Currently Viewing:
In Focus Blog

The Potential of NLP to Diagnose Childhood Asthma Earlier, Improve Future Outcomes

Allison Inserro
Using a form of artificial intelligence called natural language processing (NLP) to mine information in electronic health records (EHRs) can uncover early cases of childhood asthma more quickly, reducing the long-term morbidity of this chronic disease, improve current care, and streamline and advance new therapeutic discoveries.
 
Using a form of artificial intelligence called natural language processing (NLP) to mine information in electronic health records (EHRs) can uncover early cases of childhood asthma more quickly, reducing the long-term morbidity of this chronic disease, researchers from the Mayo Clinic proposed in a recent review. Besides creating an opportunity to identify childhood asthma earlier, they said, NLP could also streamline and advance new therapeutic discoveries.

Children and adolescents may present with a history of recurrent asthma-like symptoms, such as coughing and wheezing, which are sometimes transient. Events may be buried in EHRs, but this health data is currently too difficult and labor-intensive for providers to manually review.

NLP is a way of extracting, processing, and classifying free-text language, such as might be contained in EHR notes, and then aggregate it against a set of criteria.

In various studies, the authors have previously used 2 sets of existing criteria—the Predetermined Asthma Criteria (PAC) and the Asthma Predictive Index (API)—to create NLP algorithms to compare diagnostic performance against manual record reviews. In multiple different settings, NLP outperformed humans every time; in a real-world setting, this would allow providers to intervene earlier to prevent complications and improve overall health outcomes.

However, before NLP is used in such a way, 3 issues must be overcome, according to the review recently published in the journal Frontiers of Pediatrics.

The first issue is that clinicians must understand why early identification of asthma is important; despite the fact that in the United States over 6 million children under 17 have asthma, delayed diagnosis is common.  

The delay is not without consequence. Various studies have begun to show how early-onset asthma affects health in adulthood; improved asthma control through inhaled corticosteroids could have positive effects for those at risk from long-term lung damage. One study showed that children with wheezing episodes had lung function decline that persisted up until age 45; another demonstrated that lower lung function at age 7 was associated with chronic obstructive pulmonary disease (COPD) and asthma-COPD overlap syndrome in later years.

Children with asthma are also at increased risk for respiratory infections, such as pneumonia, pertussis, ear infections, and strep. These asthma-associated infectious and inflammatory disease comorbidities also include non-respiratory diseases, such as such as blood stream infection, appendicitis, herpes zoster, and celiac disease, the authors noted.

From a theoretical perspective, “there is a lack of consensus on asthma diagnosis,” the researchers wrote. From a social media perspective, this topic popped up on Twitter recently, as doctors talked about reactive airway disease vs asthma: why do some physicians use that term and not asthma?

 


NLP could settle such discussions.

But in order for that to happen, the authors said there is a second issue: research is needed for new informatic tools, and to refine existing ones; in addition, the implementation of NLP algorithms and other clinical informatics tools must be shared and supported widely.  

Such research could allow for the development of NLP not just to diagnose children earlier, but to stratify high-risk subsets through computational phenotyping. It could also revolutionize research capabilities by identifying subjects who meet enrollment criteria automatically, the authors noted, which could in turn lead to new treatment discoveries.

While many physicians have complaints about EHRs and while numerous flaw—and even possible harms—have been identified, the ability of NLP to pull relevant data out of an EHR can advance patient care. It’s one that is possible if partnerships develop between clinicians and researchers with informatics teams, the authors said.

Reference

Seol HY, Sohn S, Liu H, Wi C, Ryu E, Park MA, et al. Early identification of childhood asthma: the role of informatics in an era of electronic health records [published online April 2, 2019]. Front. Pediatr. doi: 10.3389/fped.2019.00113.

Related Articles

Using EHRs to Find Hidden Drug Benefits
Report: EHRs Fall Short, but So Does Physicians' Use
Study Analyzes Allergic Sensitization and Asthma from Infancy to Adulthood
Improving Outcomes for Children With Persistent Asthma
A Step in the Digital Direction: From Paper Logs to Electronic Data Capture
 
Copyright AJMC 2006-2019 Clinical Care Targeted Communications Group, LLC. All Rights Reserved.
x
Welcome the the new and improved AJMC.com, the premier managed market network. Tell us about yourself so that we can serve you better.
Sign Up