Discussion
The main results of this study indicate that advanced age and multiple comorbidities (hypertension, diabetes, cardiovascular disease, respiratory system disease, and past surgical history) are risk factors for severe patients with COVID-19. To enhance the clinical use of this risk model, we developed a risk prediction model for COVID-19 patients. To the best of our knowledge, this is the first study to attempt to establish a risk prediction model of severe patients with COVID-19.
Age, as a risk factor, for critically ill patients, has been reported in several studies (Jordan et al., 2020; Leung, 2020; Li et al., 2020; Liu et al., 2020; Wang et al., 2020). Older patients have reduced body immunity, reduced multi-organ function, and the risk of developing severe illness increases with age. This study also performed a ROC analysis on the risk factor of age (Figure S1), and its AUC reached 0.922 (95% CI = 0.852-0.991). When age is 55 years old, the Youden index reaches a maximum value of 0.711. When the patient is older than 55 years old, the risk of converting to severe illness increases significantly. In the process of modeling, we also tried to use age as a categorical variable and divide it into two groups, older age, and younger age. Although it is more convenient for clinical application when converting a continuous variable into a categorical variable, it will lose the prediction accuracy of the model. Therefore, continuous variables were still used for logistic regression analysis.
When performing the single factor analysis, there are significant differences between severe and mild patients with regards to hypertension, diabetes, cardiovascular disease, respiratory system disease, and past surgical history (P <0.1). The above comorbidities have been reported in the literature as risk factors (Aghagoli et al., 2020; Cook, 2020; Jordan et al., 2020; Klonoff et al., 2020; Leung, 2020; Li et al., 2020; Li et al., 2020; Li et al., 2020; Liu et al., 2020; Liu et al., 2020; Long et al., 2020; Wei et al., 2020; Yang et al., 2020; Zhang et al., 2020; Zheng et al., 2020; Zhou et al., 2020; Zhou et al., 2020). However, when performing multivariate logistic regression, these factors were not included in the final model, which may have a great relationship with the small sample size of this study. Therefore, we combined these comorbidities to reduce the dimensions and analyzed the number of comorbidities as risk factors. The results showed that the number of comorbidities had a significant impact on the risk of serious illness (OR = 6.067,95%CI = 1.078-34.143). The AUC obtained by the ROC analysis reached 0.871 (95% CI = 0.772-0.970) (Figure S2). Because our patients had a maximum of three comorbidities, the risk prediction model was given a maximum of 6 points (2 points for each comorbidity). It should be noted that two of the included patients had a stroke, two had chronic gastritis, one had hepatitis B, and one had gout. Because the sample size of these complications was too small, they were not considered when performing data analysis due to concern about introducing these into the model would lead to model deviation.
Several studies have reported that biochemical indicators are risk factors for severe patients (Ji et al., 2020; Lin et al., 2020; Lippi et al., 2020; Liu et al., 2020; Long et al., 2020; Mehra et al., 2020; Yao et al., 2020; Zhu et al., 2020). When extracting the data, we also obtained some laboratory indicators of the patient at the time of admission, but we did not obtain the full data of these indicators. Because some indicators might change during hospitalization, these factors were not included in the model when the model was built. From the data of mild and severe patients, leukocytes, calcitonin, and C-reactive protein all have significant differences (P <0.05), suggesting that clinical attention should be paid to these values. When the above indicators increase, it may indicate that the patient’s risk of becoming severe is increased or it has developed into a severe case. In addition, we believe that the changes in laboratory indicators may be due to changes in the patient’s condition; they may not be the cause of leading to severe coronavirus disease. That is, the changes in the index of the laboratory examination are the result of the difference in the disease condition. Due to our limited data, our view needs to be supported by more data.
It has been reported in the literature that damage to liver function (Boettler et al., 2020; Lei et al., 2020) and kidney function (Cheng et al., 2020) is a serious risk factor. However, in our study, there is only one patient with impaired renal function and three patients with mildly impaired liver function in the severe-group. The sample size is too small to obtain positive results. Univariate analysis found that the patients’ creatinine levels were significantly different between mild and severe patients, and the median levels were within the normal range in these two groups.
This study has the following limitations 1) our hospital might have given priority to treat more severe patients, which might lead to a lower proportion of mild cases in our study, 2) the sample size is limited, and our results need clinical validation, and 3) some continuous variables do not conform to the normal distribution, which will affect the power of statistical results.
In summary, advanced age and multiple comorbidities are risk factors for severe patients with COVID-19. This study provides a relatively practical tool for clinicians to identify severe high-risk cases. More attention should be paid to high-risk patients before they develop into severe cases.