Statistical approach to quality assessment in liver transplantation

Abstract

Purpose: This study investigated the utility of retrospective two one-sided cumulative sum (CUSUM) charts combined with multivariable regression analysis in liver transplantation for transplant center benchmarking.

Methods: One thousand seven hundred and forty-nine consecutive adult primary liver transplants (January 1, 1983 to December 31, 2012) were analyzed retrospectively with two one-sided CUSUM chart analysis of 90-day mortality.

Results: Three eras and two subseries in latest era 3 were identified due to graphically delineated relevant shifts in mean 90-day mortality. Delineation of eras 1, 2, and 3 coincided with relevant changes in allocation policies. CUSUM analysis detected a resurgence of higher mean 90-day mortality in era 3 after results had improved continuously over 25 years. In era 3, two subseries were identified with improving mean 90-day mortality rates from 15.4% in subseries 1 to 8.9% in the following subseries 2. The quantitative influence of independent risk factors on 90-day mortality differed markedly between all identified eras and subseries as assessed with multivariable regression analysis deployed on era-specific subcohorts.

Conclusion: The assessed methodology is able to identify meaningful center-specific eras and subseries of liver transplantation with striking alterations of the significance and weight of outcome drivers for post-transplant 90-day mortality over time. This warrants the introduction of prospective risk-adjusted two one-sided CUSUM chart analysis into quality management in liver transplantation in Germany with the goal to obtain alarm signals as early as possible.

Keywords: Mortality; Multivariable regression; Purposeful variable selection; Risk factors for survival; Risk-adjusted analysis.