Development of a Clinical Prediction Model for Recurrent Anaphylaxis

Background/Objectives: Preventing recurrent anaphylaxis is crucial for patient safety. This study aimed to identify predictive factors and develop a prediction model to estimate recurrence risk, thereby enhancing targeted preventive strategies. Methods: This prognostic prediction study used a retrospective observational cohort design, analyzing medical records from an anaphylaxis registry at Naresuan University Hospital, Phitsanulok, Thailand, between March 2011 and February 2021. We developed a prediction model using multivariable Cox proportional hazards regression analysis. Statistically significant and clinically relevant predictors were weighted into a risk score derived from hazard ratio regression coefficients. Model performance was evaluated using the area under the receiver operating characteristic curve (AuROC), calibration metrics, and decision curve analysis. Results: Over the 10-year period, 381 patients experienced 439 anaphylaxis episodes, including 58 recurrences (13.2%). The final model comprised six predictors: history of food, insect, and drug allergies; asthma; chest discomfort; and severe anaphylaxis. Corresponding risk scores were 4, 5, 5.5, 1, 2.5, and 1.5 points, respectively. Total scores ranged from 0 to 19.5 and were categorized into low (<3.0), moderate (3.0–9.0), and high (>9.0) risk groups. The high-risk group had a likelihood ratio positive (LHR+) of 4.65. The model demonstrated acceptable discrimination (AuROC 0.773 (95% CI: 0.714–0.832)) and good calibration. Bootstrap validation showed consistent performance (AuROC 0.773 (95% CI: 0.714–0.831)). Decision curve analysis indicated clinical utility across relevant threshold probabilities. Conclusions: This prediction model provides a simple, clinically applicable tool for estimating the risk of recurrent anaphylaxis and may support improved prevention and management strategies.