By Sharen Lee, Jiandong Zhou, Qingpeng Zhang and Gary Tse, on behalf of all authors
The coronavirus disease 2019 (COVID-19) has a wide range of clinical manifestations, ranging from asymptomatic to requiring life-sustaining intubation and ventilatory support. Although those who suffer from a severe disease represent only a small proportion of COVID-19 patients, it is the sudden spike of affected individuals requiring intensive care that places overwhelming burden on the healthcare systems. Thus, ever since the initiation of the pandemic, ample research efforts have been put into identifying predictors for severe COVID-19. However, few easy-for-use risk models for severe COVID-19 manifestation applicable for clinical practice are available. The majority of the current models require the interpretation of radiographic findings and details of clinical assessment, which can be difficult to execute when the healthcare system is being overwhelmed, resulting in little resources accessible for the initial assessment of patients.
To tackle this issue, we have developed a predictive risk score for severe COVID-19 based on patient demographics, comorbidities, and laboratory parameters, hence allowing the identification of patients at risk of deterioration before they manifest clinically. In our paper “Development of a multivariable prediction model for severe COVID-19 disease: a population-based study from Hong Kong”, we have analyzed consecutive COVID-19 patients in the Hong Kong region of China between January to August 2020 to identify risk factors using Cox regression, then developed the score based on the beta-coefficients found through logistic regression using these predictors. Severe COVID-19 manifestation was defined as the composite outcome of need for intensive care admission, intubation, or all-cause mortality. The score was then evaluated by five-fold cross-testing within the sample, and externally validated by a cohort from the Wuhan Asia General Hospital.
Our risk score based on age, gender, comorbidities, medication prescribed and laboratory parameters on the day of admission achieved an area-under-the-curve (AUC) of 0.85 and 0.89 under in-sample cross-testing and external validation respectively, indicating high sensitivity and specificity. Although further external validation from other populations is needed, the excellent predictive performance of our score demonstrates that the combination of simple baseline clinical and laboratory parameters can provide accurate predictions towards patient outcomes in COVID-19. It is particularly useful when COVID-19 patients can deteriorate rapidly, therefore the initial disease manifestation may not be a reliable indicator of the ultimate patient outcome. Without the inclusion of radiographic findings in the risk score, the risk stratification of COVID-19 patients can be achieved at a faster rate, since the availability of imaging is also a rate-limiting factor in patient management under the high patient load. To further increase the accessibility of our risk score, it is now included on WebMD, hence allow for easy access at any time and place as a free and easy-to-use risk-stratification tool.
Furthermore, given the strain of medical resources under the state of the pandemic, a risk-stratifying score that only requires accessible baseline investigations can serve as a screening tool for patient admission, especially since our risk score uses clinical and laboratory parameters at the time of admission. To free up space, staff, and resources in the hospital, low-risk patients can be monitored at home through telemedicine, since those of a mild disease course will likely recover fully with little medical attention needed. High-risk patients may also benefit from closer monitoring in anticipation of the likely clinical deterioration, thus achieve life-saving purposes. However, it should be kept in mind that there is always a risk of false-negative in all screening tools- a score indicative of low risk for a severe disease does not guarantee the patient to be free from adverse outcomes. Therefore, continuous symptoms monitoring with awareness to seek help at signs of deterioration.
To conclude, we have developed a simple clinical score for the risk stratification of severe COVID-19 manifestation based on clinical and laboratory parameters on admission. The excellent predictive performance of the risk score demonstrates that the combination of predictors from one’s demographic, past medical history, and baseline laboratory findings can predict disease deterioration before clinical signs and symptoms appear. The easy-to-use risk score, now available on WebMD, can serve as an accessible screening tool for the early direction of resources to these high-risk individuals and improve their prognosis.
Zhou, J., Lee, S., Wang, X., Li, Y., Wu, W.K.K., Liu, T., Cao, Z., Zeng, D.D., Leung, K.S.K., Wai, A.K.C., Wong, I.C.K., Cheung, B.M.Y., Zhang, Q.*, Tse, G.* (2021) Development of a multivariable prediction model for severe COVID-19 disease: a population-based study from Hong Kong. NPJ Digital Medicine. https://doi.org/10.1038/s41746-021-00433-4.
Please sign in or register for FREE
If you are a registered user on Nature Portfolio Health Community, please sign in