Development of a preliminary index that predicts adverse events after total knee replacement

Objective. We undertook this study to assess the relationships between hospital characteristics, volume of procedures, and perioperative outcomes after total knee replacement (TKR), and to use this information to construct a simple predictive index for perioperative outcomes that incorporates hospital- and patient-level characteristics. Methods. We studied Medicare beneficiaries who underwent TKR in 4 US states in 2000. Orthopedic surgery administrators from hospitals caring for patients in this sample were surveyed about a range of hospital characteristics. The relationships between these hospital characteristics, patient variables, and 90-day postoperative adverse events (including death, pulmonary embolus, pneumonia, deep wound infection, and acute myocardial infarction) were assessed using generalized estimating equations adjusting for hospital volume. These relationships were assessed in low- and high-risk patient groups. Variables from the final multivariate model were used to create an index that was tested against 90-day adverse event rates. Results. Three hundred twenty-seven (3.6%) of the patients undergoing TKR in our sample experienced an adverse event. In the final multivariate regression models, variables that predicted adverse perioperative events included low hospital volume (fewer than 23 TKRs in the Medicare population per year), absence of a preoperative teaching program, fewer TKRs conducted in a dedicated orthopedic surgery operating room, patient age > 70 years, male sex, and at least 1 comorbid condition. The effect of volume on perioperative adverse events was evident both in patients with few risk factors and in patients with several risk factors. An index including the 6 patient and hospital variables discriminated well, with adverse events occurring in 2.0% (95% confidence interval [95% CI] 1.4-2.7%) of patients in the lowest risk category and in 7.4% (95% CI 4.5-12.3%) of patients in the highest risk category (P for trend < 0.001). The index predicted adverse event rates both in hospitals with a low volume of TKRs and in those with a high volume of TKRs. Conclusion. Characteristics of hospital care, procedure volume, and patient-level factors are all associated with perioperative outcomes of TKR. A preliminary index combining hospital characteristics and volume is moderately predictive of adverse perioperative outcomes. The index is predictive of outcome in low- and high-volume hospitals.