Development and validation of a nomogram for predicting cough variant asthma diagnosis.

Abstract

Background: Cough variant asthma (CVA) is a specific type of asthma characterized by chronic cough as the sole or predominant symptom. Accurate diagnosis is crucial for effective treatment, yet bronchial provocation test is not always feasible in clinical settings. To identify independent predictors of CVA diagnosis, we developed a nomogram for predicting CVA. Univariate and multivariate logistic regression analyses were employed to construct the model, and the accuracy and consistency of the prediction model were subsequently validated.

Methods: We conducted a retrospective review of clinical data from 241 outpatients with chronic cough (≥ 8 weeks) who underwent bronchial provocation test at our hospital between January 2018 and December 2021. Patients were categorized into CVA group and Non-CVA group based on diagnostic criteria. Univariate analysis (chi-square and t-tests) was performed, followed by multivariate logistic regression to identify independent predictors. A nomogram was constructed using these predictors and validated using Bootstrap resampling (B = 200) to calculate the C-index. Additionally, receiver operating characteristic (ROC) curve analysis and decision curve analysis (DCA) were employed to assess the model's accuracy.

Results: Of the 241 outpatients, 156 (64.7%) were diagnosed with CVA. Multivariate analysis identified several independent predictors of CVA, including cough triggered by cold air (OR = 12.493, P = 0.019), exposure to pungent odors (OR = 3.969, P = 0.002), cough phasing (OR = 4.515, P < 0.001), history of allergic rhinitis (OR = 3.231, P = 0.018), and the percentage of the predicted value of maximum mid-expiratory flow (MMEF%pred) (OR = 0.981, P = 0.039) were independent predictors of CVA. The nomogram demonstrated good discrimination (AUC = 0.829) and calibration, with a sensitivity of 75.3% and specificity of 77.6% at the optimal cutoff. The C-index was 0.920, indicating excellent model performance.

Conclusions: We successfully developed and validated a user-friendly nomogram that accurately predicted CVA diagnosis based on clinical characteristics and pulmonary function test. This nomogram model could assist clinicians in diagnosing CVA, especially in patients without bronchial provocation test or with contraindications to bronchial provocation test.