Clinical prediction model for mortality of adult diabetes inpatients with COVID-19 in Wuhan, China: A retrospective pilot study

Graphical abstract

The novel coronavirus disease 2019 (COVID-19) originated in Wuhan, China, and has since spread expeditiously across the global [1,2]. Recent reports indicated that 19% of adult inpatients with COVID-19 had a comorbidity with diabetes [3]. Additionally, it was found that diabetes is a compounding factor that exacerbates the fatality rate of this patient subgroup. Therefore, it is important to recognize the risk factors between diabetes and fatality in patients with COVID-19.

This was a retrospective case-control study with a confirmed novel coronavirus infected pneumonia (NCIP) diagnosis between January 10 and February 20, 2020, which was approved by the institutional ethics board of The Fifth Hospital of Wuhan, Wuhan, China. For each non-survivor with pre-existing diabetes (n = 26), two subjects (n = 52) were matched based on age, sex, and date of admission among discharged adult inpatients infection with diabetes. Using the 78 patient medical records sampled, two independent operators (J. Guo and X.L. Yang) conducted the study and analyzed the medical data of patients. Among all patients, the mean age was 51.65 ± 13.28 years and 63.6% were male.

1. Risk factors for mortality of NICP inpatients

In univariable analysis, in-hospital death was correlated with serum level of alanine aminotransferase (ALT), activated partial thromboplastin time (APTT), urea nitrogen, creatinine kinase-myocardial band (CK-MB), creatine kinase (CK), lactate dehydrogenase (LDH), creatinine, direct bilirubin, D-dimer, myoglobin, total bile acid, total bilirubin, cardiac troponin I, white blood cell (WBC) count, lymphocyte percentage, neutrophil percentage, eosinophils, hemoglobin, and platelets at hospitalization admission. The multivariable analysis showed that increased APTT (OR = 1.494, 95%CI 1.062–2.101), urea nitrogen (OR = 1.200, 95%CI 1.056–1.364), WBC count (OR = 2.233, 95%CI 1.166–4.280), and LDH (OR = 1.014, 95%CI 1.001–1.028) were significantly associated with the mortality of NICP patients.

2. Prediction model

Based on the results above, our proposed prediction model incorporated only with APTT, urea nitrogen, LDH, and WBC count. This model predicted 83.6% of all NCIP inpatients combined total death (Fig. 1 ). The p-value for the Hosmer and Lemeshow test was 0.45, confirming the null hypothesis of a good fit for the final model. At the optimal cut-off of the final prediction model, sensitivity and specificity were 84.7% and 74.8%, respectively.

Fig. 1.

Receiver operating characteristic curve for the final prediction model, based on APTT (activated partial thromboplastin time), urea nitrogen, LDH (lactate dehydrogenase), and white blood cell count. The area under the curve is 0.836, meaning that 83.6% of the in-hospital death of patients with COVID-19 can be explained by the model.

This study indicates that the variables APTT, WBC count, LDH, and blood urea nitrogen can be assessed during hospitalization for good prediction of mortality for adult diabetes patients with COVID-19. Risk prediction modeling could be a useful tool for prioritizing monitoring of diabetes patients with COVID-19. The finding suggests that the risk of fatality increased with elevated levels of APTT, WBC count, LDH, and blood urea nitrogen. These results were supported by recent studies of patients with COVID-19 [4]. In the report by Zhou et al. [5], high Sequential Organ Failure Assessment (SOFA) score and d-dimer greater than 1 μg/mL at an early stage were the potential risk factors for mortality with COVID-19. In the present study, we found elevated levels of APTT, platelet, and D-dimer were also related to mortality by univariable analysis. Prolonged APTT was caused by COVID-19 infection, and it might be more valuable to reflect coagulopathy status for patients with diabetes. Although SOFA may be a reasonable model, it may have excluded some important data because the collection of information was limited at the early stage of COVID-19.

This is the first report to create a prediction model based on the variables APTT, WBC count, LDH, and blood urea nitrogen for adult diabetes patients with COVID-19. This simple model crucially highlights parameters that can be analyzed to predict risk of fatality for this subgroup of patients.

Ethics approval and consent to participate consent for publication

This study was only a retrospective report, so there were no patients or others.

Consent for publication

Not applicable.

Availability of data and material

All data generated or analyzed during this study are included in this published article.

Funding

Science and Technology Department of Sichuan Province (2018SZ0224).

Authors' contributions

The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted.

Minghua Su, Jing Yuan, Jieru Peng wrote the paper and conducted this study. Mengjun Wu, Yousheng Yang and Yong G, Peng designed the study, reviewed, and edited the manuscript.

Declaration of competing interest

The authors declare that they have no competing interests.