Abstract:  Objective To establish risk scoring models for predicting emergency allergic conjunctivitis surges in urban Beijing each spring and to provide a quantitative tool for estimating medical risk. Design Retrospective study. Participants A total of 13 086 patients who presented to the Ophthalmic Emergency Department of Beijing Tongren Hospital for allergic conjunctivitis in March 2021 to 2025 were enrolled. Methods Associations between the emergency allergic conjunctivitis surges and pollen concentration, air quality index, fine particulate matter, inhalable particulate matter, maximum and minimum temperatures, date, precipitation, wind grade were analyzed by univariate analysis. Variables significant in univariate analysis were selected for a binary logistic regression model and a multi-index superposition model, with an average probability model serving as the control. The efficiency of the models was evaluated by threat score, probability of detection and false alarm rate. Main Outcome Measures Threat score, the probability of detection and the false alarm rate. Results Univariate analysis showed that the emergency allergic conjunctivitis surges were significantly correlated with precipitation (χ2=6.749), maximum (t=4.103) and minimum temperatures (t=3.705), natural logarithm of pollen concentration (t=9.043), and date (t=2.287) (all P<0.05); whereas wind grade (χ2=5.074), air quality index (t=-0.364), natural logarithm of inhalable particulate matter (t=0.506)and fine particulate matter (t=-0.952) were not significant (all P>0.05). In the binary logistic regression model, there was a statistically significant association between the emergency allergic conjunctivitis surges and the natural logarithm of pollen concentration (Wald=12.397), precipitation (Wald=7.978), and maximum temperature (Wald=3.868) (all P<0.05). The most important factor in the model was pollen concentration, followed by precipitation and highest temperature. At a cut-off of 0.3, the model achieved a threat score of 55.6%, a probability of detection of 66.7%, and a false alarm rate of 23.1%. The optimized multi-index superposition model incorporating the natural logarithm of pollen concentration, precipitation, and maximum temperature, yielded a threat score of 64.7%, a probability of detection of 73.3%, and a false alarm rate of 15.4%. Both the binary logistic regression and multi-index superposition models significantly outperformed the average-probability model. Conclusion Risk scoring models for the emergency allergic conjunctivitis surges can be established based on pollen concentration, precipitation, and maximum temperature, with the optimized multi-index superposition model reaching a detection probability of 73.3%. This model may inform daily health management and medical-resource preparation.

Key words: Ophthalmic emergency, Allergic conjunctivitis, Pollen allergies, Risk assessment