# Multiscale Modeling of Hospital Length of Stay for Successive SARS-CoV-2 Variants: A Multi-State Forecasting Framework

## TL;DR

This study models hospital stays for different SARS-CoV-2 variants and age groups to improve pandemic resource planning.

## Contribution

A multi-state forecasting framework that integrates variant-specific and age-based hospital length of stay dynamics.

## Key findings

- Omicron-phase hospital stays were 5–8 days, shorter than the 10–14 days in earlier phases.
- Older patients (65+) had longer stays (8–12 days) and more prolonged admissions (over 30 days).
- The model can support real-time bed-availability systems and variant-specific resource planning.

## Abstract

Understanding how hospital length of stay (LoS) evolves with successive SARS-CoV-2 variants is central to the multiscale modeling and forecasting of COVID-19 and other respiratory virus dynamics. Using records from 1249 COVID-19 patients admitted to Chungbuk National University Hospital (2021–2023), we quantified LoS across three distinct variant phases (Pre-Delta, Delta, and Omicron) and three age groups (0–39, 40–64, and 65+ years). A gamma-distributed multi-state model—capturing transitions between semi-critical and critical wards—incorporated variant phase and age as log-linear covariates. Parameters were estimated via maximum likelihood with 95% confidence intervals derived from bootstrap resampling, and Monte Carlo iterations yielded detailed LoS distributions. Omicron-phase stays were 5–8 days, shorter than the 10–14 days observed in earlier phases, reflecting improved treatment protocols and reduced virulence. Younger adults typically stayed 3–5 days, whereas older cohorts required 8–12 days, with prolonged admissions (over 30 days) clustering in the oldest group. These time-dependent transition probabilities can be integrated with real-time bed-availability alert systems, highlighting the need for variant-specific ward/ICU resource planning and underscoring the importance of targeted management for elderly patients during current and future pandemics.