialized investigations, new interventions intended to decrease risk, modifications in planned operative procedures, or intensification of postoperative monitoring; and facilitate fairer comparisons of outcomes between providers and hospitals. A preoperative risk score formally integrates several pieces of clinical information (e.g., age, comorbid disease, laboratory tests) to arrive at an overall estimate of an individual patient’s expected risk for specific postoperative adverse events. A good risk score should be simple to incorporate in clinical practice, reliable when applied by different raters, and accurate at predicting postoperative risk. Several analytical methods (e.g., receiver operating characteristic curves, likelihood ratios, risk reclassification tables, observed vs predicted plots) are required to characterize the relevant domains that encompass the prognostic accuracy of a risk score. External validation is critical in determining whether the predictive accuracy of a risk score is preserved when applied to new settings, populations, or outcome events.