The factor analysis approach to mortality prediction in COVID-19 severe disease using laboratory values: a retrospective study

Authors : Umut KARA, Fatih ŞİMŞEK, Mehmet Özgür ÖZHAN, Mehmet Emin INCE, Gökhan ÖZKAN, Serkan ŞENKAL, Ahmet COŞAR

Pages : 528-533

Doi:10.32322/jhsm.1052191

View : 14 | Download : 9

Publication Date : 2022-03-15

Article Type : Research Paper

Abstract :Aim: Factor analysis is a statistical approach used mainly in social science scale development systems. The aim of this study was to evaluate the performance of factorial structures formed by laboratory values in predicting mortality in severe COVID-19 patients. Material and Method: The study included 281 patients diagnosed with ‘‘severe coronavirus infection’’ according to the WHO COVID-19 clinical management guideline who were treated in a 13-bed adult tertiary-level critical care unit of a tertiary level hospital. For a total of 23 variables insert ignore into journalissuearticles values(ALT, AST, BUN, creatinine, Na, K, LDH, CRP, CK, ferritin, D-dimer, INR, TB, Glu, NLR, WBC, fibrinogen, % NEU, PLT, HTC, % LYM, TLC, Alb);, laboratory values were collected. A two-step method was used to determine if exploratory factors might be used in place of laboratory variables. First, the ability of individual laboratory variables to predict mortality was obtained by analysis of the receiver operating characteristic insert ignore into journalissuearticles values(ROC); analysis. Then, the ability of factors created from these variables to predict mortality was measured using ROC analysis. The area under curve insert ignore into journalissuearticles values(AUC); values were compared between the two conditions. Results: The Kaiser-Meyer-Olkin insert ignore into journalissuearticles values(KMO); value calculated using factor analysis on the variables was found to be 0.661. The significance level of the Bartlett’s Test was 0.6. A five-factor structure was created based on the Scree Plot. The fifth factor, which included CRP, fibrinogen, and ferritin, was the highest for predicting mortality insert ignore into journalissuearticles values(AUC: 0.677);. According to the individual laboratory variables, the first factor comprising TLC, CK, and NLR, had the most remarkable success insert ignore into journalissuearticles values(AUC: 0,642);. Conclusions: The factor analysis approach can be used to present an alternative perspective for predicting mortality in COVID-19 critical disease. The factor including CRP, fibrinogen, and ferritin predicted mortality at the highest rate in this study.
Keywords : Covid 19, factor analysis, critically ill, mortality, Coronavirus disease 2019