Establishment of nomogram prediction model for patients with chronic obstructive pulmonary disease based on Lasso feature screening.

Background and objective: To establish a nomogram prediction model for patients with chronic obstructive pulmonary disease (COPD) based on Lasso feature screening using acoustic features and general clinical data, as well as a risk warning model for patients with acute exacerbation of COPD (AECOPD), and to investigate the performance and value of these two models. Methods: A total of 240 male COPD patients, including 41 patients with acute exacerbation, and 82 healthy control male volunteers were enrolled as subjects from October 2022 to January 2024. Acoustic features and general clinical data were collected. Lasso regression was used to screen variables related to COPD and AECOPD diagnosis, and nomogram models were separately established and verified by the receiver operating characteristic (ROC) curve, calibration curve, and decision curve. Results: Variables related to COPD diagnosis screened by Lasso regression included age, smoking history, a_Jitter, e_MFCC1, e_F2 frequency, i_H1-A3, i_F1 amplitude, o_F1 amplitude, and u_MFCC4， and the variables related to AECOPD included expectoration, mMRC grade, i_Jitter, i_F2 frequency, i_Alpha Ratio, and u_H1-H2. The ROC Curve showed that the Area Under the Curve (AUC) of the COPD nomogram model was 0.95, and the AUC of the AECOPD risk warning model was 0.83. The calibration curve indicated that nomogram models showed reasonable consistency, and the Mean Absolute Error (MAE) values were 0.026 and 0.028, respectively. The decision curve indicated that nomogram models showed good benefit, and the benefit thresholds were nearly full threshold, and 0.11-81 and 0.88-0.99, respectively. Conclusion: The nomogram models for COPD prediction and risk warning of AECOPD can be used as a clinical auxiliary diagnostic and early screening method, providing new insights into the intelligent auscultation of COPD.