Top-line data from the latest trial of ResApp’s smartphone-based diagnostic algorithm for respiratory diseases suggest that the machine learning tool’s conclusions generally align with clinical diagnoses when employed for younger patients.
The double-blind, prospective Breathe-Easy pediatric study was conducted among a cohort of 585 Australian children whose ages ranged from 29 days to 12 years, more than a quarter of whom were aged younger than two years. All participants presented to one of two participating hospitals with signs of respiratory disease.
ResApp’s offering uses a smartphone’s microphone to record the sound of a patient’s cough, which is then run through the algorithm to reach a diagnosis. For many respiratory conditions, the tool also takes patient-reported symptoms into consideration. Researchers compared diagnoses made by the algorithm with those from a clinical adjudication committee to determine positive and negative percent agreement for a number of respiratory conditions.
According to ResApp, positive percent agreement ranged from a high 97 percent for asthma and reactive airway disease decisions to a low 79 percent for primary upper respiratory tract disease decisions. These two conditions also comprised the highest and lowest negative percent agreement, with asthma and reactive airway disease reaching 91 percent and primary upper respiratory tract disease sitting at 80 percent.
All other positive and negative percent agreement data were as follows: 83 percent positive and 82 percent negative for lower respiratory tract disease, 88 percent positive and 86 percent negative for croup (for which the algorithm did not consider patient-reported symptoms), 87 percent positive and 85 percent negative for pneumonia, and 84 percent positive and 81 percent negative for bronchiolitis.
“Even with all the information available to us in a modern, well-equipped hospital we still face difficulties in diagnosing childhood respiratory diseases,” Dr Paul Porter, the study’s principal investigator at Joondalup Health Campus and Princess Margaret Hospital, said in a statement. “These excellent prospective results, obtained without the need to examine the patient or perform investigations, will give clinicians significantly more confidence in their diagnosis of respiratory disease, especially in telehealth and emergency department settings.”
In a statement, ResApp CEO and Managing Director Tony Keating said that the company will be relying on these data as the foundation for its upcoming CE Mark and Therapeutic Goods Administration regulatory submissions.
“The outstanding results delivered by the Breathe Easy prospective study are consistent with our expectations, and for diseases such as pneumonia and asthma have even exceeded our expectations,” Keating said. “This formal confirmation of ResApp’s algorithms’ robust performance will allow us to provide key diagnostic products in an array of clinical settings, especially in telehealth — one of the fastest growing segments in healthcare — where up to 50 percent of consultations are respiratory related.”
Last year, ResApp faced a major setback after the data that came back from its landmark clinical trial, conducted at major institutions like Massachusetts General Hospital, Cleveland Clinic, and Texas Children’s Hospital, was proclaimed useless due to “procedural anomalies.” Keating told MobiHealthNews that the company has since learned a number of key lessons following the botched trial, and now expects to have the results of this larger, US-based clinical trial within the month.