Predicting models of inpatient death risk accompanied by coronavirus disease in healthcare establishments as an additional tool for decision-making

Abstract

we aimed to analyse risk prediction models and propose a new model for predicting in-hospital death risks. Materials and methods. We conducted a retrospective case-control study, analysing cases of hospitalisations of patients with severe and moderate COVID-19 from 2020 to 2021 (n=129). Results. We found that such factors significantly influence mortality risk: age (OR 0,866; 95% CI 0,8–0,9; p<0,001), lymphocyte absolute ratio (OR 0,000144; 95% CI 0.00000513-0.00407; p<0,001), C-reactive protein (OR 1,2; 95% CI 1,010-1,030; p<0,001), albumin baseline (OR 0,796; 95% CI 0,661-0,959; p<0,05), minimal albumin (OR 0,716; 95% CI 0,593-0,864; p<0,001), eGFR minimal (OR 0,951; 95% CI 0,93-0,972; p<0,001), INDEX PLRI score (OR 1,7; 95% CI 1,3–2,2; p<0,001), PADUA score (OR 4,49; 95% CI (2,25-8,94; p<0,001), respiratory insufficiency (OR 22,6; 95% CI (7,79-65,6; p<0,001), parenchymal involvement on multisectoral computer tomography (MSCT), % (OR 1,04; 95% CI 1,02-1,060; p<0,001), severity of lung damage on MSCT (pulmonary parenchymal involvement) over 50% (OR 4,96; 95% CI 2,08-11,8; p<0,001), hypertension in the medical history (OR 2,38; 95% CI 1,1–5,1; p = 0,026). Conclusion. We used models to predict the risk of in-hospital death. The area under the curve is 0.976, with a 95% confidence interval (CI) of 0.951-1. At the threshold point, 0.366, sensitivity is 95%, and specificity is 92,6%. We created a web version of the COVID-19 lethality calculator, which also works in Excel and could be helpful for viral or bacterial pneumonia. The calculator is available online. We propose to focus on clinical conditions and underlying comorbidities in decision-making despite the absence of data on the decompensation of diabetes mellitus, as we did not find any difference in the groups in the level of HbA1c (p=0.0662). Respiratory insufficiency could worsen progressively, so it is necessary to monitor clinical data. We analysed the presence of hypertension, diabetes mellitus and cardiovascular diseases (ischemic heart diseases, stroke, myocardial infarction, etc.) in medical history. We didn’t focus on decompensation for diabetes or destabilisation of heart diseases as in the pandemic, the presence of SARS-CoV-2 could rapidly influence the severe course of COVID-19, which was proved in numerous studies and clinical recommendations. If there are enough resources, it is advisable to hospitalise patients with noncommunicable diseases after assessment of risk before SpO2 rapid decline. In the discussable cases, a Calculator for evaluating underlying conditions could be used as an additional tool (the area under the curve is 0.766, 95% CI 0.548 - 0.984). At the threshold of 0.244, sensitivity is 87,5% and specificity – 68,8%. We suggest adding information on hospital admission criteria concerning underlying conditions rather than age factors. As in the elderly population, we received comparable results in risks in younger individuals with signs of metabolic syndrome or other non-communicable diseases. Further study is necessary to assess body mass index (BMI) as in our cohort, there was minor information on anthropological data. For a better understanding of the influence of adipose tissue on inflammatory laboratory results, we should use international study data, focus on outcomes assessment for the Ukrainian population, and assess risk individually.