Reflecting, Not Defining: Limited Prognostic Value of Serum Creatinine upon Admission in Computed Tomography–Based Mortality Models for COVID-19

Abstract

Objective: To challenge the perceived independent prognostic value of serum creatinine upon admission in
hospitalized COVID-19 patients by evaluating its additive role in mortality prediction models that already
include direct measures of pulmonary involvement and systemic inflammation.

Methods: This retrospective cohort study included 124 adults hospitalized with polymerase chain reaction–confirmed COVID-19 across 4 centers in Türkiye between March and May 2020. All patients underwent non-contrast chest computed tomography (CT) and baseline laboratory testing within 24 hours of
admission. Computed tomography severity was quantified using a validated 6-zone scoring system. Four
logistic regression models were constructed: model A: CT severity score; model B: CT+creatinine; model C:
CT+age, diabetes, neutrophil-to-lymphocyte ratio, lactate dehydrogenase; model D: model C+creatinine.
Models were evaluated by area under the curve (AUC), Brier score, accuracy, and specificity.

Results: Creatinine levels were significantly higher in non-survivors (1.12 ± 0.41 vs. 0.83 ± 0.25 mg/dL,
P = .001). Model A yielded an AUC of 0.857; adding creatinine in model B raised the AUC modestly to
0.882. Model C, incorporating systemic and clinical variables, achieved the highest performance (AUC
0.975). Model D, which added creatinine to model C, showed no further improvement in discrimination
(AUC 0.975) or calibration (Brier score 0.032).

Conclusion: Despite its widespread clinical use and association with adverse outcomes, serum creatinine
offers limited incremental prognostic value when CT-based severity scoring and inflammatory markers
are included. These findings suggest that creatinine reflects—but does not independently define—disease
severity in COVID-19.

Cite this article as: Şahutoğlu T, Şahutoğlu E. Reflecting, not defining: limited prognostic value of serum creatinine upon admission
in computed tomography–based mortality models for COVID-19. Cerrahpaşa Med J 2025, 49, 0043, doi:10.5152/cjm.2025.25043.