Cleveland Clinic researchers have used outpatient EHR data to build AI algorithms that accurately predict worsening asthma before it manifests in symptoms.
The tools could be used by primary care providers to head off adverse respiratory events and suboptimal healthcare utilization, according to a study published this month in Chest.
Joe Zein, MD, PhD, and colleagues trained and tested three machine learning models on outpatient data from more than 60,300 asthma patients followed at the Cleveland Clinic from 2010 to 2018.
They found three illustrative outcomes—treatment with oral corticosteroids, ER visits and hospitalization—were all best predicted by a machine learning algorithm called Light Gradient Boosting Machine (LightGBM), which specifically fits AI to predictive analytics.
“To our knowledge, we are the first to use outpatient data to build predictive models that enable primary care providers to identify high-risk asthma patients,” the authors write in their discussion. “Machine learning algorithms can be incorporated in EHRs to build predictive models using real-world data that account for local population characteristics.”
Zein and co-authors note that, in asthma, these capabilities could, for example, help identify patients who need to be treated at specialized asthma centers rather than given intensified therapeutics locally.
Theoretically, Zein et al. add, asthma-specific AI might also help warn patients of impending turns for the worse so they can take preventative measures.
The authors cite the widespread use of electronic health records, not least among patients, as instrumental in setting the table for these burgeoning advances in asthma care.