A novel scoring system for early assessment of the risk of the COVID-19-associated mortality in hospitalized patients: COVID-19 BURDEN

Corona Virus Disease 2019 (COVID-19) caused by the Severe Acute Respiratory Syndrome Coronavirus-2 (SARS-CoV-2) is highly a contagious disease, with symptoms ranging from those of common flu including fever, cough, and congestion of the nasal cavity to very severe respiratory symptoms [1]. As the pandemic spread, other symptoms such as loss of taste and smell (anosmia) have also emerged [2, 3]. Patients with the severe form of the disease can experience a large range of symptoms arising from the host immune response against the infection including serious respiratory disease and pneumonia, vascular and hemodynamic disorders, and metabolic dysfunction [4‐7]. Patients can also present normal or abnormal leukocyte counts, lymphopenia, or thrombocytopenia, with extended activated thromboplastin time, and elevated C-reactive protein (CRP) [5, 8, 9]. Those most at risk are the elderly and people with preexisting medical conditions, such as cardiovascular disorders and diabetes mellitus [10‐12]. The exact mechanisms behind the disease and why some remain asymptomatic carriers, while other patients develop severe diseases with unfavorable outcomes are still poorly understood [13].

Here, we present details of the patients with laboratory-confirmed COVID-19 pneumonia to shed light on the specifications of patients who experienced in-hospital mortality and explore the risk factors that might facilitate early screening. Finally, we develop a predictive model for early detection of COVID-19 patients with a high risk of mortality using analysis of clinical features and laboratory data on admission, allowing timely identification and intervention.

We retrospectively investigated 480 adult patients, aged 21–95, who were admitted to Faghihi Teaching Hospital from September 23, 2020, to November 21, 2020, with a positive RT-PCR COVID-19 test. The patients who did not complete their course of hospitalization and were released with their own consent were excluded. The study protocol was approved by the Ethics Committee of Shiraz University of Medical Sciences (IR.SUMS.MED.REC.1400.382). Written informed consent was obtained from all participants on admission.

Corresponding medical records were thoroughly examined. Variables investigated in this study for correlation with the outcome of hospitalization (discharge or death) included demographics (gender and age), previous medical conditions (smoking/opium abuse, hypertension, HTN, and diabetes mellitus, DM, ischemic heart disease, IHD, and hyperlipidemia, and HLP), previous drug history (angiotensin-converting-enzyme inhibitors/angiotensin II receptor blockers; ACEIs/ARBs, calcium channel blocker; CCB, beta blockers; BB, acetylsalicylic acid; ASA, and Statins) the patients’ condition on admission (systolic blood pressure; SBP, diastolic blood pressure; DBP, pulse rate; PR, respiratory rate; RR, temperature; T, O₂ saturation; O₂ Sat, presence of dyspnea, cough, chest pain, fever, malaise, anorexia, nausea and/or vomiting; N/V, other gastrointestinal symptoms, and the interval between disease onset and hospital admission), the laboratory findings of the sample obtained on admission (white blood cell count; WBC, absolute neutrophil count; ANC, absolute lymphocyte count; ALC, hemoglobin; Hb, prothrombin time; PT, partial thromboplastin time; PTT, blood urea nitrogen; BUN, creatinine; Cr, sodium; Na, potassium; K, aspartate transaminase; AST, alanine transaminase; ALT, alkaline phosphatase; ALP, albumin; Alb, total bilirubin; TB, direct bilirubin; DB, creatinine phosphokinase; CPK, lactate dehydrogenase; LDH, erythrocyte sedimentation rate; ESR, C-reactive protein; CRP), and total length of stay in the hospital.

The COVID-19 pandemic is a public health concern with dire health, environmental, and economic consequences [15]. Currently, the facilities required for the hospitalization of patients with COVID-19 are limited. This limitation is especially pronounced in ICUs for patients requiring mechanical ventilation [16]. It is necessary to develop prediction models that can be used by clinicians, to estimate the severity of the disease and prognosis of each patient at an early stage of the disease, which will grant a better allocation of resources and a wider window for interventions. This tool also potentially helps to reduce hospital care costs and improves its quality in the health care units [17‐19]. Applying such prediction models is a useful strategy for the early screening of high-risk patients in crowded care centers during the COVID-19 outbreak. Risk prediction models are increasingly utilized in medical practice to help practitioners promote healthcare quality. Following extensive validations in independent samples, a predictive score can be used routinely.

Thus far, several prognostic models with varying clinical applicability and performance have been developed for COVID-19. Among them, a few have a low risk of bias and low concern for applicability. The most well-validated models are developed by Clift et al. and Knight et al. with the potential of being applied by clinicians in public health and clinical settings [30, 31]. Age, sex, RR, O₂ sat, level of consciousness, BUN, BMI, and comorbidities are among the variables incorporated into the aforementioned models. On top of that, a few models developed before the pandemic were applied to COVID-19 prognostication. Most notably APACHE II had very promising results (area under the curve, 0.966) in predicting in-hospital mortality [32].

