Journal of Medical Internet Research - International Scientific Journal for Medical Research, Information and Communication on the Internet

Journal of Medical Internet Research - International Scientific Journal for Medical Research, Information and Communication on the Internet

Medication errors affect approximately 56 000 children treated by emergency medical services (EMS) each year in the US, with many drugs administered outside the proper dose range.¹ In addition, many errors likely go underreported because of failure to recognize them or reluctance to report them.² In 2017, the World Health Organization called for a reduction by 50% of serious and avoidable medication-associated harm in all countries during the ensuing 5 years.³ Emergency care is an environment with a high risk for medication errors, particularly in critical pediatric situations, such as out-of-hospital cardiopulmonary resuscitation.⁴ In this setting, the combination of limited safeguards and resources⁵ places children at higher risk than adults for life-threatening prehospital medication errors.^5,6 Factors associated with increased risk for children include little exposure of paramedics to critically ill children, an increased cognitive load owing to emotional stress and time pressure, and pediatric-specific, age-related variations in pharmacokinetics, with the need for an individual, weight-based dose calculation and drug preparation for each child.^1,7,8 Among other drugs, epinephrine has the highest rate of incorrect dose administration, with up to 68% of preparations containing an error and a mean error overdose of 808%.^9,10 Similarly, a study¹¹ indicated a frequency of medication errors by paramedics of 49% to 63%, with miscalculation as a primary cause.

The interpretation of lung auscultation is highly subjective and relies on non-specific nomenclature. Computer-aided analysis has the potential to better standardize and automate evaluation. We used 35.9 hours of auscultation audio from 572 pediatric outpatients to developDeepBreath: a deep learning model identifying the audible signatures of acute respiratory illness in children. It comprises a convolutional neural network followed by a logistic regression classifier, aggregating estimates on recordings from eight thoracic sites into a single prediction at the patient-level. Patients were either healthy controls (29%) or had one of three acute respiratory illnesses (71%) including pneumonia, wheezing disorders (bronchitis/asthma), and bronchiolitis). To ensure objective estimates on model generalisability,DeepBreathis trained on patients from two countries (Switzerland, Brazil), and results are reported on an internal 5-fold cross-validation as well as externally validated (extval) on three other countries (Senegal, Cameroon, Morocco).DeepBreathdifferentiated healthy and pathological breathing with an Area Under the Receiver-Operator Characteristic (AUROC) of 0.93 (standard deviation [SD] ± 0.01 on internal validation). Similarly promising results were obtained for pneumonia (AUROC 0.75 ± 0.10), wheezing disorders (AUROC 0.91 ± 0.03), and bronchiolitis (AUROC 0.94 ± 0.02). Extval AUROCs were 0.89, 0.74, 0.74 and 0.87 respectively. All either matched or were significant improvements on a clinical baseline model using age and respiratory rate. Temporal attention showed clear alignment between model prediction and independently annotated respiratory cycles, providing evidence thatDeepBreathextracts physiologically meaningful representations.DeepBreathprovides a framework for interpretable deep learning to identify the objective audio signatures of respiratory pathology.