Simple blood and lung tests may predict airway hyperresponsiveness before provocation testing

Researchers have developed an interpretable prediction model, together with a simplified scorecard, that estimates the likelihood of bronchial hyperresponsiveness using only routine pulmonary function measurements and blood inflammatory markers. The work, published in BMC Medical Informatics and Decision Making, was motivated by the fact that bronchial provocation testing, while valuable for assessing airway hyperresponsiveness, can be difficult to deploy in busy outpatient settings.

The study retrospectively included 3,993 outpatients with asthma-suggestive symptoms who had undergone bronchial provocation testing. All participants had a post-bronchodilator FEV1/FVC ratio in the normal range, no substantial smoking exposure, and high-resolution CT findings that were either normal or consistent with asthma. The cohort was split into training and test sets in an 8:2 ratio. Overall, 718 participants, or 18.0 percent, had bronchial hyperresponsiveness.

The team built several candidate models using demographics, blood count markers and spirometric indices: logistic regression, LASSO, random forest, XGBoost, LightGBM, a spline logistic regression, and two explainable boosting machine models. Feature selection, hyperparameter tuning and threshold selection were carried out in the training set, while discrimination, calibration and decision curve analysis were assessed in a randomly held-out internal test set.

In the test set, the two explainable boosting machine models showed discrimination similar to the gradient-boosting and random-forest models, with an area under the curve of approximately 0.81, and numerically more favorable calibration; the main-effect EBM model had an expected calibration error of 0.0233. The researchers noted that differences in calibration among the leading models were small. Higher eosinophil count and lower MEF50%, FEV1/FVC and MMEF% contributed most to the predicted risk.

From the main-effect EBM model, the team derived a simplified scorecard intended for pretest risk stratification. The scorecard retained most of the predictive performance of the full model, achieving an AUC of 0.7964 and an AUPRC of 0.4797.

The authors concluded that in this selected pretest outpatient population, routine clinical variables predicted bronchial hyperresponsiveness with good discrimination and with calibration comparable to that of the best alternative models, and that the derived scorecard may support risk stratification before bronchial provocation testing. They emphasized that external validation is required before any clinical deployment.

Subject of Research: Predicting bronchial hyperresponsiveness from routine pulmonary function and blood inflammatory markers

Article Title: Prediction of bronchial hyperresponsiveness from routine pulmonary function and inflammatory markers: an interpretable model and scorecard

Article References: Wen, X., Wu, J., Bao, W., Zheng, N., Dong, Z., Lin, Y., Zhang, M., & Yin, J. (2026). Prediction of bronchial hyperresponsiveness from routine pulmonary function and inflammatory markers: an interpretable model and scorecard. BMC Medical Informatics and Decision Making. https://doi.org/10.1186/s12911-026-03910-6

Image Credits: AI Generated

DOI: 10.1186/s12911-026-03910-6

Keywords: bronchial hyperresponsiveness, bronchial provocation testing, asthma, spirometry, eosinophils, machine learning, explainable boosting machine, scorecard, risk stratification, pulmonary function tests, model calibration, Prediction