Background: This study aimed to establish a Severity Scale for influenza and other acute respiratory infections (ARI), requiring hospitalization, for surveillance and research purposes (the SevScale). Such a scale could aid the interpretation of data gathered from disparate settings. This could facilitate pooled analyses linking viral genetic sequencing data to clinical severity, bringing insights to inform influenza surveillance and the vaccine strain selection process. Methods: We used a subset of data from the Global Influenza Hospital Surveillance Network database, including data from different geographical areas and income levels. To quantify the underlying concept of severity, an item response model was developed using sixteen indicators of severity related to the hospital stay. Each patient in the dataset was assigned a Severity Score and a Severity Category (low, medium, or high severity). Finally, we compared the model scores across different subgroups. Results: Data from 9 countries were included, covering between 4 and 11 seasons from 2012 to 2022, with a total of 96,190 ARI hospitalizations. Not for all severity indicators data was available for all included seasons. Subgroups with a high percentage of patients in the High Severity Category included influenza A(H1N1)pdm09, age ≥50, lower-middle income countries, and admission since the start of the COVID-19 pandemic. Conclusions: The initial model successfully highlighted severity disparities across patient subgroups. Repeating this exercise with new, more complete data would allow recalibration and validation of the current model. The SevScale proved to be a promising method to define severity for influenza vaccine strain selection, surveillance and research.