Accurate assessment of COVID-19 severity is essential for predicting patient outcomes, as it allows clinicians to identify individuals at high risk of mortality early in the disease course. This prognostic information directly guides triage decisions, ensuring that limited hospital resources are allocated to those who need them most.

This retrospective study included 160 hospitalized COVID‑19 patients aged over 18 years who were admitted in 2020 to Iranshahr University of Medical Sciences hospitals. Overall in-hospital fatality was 14.4%, and all deaths occurred in the critical group. Among demographic factors, older age and lower educational level (P <0.01) were significantly associated with increased disease severity, whereas sex and occupational status demonstrated non significant trends toward greater disease severity. Among behavioral factors smoking showed positive but statistically non significant (p=0.067) associations with COVID 19 severity. Lower fruit consumption was significantly associated with more severe outcomes (P = 0.021). Among comorbidities, cancer, hypertension and cardiovascular disease were strongly associated with severe and critical illness (P < 0.05) and including these factors in ANN increased the accuracy of diagnostic critical cases.

The ANN analysis revealed that the national COVID-19 severity protocol possesses exceptional predictive power for determining fatality (AUC = 0.99, CI95%: 0.98–1.00). These findings highlight the critical role of such protocols in optimizing patient triage and hospitalization priorities, and suggest that these clinical measures should be integrated with a comprehensive understanding of the social and behavioral factors driving mortality.