Risk score development
To facilitate the clinical usage, each risk factor was assigned a risk
score according to the multivariable logistic regression results
(Table 2 ). The smallest OR value, age variable, was selected as
a reference value. Considering that the regression coefficient of age is
very small, 0.093, we divided age into several age groups based on the
unit of 10 years. The youngest patient in this study was 20 years old,
and 2 points were assigned to the 20-29 years old patients. The score
increases by 1 point for every 10 years of age, and so on. The
regression system of age was used to calculate the relative multiple of
the comorbidity regression coefficient. After rounding, the scores for
comorbidities were clearly defined. Each additional comorbidity scores 2
points with a maximum of 6 points under comorbidity. In this study, the
maximum number of comorbidities was 3, and the model’s prediction range
should not exceed the data range used to develop the model.Table 3 shows the detailed score assignment rules.
The total risk score ranged from 2 to 14, with corresponding predicted
probabilities of severe patients ranging from 2.11%-99.98%
(Figure 2 ). The risk scores were divided into three levels
(low-risk, moderate-risk, high-risk) to facilitate the clinical use of
this risk model: low-risk (score 2-4, severe patients incidence
3.47%-17.92%), moderate-risk (score 5-6, severe patients incidence
35.62%-56.98%), and high-risk (score 7-14, severe patients incidence
77.05%-99.95%).