A new COVID-19 prediction scoring model for in-hospital mortality: experiences from Turkey, single center retrospective cohort analysis

Abstract

Objective: Although many studies reported prognostic factors proceeding to severity of COVID-19 patients, in none of the article a prediction scoring model has been proposed. In this article a new prediction tool is presented in combination of Turkish experience during pandemic.

Materials and methods: Laboratory and clinical data of 397 over 798 confirmed COVID-19 patients from Gülhane Training and Research Hospital electronic medical record system were included into this retrospective cohort study between the dates of 23 March to 18 May 2020. Patient demographics, peripheral venous blood parameters, symptoms at admission, in hospital mortality data were collected. Non-survivor and survivor patients were compared to find out a prediction scoring model for mortality.

Results: There was 34 [8.56% (95% CI:0.06-0.11)] mortality during study period. Mean age of patients was 57.1±16.7 years. Older age, comorbid diseases, symptoms, such as fever, dyspnea, fatigue and gastrointestinal and WBC, neutrophil, lymphocyte count, C-reactive protein, neutrophil-to-lymphocyte ratio of patients in non-survivors were significantly higher. Univariate analysis demonstrated that OR for prognostic nutritional index (PNI) tertile 1 was 18.57 (95% CI: 4.39-78.65, p<0.05) compared to tertile 2. Performance statistics of prediction scoring method showed 98% positive predictive value for criteria 1.

Conclusions: It is crucial to constitute prognostic clinical and laboratory parameters for faster delineation of patients who are prone to worse prognosis. Suggested prediction scoring method may guide healthcare professional to discriminate severe COVID-19 patients and provide prompt intensive therapies which is highly important due to rapid progression leading to mortality.