Developing a Passive Digital Marker for the Prediction of Childhood Asthma Treatment Response

Undertreatment of childhood asthma is prevalent, and often the right treatment for an initial diagnosis—the “incident case”—is unknown, hence the widespread use of therapeutic trials to determine an evidence-based treatment strategy. Two-thirds of incident childhood asthma cases continue to have persistent symptoms even after treatment initiation. Missed opportunities for early and efficacious treatment contribute to increased risk of childhood asthma-associated morbidity, such as uncontrolled asthma, which exerts a substantial burden on patients, families, and the healthcare system. Actively reviewing electronic health records (EHRs) to identify factors that could inform treatment decisions can be costly, time consuming, error-prone, and infeasible in busy clinics. 

This research is predicated on the notion that applying novel machine learning (ML) methodologies to increasingly available EHR risk and prognostic data can generate predictive analytics and insights regarding childhood asthma treatment response. Clinicians can then use such insights toward effective treatment decision making at the point of care, including more proactive and personalized treatment, for improved patient-centered outcomes.

Researchers will build on previous work and develop a Passive Digital Marker for Treatment Response (PDM-TR), a ML algorithm that can retrieve and synthesize existing mother-child dyad data from an EHR to provide an objective and quantifiable marker of a child’s future response to first-line asthma treatment. 

The specific aims of the research are as follows: 

• Develop and evaluate the predictive performance of a PDM-TR among children with an incident asthma diagnosis.
• Examine whether the incidence of asthma control is different among children with distinct asthma phenotypes (i.e., allergic versus nonallergic) when exposed to the same first-line treatment.

This retrospective cohort study will use data from the Indiana Network for Patient Care (INPC) EHR databases with a focus on pediatric and primary care practices. Eligible cases will include children up to age 11 who were born between 2005-2019 and have had incident asthma diagnosis identified, with followup starting on the date of the diagnosis up until December 31st, 2019. The study team will develop, validate, and evaluate the ability of the PDM-TR to provide an accurate prediction of treatment response. The algorithm will learn from existing EHR data to predict whether or not a specific treatment may be successful for an individual with a specific set of attributes, such as demographics, history of allergy sensitization, eczema, lung function, and body mass index. 

Researchers hypothesize that, when applied to risk and prognostic EHR data derived from incident asthma cases exposed to first-line treatments, the PDM-TR will predict asthma control at 2 to 3 months with high accuracy: greater than or equal to 80 percent sensitivity and greater than or equal to 80 percent specificity. Upon successful development of the PDM-TR, future studies will evaluate the feasibility of implementing the tool in clinical practice for treatment response prediction and its efficacy in improving asthma control compared to current practice.