Stability of county rankings in liver transplant outcomes

Abstract

Background: Evidence of geographic differences in liver transplantation (LT) outcomes has been proposed as a reason to include community characteristics in risk-adjustment of transplant quality metrics. However, consistency of rankings in LT outcomes for small geographies, such as counties, has not been demonstrated. We sought to evaluate the stability of county rankings on 1 year survival after LT. Methods: We used United Network for Organ Sharing data to identify adults ≥18 years of age undergoing first-time LT in 2002–7 (era 1) and 2008–13 (era 2). County-specific 1-year survival was calculated using the Kaplan-Meier method for counties with ≥5 LT performed in each era. Agreement between high-risk designation according to era was calculated using the kappa statistic and Spearman correlation coefficient. Results: The analysis included 772 counties, with high mortality risk (bottom decile) defined as ≥25% mortality in era 1 and ≥20% mortality in era 2. County 1-year mortality rates were not correlated between the two eras (Spearman ρ = 0.03, p = 0.474). Comparing high risk designation across eras, 142 (18%) counties were classified as high-risk in one but not both eras; whereas only 10 (1%) counties were classified as high-risk in both eras. The kappa statistic (κ = 0.02, p = 0.261) confirmed poor agreement between era-specific designations of county high risk. Conclusion: There was poor agreement on county rankings of LT outcomes across transplant eras, with very few counties consistently classified as high risk. The instability of county-level outcomes suggests limited utility for county characteristics to predict excess risk of 1-year mortality after LT.