<![CDATA[Background: COVID-19 remains a global public health concern. Differences in disease severity and hospitalization have been observed between vaccinated and unvaccinated individuals. This study aimed to evaluate COVID-19 vaccination status and clinician-assessed clinical severity, as well as artificial intelligence (AI)-classified lung parenchymal involvement. Methods: This retrospective cohort study included COVID-19 patients from January to December 2022. Clinical severity at admission was determined by World Health Organization criteria (mild, moderate, severe, critical). Lung parenchymal involvement was classified using a previously validated VGG16-based deep learning model applied to admission chest X-ray images. Vaccination status, verified through the national Satu Sehat registry, was categorized as unvaccinated, single-dose, or two-dose, without regard to time since last vaccination and vaccine type cause not consistently available. Results: A total of 153 patients were included, 31 patients (20%) unvaccinated, 22 patients (15%) single-dose, and 100 patients (65%) two-dose recipients. Critical disease occurred in 12.9% of unvaccinated patients based on clinician assessment. No patients were classified as having severe lung parenchymal involvement by the AI model. Moderate lung involvement was more frequent in unvaccinated patients (29.0%) compared with single-dose (27.3%) and two-dose groups (7.0%). In multivariable ordinal logistic regression, two-dose vaccination was independently associated with lower odds of higher based on WHO clinical severity (adjusted odds ratio [aOR]=0.41; 95% confidence interval [CI]=0.19–0.86; P=0.018) and lower odds of more severe AI-classified lung involvement (aOR=0.32; 95% CI=0.14–0.71; P=0.005). Conclusion: Two-dose vaccination was associated with lower clinician-assessed severity and reduced AI-classified lung parenchymal involvement at admission in this retrospective hospital-based cohort.]]>
