2015 ISAKOS Biennial Congress ePoster #1444
Threshold Analysis of the Volume-Outcomes Relationship in Total Knee Arthroplasty
Sean Wilson, New York, NY UNITED STATES
Robert G. Marx, MD, MSc, FRCSC, New York, NY UNITED STATES
Ting Jung Pan, MPH, New York, NY UNITED STATES
Stephen L. Lyman, PhD, New York, NY UNITED STATES
Hospital for Special Surgery, New York City, New York, USA
FDA Status Not Applicable
Summary: Studies have associated higher surgeon volumes with better outcomes for many procedures, including total knee arthroplasty (TKA), but common approaches to volume stratification are arbitrary. Stratum specific likelihood ratio (SSLR) analysis of a receiver operating characteristic (ROC) curve offers a rigorous statistical approach to volume stratification and established meaningful thresholds.
Introduction/Objective: In total knee arthroplasty (TKA), higher surgeon procedural volume has been associated with lower mortality, infection, and transfusion rates, shorter procedure times, decreased lengths of stay, and superior patient-reported outcomes. Although procedural volume is gaining importance as a convenient structural measure of quality for numerous procedures, existing studies rely on arbitrary volume thresholds to distinguish between “high,” “medium,” and “low” volume surgeons. In order for volume-outcomes research to have clinical utility, a more rigorous statistical approach to volume stratification is required. Receiver operating characteristic (ROC) curves are used routinely in diagnostic testing research to establish thresholds for test results. Stratum specific likelihood ratio (SSLR) threshold analysis has been proposed as an alternative approach to ROC curve analysis. Applied to volume-outcomes research, SSLR analysis has the potential to generate multiple discrete volume strata.
Material/Methods: Using an administrative database, we identified 233,859 patients who underwent a single primary TKA between 1997 and 2009 in New York State, USA. Stratum specific likelihood ratio (SSLR) threshold analysis of a receiver operating characteristic (ROC) curve established multiple discrete volume strata with significantly different risks of revision within two years. We used a Cox proportional hazard model and logistic regression to control for various patient level factors and measure the effect of these volume strata on the risks of early revision, 90-day complication, and 90-day mortality.
SSLR analysis of the ROC curve for early revision rates by surgeon volume generated three discrete volume strata: <13 TKAs per year (low volume), 13-59 (medium), and 60+ (high). Unadjusted revision rates were 3.3%, 2.6%, and 2.0% for low, medium, and high volume surgeons, respectively. Early revision rates were significantly higher for low (HR 1.54[1.42,1.68]) and medium volume surgeons (HR 1.24[1.17,1.32]) compared to their high volume counterparts. The rates of complication within 90 days were 10.4%, 8.7% and 7.1% for low, medium, and high volume surgeons, respectively . Rates of complication were significantly higher for low (OR 1.64[1.56,1.71]) and medium volume surgeons (OR 1.32[1.28,1.36]) compared to high volume. 90-day mortality was similarly higher for low (OR 1.90[1.53,2.34]) and medium volume surgeons (OR 1.18[1.00,1.39]) compared to high volume, with unadjusted mortality rates of 0.51%, 0.29%, and 0.21%, respectively.
Discussion/Conclusion: SSLR analysis of the ROC curve for early TKA revision rates established meaningful surgeon procedural volume thresholds at 13 and 60 TKAs per year. Higher surgeon volume strata were associated with lower risks of 2-year revision, 90-day complication, and 90-day mortality after TKA when compared to lower volume strata. Our findings reinforce existing evidence of the surgeon volume effect in TKA surgery and support SSLR analysis as a more refined and rigorous statistical approach to volume stratification and an improvement over arbitrary threshold assignment.