Prediction models for pulmonary function during acute exacerbation of chronic obstructive pulmonary disease

Objective: The pulmonary function test is an effort-dependent test; however, during acute exacerbation of chronic obstructive pulmonary disease (AECOPD), patients are unable to effectively cooperate due to poor health. The present study aimed to establish prediction models that only require demographic and inflammatory parameters to predict pulmonary function indexes: forced expiratory volume in one second (FEV1) and forced vital capacity (FVC). Approach: The goal was to establish prediction models based on multi-output support vector regression. A total of 143 subjects received a peripheral blood examination and pulmonary function test. The demographic and inflammatory parameters were used as input features, and FEV1 and FVC were used as the target features in prediction models. Three models (mixed model, severe model and nonsevere model) were established with FEV1 < 1 l as the threshold of severe episodes of AECOPD. The values of FEV1 and FVC from the pulmonary function tests were compared with the prediction models to validate the performances of the developed prediction models. Main results: The severe and nonsevere models' prediction performances were better than that of the mixed model. The mean squared errors were lower than 0.05 l(2), and the decision coefficients (R-2) were higher than 0.40. The two-tailed t-test results showed that for both severe and nonsevere models, the absolute percentage errors of FEV1 and FVC were within 10%. Significance: Our study shows the feasibility of predicting the pulmonary function indexes FEV1 and FVC with demographic and inflammatory parameters when the pulmonary function test fails to be implemented, which is beneficial for the treatment of AECOPD.