Machine Learning Tool Identifies Asthma, COPD Better Than a Physician, Study Finds


A diagnostic algorithm was evaluated for its ability to gauge the presence of asthma or chronic obstructive pulmonary disease (COPD), in a new analysis.

An asthma/chronic obstructive pulmonary disease (COPD) differentiation classification tool was able to identify patients 35 years and older with asthma or COPD more accurately than primary care physicians (PCPs) and pulmonologists, a new study has found.

“The primary objective of this study was met; the average diagnostic accuracy of the AC/DC tool for these diagnoses was superior to that of PCPs and noninferior and superior to that of pulmonologists,” wrote the researchers of this study. “Furthermore, the diagnostic accuracy of pulmonologists was superior to that of PCPs, as might be expected by virtue of their medical specialization.”

The results of this multinational observational study were published in Journal of Allergy and Clinical Immunology.

Asthma and COPD are chronic respiratory diseases with overlapping symptoms and diagnostic criteria, which can pose a challenge for PCPs and pulmonologists to differentiate and diagnose. Using the Asthma/COPD Differentiation Classification (AC/DC) tool, the researchers of this study aimed to evaluate the accuracy of this machine learning tool in diagnosing patients with asthma, COPD, and asthma-COPD overlap (ACO) compared with PCPs and pulmonologists.

The tool initially included clinical characteristics of more than 400,000 patients 35 years or older, with a diagnosis of asthma, COPD, or ACO by a specialist (pulmonologists/allergists) between 2010 and 2017.

From the data, 12 variables were identified as the most impactful:

  • Forced expiratory volume per 1 second (FEV1)
  • FEV1/forced vital capacity (FEV1/FVC)
  • Smoking pack-year
  • Age at onset of respiratory disease
  • Body mass index
  • Dyspnea
  • Wheeze
  • Cough
  • Diagnosis of allergic rhinitis
  • Current smoker
  • Never smoked
  • Diagnosis of chronic rhinitis

In total, 119 patients were evaluated and screened by an expert panel of 3 pulmonologists and 4 PCPs from 5 countries. Of the total, patients had asthma (n = 53), COPD (n = 43), ACO (n = 7), or other (n = 13). Cases were evaluated by 180 PCPs and 180 pulmonologists from 9 countries and by the AC/DC tool to compare accuracy for diagnosing asthma, COPD, or ACO.

The average diagnostic accuracy of the AC/DC tool was superior to that of PCPs (mean difference, 24%; 95% CI, 17%-29%; P < .0001) and was noninferior (mean difference, 12%; 95% CI, 6%-17%; P < .0001) or superior (P = .0006) to that of pulmonologists. Additionally, average diagnostic accuracies were 73% (AC/DC tool), 50% (PCPs), and 61% (pulmonologists) vs the expert panel's diagnosis, respectively.

The researchers acknowledge some limitations to this study, including that it only studied patients 35 years and older; the AC/DC tool cannot be used among younger patients. Furthermore, the AC/DC tool was not intended to be used to diagnose patients on their own. Rather, it should be used as an aid in addition to spirometry when diagnosing asthma, COPD, or ACO in patients.

“Subject to these validation and safety considerations, the tool has the potential to support a range of clinicians, including nurse practitioners, PCPs, pulmonologists, and respiratory experts functioning across health care facilities," wrote the researchers, "such as mini-clinics, outpatient or satellite care centers, and large hospitals, in distinguishing between asthma and COPD in patients 35 years and older, in whom other causes of respiratory symptoms have been excluded.”


Kocks JWH, Cao H, Holzhauer B, et al. Diagnostic performance of a machine learning algorithm (asthma/chronic obstructive pulmonary disease [COPD] differentiation classification) tool v primary care physicians and pulmonologists in asthma, COPD, and asthma/COPD overlap. J Allergy Clin Immunol Pract. Published online January 28, 2023. doi:10.1016/j.jaip.2023.01.017

Related Videos
Rachel Dalthorp, MD
dr jennifer vidrine
dr brian koffman
Melissa Jones, MD on Artificial Intelligence and Sleep Studies
dr surya bhatt
Amy Shapiro, MD, medical director, Indiana Hemophilia and Thrombosis Center
dr thomas hall
dr surya bhatt
RAvin Ratan, MD, MEd, MD Anderson
Related Content
CH LogoCenter for Biosimilars Logo