Development and External Validation of a Prediction Model for Critical Outcomes of Patients With Confirmed COVID-19 in the Hospital Setting Based on Demographic Characteristics, Medical Conditions and Oral Health

Background The novel coronavirus disease 2019 (COVID-19) pandemic has presented an important and urgent threat to global health since its outbreak in December 2019. The COVID-19 does not only affect the respiratory tract, but also other organs in human body including lungs, liver, kidney, heart, vessels, and other organs (1). Respiratory failure and acute respiratory distress syndrome (ARDS) are the most common serious complications of COVID-19 infection (2). The crude mortality rates (CMRs) of COVID-19 varies in different locations, which ranges from 37.0/100,000 to 167.6/100,000 in several European countries up to 30th August 2020 (3). The in-hospital mortality of COVID-19 was reported to be 17.1% based on 33 studies from 13,398 patients (4), and it was 2.9 times higher than that of influenza based on the French national administrative database (5). It was reported that 26% of the COVID-19 patients admitted to ICU with severe status and 31% of the patients who admitted to ICU died based on 37 studies from 24,983 patients (6).

There are several risk factors on patients' demographic characteristics and underlying medical conditions which were shown be associated with the critical outcomes of COVID-19 (7). In addition, poor oral health, in particular periodontitis, was also shown to be associated with the critical outcomes of COVID-19 (8). Marouf et al. showed that COVID-19 patients with periodontitis had 8 times higher odds of death and 3.5 times higher odds of ICU admission than those without periodontitis based on a case-control study (9). This may be because periodontal disease could enhance cytokine release via altered microflora, expression of multiple viral receptors, bacterial superinfection, and aspiration of periodontal pathogens (10). The increased production of pro-inflammatory cytokine, which is referred to as cytokine storm, is the foremost cause of the adverse events of COVID-19 (10).

Because of the high contagiousness, high ICU admission rate, and high mortality of COVID-19, it has led to tremendous increases in the demand for hospital beds and shortage of medical equipment. Therefore, there is an urgent need for a pragmatic risk stratification tool that allows the early identification of the COVID-19 patients who are likely to be at highest risk of ICU admission and death (11). This can help clinicians and policymakers make the decisions on the management and optimize resource allocation. A recent review identified multiple prediction models which have been developed for prediction of prognosis of COVID-19 patients (12). Those prediction models varied in their predictors and performance of the models. A large number of prediction models reflected difficulties in their application in the rapid risk stratification for general COVID-19 patients at their first intake in hospitals because some predictors cannot be easily obtained without professional devices or lab tests, such as C reactive protein, peripheral oxygen saturation, and urea level. Many prediction models showed moderate performance in aspects of discrimination and calibration, and no benefit to clinical decision making (12). In addition, the dental variables were not considered the potential predictors in the previously developed models.

Therefore, the aim of the present study is to develop and externally validate a prediction model for the critical outcomes of COVID-19 patients using predictors which can be easily obtained in clinical practice, including patients' demographic characteristics, self-reported medical conditions, and oral health.

Materials and Methods Participants We include hospitalized patients and outpatients from the Isala Hospital (Zwolle, the Netherlands) with confirmed COVID-19 who visited the Department of Oral and Maxillofacial Surgery (OMFS) between March 2020 and May 2021, if they have had a dental panoramic radiograph (OPG), obtained up to a maximum of 5 years until the end of the current study. The patients are used to develop the prediction model (derivation cohort).