In this survey, we designed a simply calculated clinical risk score, using the patient's clinical characteristics and laboratory data, which can be used as a predictor to predict mortality, the most important outcome in the course of hospitalization and implicitly related to the severity of the disease, the need for hospitalization, and the possibility of ICU admission requirement during the hospitalization. Low DBP (≤ 75 mmHg), prolonged PT (> 16.2 s), increase in the level of LDH (> 731 U/L), BUN > 23 mg/dl, elevated CRP (> 73.1 mg/L), and decrease in oxygen saturation (< 84%) were identified as risk factors for disease severity among 480 adult patients.

Severe hypoxemia (O2 sat < 84%) had the highest odds ratio (OR = 9.19) among the 6 risk factors determining the mortality in our survey. This finding was consistent with a cohort study by Bahl et al. among 1461 patients in which O2 sat of ≤ 88% was associated with a higher mortality rate [20]. In the review conducted by Petrilli et al. among 4103 patients, oxygen saturation of < 88% (OR = 6.99) was also introduced as the most critical factor predicting the outcome of hospitalization on admission [21].

Recently, several studies have been conducted in different parts of the world to develop a simple scoring system to predict the prognosis and outcome of COVID-19 [19, 22‐25]. The first scoring system to predict the severity of COVID-19, incorporating age, glomerular filtration rate (GFR), WBC, neutrophil count, and myoglobin, was developed by Zhang et al. in 2020 among 80 patients [22]. The low sample size was one of the limitations of their study. Moreover, the aforementioned model did not include any of the vital signs, which were shown to be quite important in determining the clinical course of COVID-19 patients. The findings of the study conducted by Altschul et al. are also in agreement with the present study. Old age (especially ≥ 80 years) mean arterial blood pressure ≤ 60 mmHg, O₂ sat < 94%, BUN > 30 mg/dl, CRP > 10 mg/dl, and INR > 1.2 were identified as six risk factors that affect the mortality rate of COVID-19 patients. Their study was conducted in three major referral hospitals in New York City. However, the study data is limited to their urban population and may not be fully generalizable to other settings with different backgrounds [24]. In another study which was published in July 2020, old age, the presence of coronary heart disease, high procalcitonin, lymphopenia, and high d-Dimer level were associated with a high mortality rate in the affected population [25].

Having a precise, inexpensive, accessible, and straightforward prediction model will improve the patients' triage, and help to identify high-risk individuals, initiate timely interventions, and allocate resources efficiently [22, 26]. One of the advantages of our model is that it does not include lung high-resolution computed tomography (HRCT) scans, which contributes greatly to the health financial system during this pandemic. Rather, we used more accessible clinical and laboratory metrics to predict disease severity, which will reduce the costs, improve the timing of the decision-making in the initial screening, and lift the burden from the already exhausted imaging facilities. HRCT can be conducted later during the hospitalization of the patients rather than on admission in the high risk patients; however, the present study is not concerned with this issue and this hypothesis requires further investigation on the effect of applying prediction models on the burden imposed on various units in hospitals.

CRP, Oxygen saturation variation, increased PT, DBP, and BUN, and raised LDH (COVID-19 BURDEN) were detected as six factors for COVID-19 infection severity. Using this prediction model, the severity of the disease in the early stages of the disease can be estimated and helps to reduce health costs and improve the quality of patient care in the health care units. The great utility of clinical prediction models have been demonstrated previously in several instances (e.g., CURB65, MELD, etc.) [27, 28].

The present survey was mainly limited by studying COVID-19 patients in a single center retrospectively. The management of the patients included in the present study was based on the most recent version of the WHO guideline for COVID-19 management at the time (https://​www.​who.​int/​publications/​i/​item/​WHO-2019-nCoV-Clinical-2022.​2) [29]. Furthermore, our model did not make use of viral load in predicting the mortality in COVID-19 patients. Viral load is among the most important parameters in the setting of COVID-19 infection; however, it is mostly limited to experimental settings and not implemented in the real-world clinical setting. Moreover, it is not easy-to-obtain and not available early in the admission of the patients, if at all. As a result, incorporating it into a screening model will defeat its purpose and make triage based on the model impossible. It is quite common for patients to receive non-invasive ventilation or high-flow nasal cannula before their screen either in the emergency department or during transportation with emergency medical service. In our study, although the clinicians conducting the primary screening abided by the recommendations for cessation of therapy before recording O₂ saturation, it is plausible to assume that the O₂ saturation was falsely recorded in a set of patients, inclining toward a higher O2 saturation. Undoubtedly, further studies with larger sample sizes are needed in settings with diverse backgrounds to validate the COVID-19 BURDEN model. Moreover, many more studies are still required to better understand the disease itself. Finally, simple models similar to that outlined in the present paper can be beneficial in other aspects of the care for COVID-19 patients and their needs including ICU admission requirements, unplanned readmission, its long-lasting health effects, etc.