In this study, we identified nine routine clinical indexes, including age, dyspnea, anorexia, NLR, PLT, AST, ALB, and CRP as independent risk factors for COVID-19 fatality. Two predictive score nomograms were established based on the above variables to evaluate the mortality risk of COVID-19. Nomo1 and Nomo2 predicted mortality with a larger AUC than that for the MuLBSTA score in the training cohort, also with satisfactory AUCnomo1 of 0.92 (95% CI 0.86–0.98) and AUCnomo2 of 0.89 (95% CI 0.83–0.96) in external validation cohort. To the best of our knowledge, this is the largest study with external validation aimed to construct nomograms that includes routine clinical indicators to predict COVID-19 mortality risk.
Age is an acknowledged risk factor associated with disease severity and prognosis of COVID-19 [16]. In our study, the median ages in survivors and non-survivors were 60.0(49.3 ~ 68.0) and 69.0(62.0 ~ 78.0), respectively. A retrospective study of 1099 COVID-19 patients carried out in China showed the median ages were 45.0 (34.0–57.0) in non-severe patients and 52.0 (40.0–65.0) in severe patients [5]. Mortality increased sharply to 7.8% in aged patients over 80, while the overall death rate was estimated to be 0.66% [17]. Similar results have been characterized in other studies [18]. Moreover, a study of 577 patients identified age over 60 as an independent prognostic factor for 12-day mortality [19].
Although within the normal range, in our study, WBC and neutrophil counts were significantly higher in non-survivors when admitted. We also noted lymphocytopenia in non-survivors. Lymphocytopenia had been reported as a characteristic of COVID-19 since the lymphocyte count in ICU patients was 0.4 (0.2–0.8) compared with 1.0 (0.7–1.1) in non-ICU patients [20]. Furthermore, lymphocyte count was integrated in a predictive model for COVID-19 fatality [21]. The net effect of elevated neutrophils and decreased lymphocytes resulted in raised NLR. In other studies, NLR ≥ 2.22 had been used to recognize COVID-19, and NLR ≥ 4.06 was an indicator of severe disease [22]. In our study, both WBC and NLR were identified as independent risk factors associated with COVID-19 mortality, we included NLR in the reduced model for the robustness of NLR used to predict mortality risk of COVID-19 in other study [11].
CRP is elevated in response to inflammatory disease to protect against infection, to clear damaged cells and to regulate the inflammatory response. In our study, the CRP was significantly elevated in nonsurvivors compared with survivors (52.38 (7.76 ~ 132.7) vs 2.46(0.92 ~ 10.07) mg/L). CRP is significantly elevated in deaths compared with recovered patients with severe diseases (113 [69.1-168.4] vs 26.2 [8.7–55.8]) in a retrospective study delineating the clinical characteristics of 113 nonsurvivors with COVID-19 [23].
Dyspnea is a symptom which can reflect the severity of the disease. Dyspnea has been reported to be associated with increased risk of developing ARDS in another study [24]. We observed that 18 (25.35%) non-survivors were common type when admitted, who reported symptoms of dyspnea but not respiratory distress, ahead of their subsequent disease progression into severe type. The involvement of dyspnea in the score compensates for underestimating death in patients in the early stage before disease progression.
PLT count was reported to be lower in severe COVID-19 patients [5]. In our study, we also found that PLT was lower in nonsurvivors compared with survivors (190 [87 ~ 265] vs. 227 [184 ~ 280], p < 0·01). However, PLT was reported to be significantly higher in ICU patients than in non-ICU patients [20], and the study carried out in Jin Yin-tan Hospital that enrolled 52 critically ill patients showed that the non-survivors had elevated PLT [25]. We speculate that the difference between studies may be influenced by selection bias and the number of patients enrolled. In our study, PLT is a protective factor for survival.
Gastrointestinal involvement has been observed in COVID-19 patients [26–27]. Anorexia was reported to be associated with ICU admission for the patients with COVID-19 [28]. In our study, anorexia is an independent predictor of death. Regarding liver function, AST and TB were higher while ALB was lower in non-survivors compared with survivors, but ALT was insignificant between the two groups. A meta-analysis including 35 studies of 6686 COVID-19 patients showed that the pooled prevalence of digestive system comorbidities was 4%, and the pooled prevalence of abnormal liver function was 19%. ALT, AST, and TB predicted severe cases with pooled ORs of 1.89 (1.30–2.76), 3.08 (2.14–4.42) and 1.39 (0.78–2.47), respectively [24].
The MuLBSTA score was previously used to predict the mortality risk of viral pneumonia. This model was established mainly by patients with influenza pneumonia and other viral pneumonias except for SARS-CoV-2 infection. Seven parameters, including multilobular infiltrates, lymphocytes, bacterial coinfection, acute smoker, former smoker, hypertension and age ≥ 60 years were included. It has been reported that the deaths with COVID-19 had high MuLBSTA scores [29]. In our study, non-survivors had higher MuLBSTA scores than survivors (11[7 ~ 13] vs 7[5 ~ 9], P < 0.001). Although the MuLBSTA score was higher in non-survivors, the AUC of the MuLBSTA score was 0.814 (95% CI 0.76–0.868), with a sensitivity of 40.91% (28.79–53.03%). The poor sensitivity of the MuLBSTA score made it unsuitable for prediction of death. By adjusting the optimal cut-off value from the reported 12 to 10.5 according to the Youden Index, the sensitivity of the MuLBSTA score increased to 66.67% (95% CI 54.55%-77.27%). Compared with the MuLBSTA score, besides the advantages of our model in discrimination and calibration, our models calculated individualized death probability rather than assigning cases into low or high-risk groups.
The present study has several advantages. First, the study has a large sample size, recruiting patients from late January to early April, representative of the COVID-19 epidemic in Wuhan, China. Second, all variables included in the nomogram are routine clinical indexes, making it applicable in most medical institutions worldwide. Third, two nomograms were built for different purpose, Nomo1 is more robust and Nomo2 is more convenient, clinician may choose either one appropriate according to the situation.
This study also has some limitations. First, it is a retrospective study, bias is inevitable, the results should be interpreted carefully as an exploratory study. Second, since the study was carried out in a single city (Wuhan, China), the results are not fully representative. The predictive potency of the nomogram needs to be verified in other medical facilities outside of Wuhan. However, despite these limitations, we have successfully identified in-hospital mortality risk factors of COVID-19 and have constructed predictive nomograms to estimate the in-hospital mortality risk of COVID-19 based on routine clinical indicators.