Pulled from PubMed
Manuel Castillo-Angeles 1, Tanujit Dey 2, Jill Stinebring 3, Ali Salim 4, Joel T Adler 5
Background: Efforts to ameliorate the current organ shortage have predominantly focused on improving processes among the nation’s organ procurement organizations. However, there is no consensus on how to measure the quality of an individual hospital’s organ donation process. Therefore, we sought to create a unified measurement, utilizing existing metrics, to give a holistic view of hospital organ donation performance.
Methods: This composite score was developed through a multistage process: (1) an already-conducted extensive literature search and systematic review to identify measures of organ donation performance; (2) a modified-Delphi panel to determine the most relevant variables; (3) a random forest machine learning algorithm to rank these consensus metrics based on variable importance; and (4) an unsupervised learning method, principal component analysis, to construct the composite score and generate rankings for 204 hospitals within a single local organ procurement organization.
Results: Our initial systematic review identified several metrics, of which the Delphi panel reached a consensus on 12 organ donation metrics, 10 hospital characteristics, and 4 interventions. Machine learning analysis identified the final variables to be included in the score (10 metrics and 2 hospital characteristics). The most important variables were trauma center designation, the presence of an organ donation council, and the proportion of brain death testing. Using organ donation data from hospitals within a single organ procurement organization, principal component analysis was then applied to construct the composite scoring system and rank hospitals accordingly.
Conclusion: Using a novel approach that combines machine learning and unsupervised learning, we integrated key metrics into a composite score to assess donor hospital performance.