A Clinical Risk Score Based on Albumin and Electrolyte Levels for Predicting Death Risk in Hospitalized Elderly COVID-19 Patients

Abstract

Background: The Omicron subvariants of SARS-CoV-2 spread rapidly since 2021. Following China's relaxation of containment measures in December 2022, a surge in COVID-19 cases poses a public health threat. Early identification of elderly COVID-19 patients at death risk is crucial for optimizing treatment and resource use.

Objective: To develop a clinical score for predicting death risk in elderly COVID-19 patients at hospital admission, based on a cohort from the Second Hospital of Shandong University.

Methods: We established a retrospective cohort of hospitalized COVID-19 patients from November 1, 2022, to March 31, 2023. Cox regression identified prognostic factors, leading to the development of a nomogram-based prediction model and a clinical risk score. Patients were classified into low- and high-risk groups using optimal segmentation thresholds, with survival curves generated by the Kaplan-Meier method. An online risk calculator was developed to facilitate real-time risk assessment in clinical settings.

Results: The cohort included 1413 hospitalized COVID-19 patients. Elderly patients (≥60 years, N = 971) had a high mortality rate of 18.13%. Four independent predictors of mortality were identified: age (HR = 1.07), serum albumin (HR = 0.88), serum potassium (HR = 0.35), and serum sodium (HR = 0.91). The developed risk score demonstrated strong predictive performance and effectively stratified patients into risk categories.

Conclusion: We developed a validated clinical risk score integrating age, serum albumin, potassium, and sodium levels to predict mortality in hospitalized elderly COVID-19 patients. This scoring system enables early risk stratification, assisting clinicians in decision-making and optimizing patient management.

Keywords: albumin; clinical risk score; death risk; electrolyte levels; hospitalized elder patients with COVID-19.

Figure 1

Nomogram for COVID-19 prediction model. The four variable points are allocated for age, ALB, K and Na axis, and the top line points are the value for each variables. The sum of the corresponding points for the variables is the total points. The risk scores calculated by the nomogram correspond to the risk stratification, and 7-, 14-, 28-day survivals are visualized.

Figure 2

Model performance of nomogram. (A) The receiver operating characteristic (ROC) curves of the nomogram at 7 days, 14 days and 28 days. (B) Harrell’s concordance index (C-index) of the nomogram. (C) Calibration curves for predicting patient survival at 7 days, 14 days and 28 days.

Figure 3

Determine the optimal segmentation threshold of risk score and risk stratification from COVID-19 prediction model. (A) Determine the optimal segmentation threshold of risk score by the corresponding log-rank P-values of COVID-19 patients. (B) Risk factor association chart. The graph above shows the predicted risk scores for each patient ranked in descending order. Two groups are differentiated by the optimal segmentation threshold cutoff value: low-risk (green) and high-risk (red). The graph below presents the relationship between the predicted value-at-risk ranked patients and the survival time. The green dots represent the living patients, and the red dots represent the dead patients. (C) Kaplan–Meier curves for overall survival of high-risk and low-risk patients based on the optimal segmentation threshold.

Figure 4

The online calculator for predicting the outcome of elder patients with COVID-19.