Development and Validation of the Risk of Exacerbation in Severe Asthma (RESA) Model

Abstract

Background: Severe asthma (SA) is associated with frequent exacerbations and high treatment costs.

Objectives: To develop and validate an individualized risk calculator for severe exacerbations in SA, and evaluate its clinical utility for guiding personalized clinical decisions.

Methods: Patients with SA were identified from combined data from the International Severe Asthma Registry (2015-2022) and NOVEL observational longiTudinal studY (2016-2023) across 30 countries and regions. The prediction end point was the 12-month risk of 1 or more or 2 or more severe exacerbations. Using expert input and Bayesian network analysis, 11 routinely measured predictors were identified, measured within the past 12 months. A mixed-effects, zero-inflated negative binomial model was developed, adjusting for between-country variability and biologic drop-in effects. Internal-external cross-validation was performed using the natural clustering by country settings.

Results: Data from 9911 patients with SA were used. Essential predictors included age, sex, past 12-month severe exacerbations, asthma control, chronic rhinosinusitis, FEV₁ to forced vital capacity ratio, percent predicted FEV₁, blood eosinophils, fractional exhaled nitric oxide, and long-term oral corticosteroid and macrolide use. The model also adapted setting-specific baseline risks. In the internal-external cross-validation, across broad geographical and health care variability, the model showed excellent calibration and informative, generalizable discrimination (pooled area under the time-dependent receiver-operating characteristics curve of 0.63 [95% CI, 0.60-0.66] for ≥1 and 0.68 [95% CI, 0.64-0.72] for ≥2 exacerbations). Decision curve analysis showed clear net benefit across risk thresholds.

Conclusions: The Risk of Exacerbation in Severe Asthma model quantifies SA exacerbation risk using routinely available predictors and demonstrates potential clinical utility.

Keywords: Clinical decision support; Clinical utility; Exacerbation risk; Individualized risk; Internal-external cross-validation; Precision medicine; Prediction model; Real-world data; Severe asthma.