Artificial Intelligence Diagnostic Decision Support to Reduce Antimicrobial Prescriptions in Young Children With Colds

Description

Participants and Setting This will be a 12-month, within-subject design, single center study of 300 children 6 to 24 months of age presenting to their primary care providers at Children’s Community Pediatrics offices with upper respiratory symptoms. Exclusion criteria are children with tympanostomy tubes or purulent otorrhea, absence of upper respiratory symptoms, or who are currently taking antimicrobials.

Design and Outcomes

Using a double blind, within-subject design, each child’s ear will be assessed by the AI app and a standard clinical exam. The primary outcome measure will be antimicrobial prescription rates derived from 150 paired images that each have an AI app and clinician diagnosis. We will secondarily assess acute otitis media (AOM) diagnosis rates. If the AI app diagnoses AOM it will always prescribe an antimicrobial. Secondary outcomes are described below:

* Implementation challenges reflected by the proportion of uninterpretable images (cerumen, uncooperative patient, poor technique).
* Symptom resolution and side effects of antimicrobial use for 10 days after enrollment, regardless of whether children are diagnosed with AOM or prescribed antimicrobials. We will monitor symptoms of AOM daily using a validated symptom scale entered every evening by parents in electronic diaries which we have used in many other studies. If any participant increases their symptom score by >20% at any time, we will contact them and offer a visit. Rates of protocol-defined diarrhea and diaper rash which are the most common side effects of antimicrobial use in this age group will be assessed.
* AOM recurrences by reviewing the medical record for 3 months following enrollment.

Sample Size Calculation Using paired observations (AI app vs clinician antimicrobial prescription recommendations), an estimated 300 children 6 to 24 months of age presenting to primary care practices with upper respiratory symptoms will need to be enrolled to derive 150 paired interpretable images to detect a 10% difference in AOM diagnosis and subsequent antimicrobial prescription rates between the AI app and clinicians, assuming a power of 80% and two-sided p-value 20% at any time, they will be contacted and offered a visit. Rates of protocol-defined diarrhea and diaper rash which are the most common side effects of antimicrobial use in this age group will also be assessed. Finally, AOM recurrences will be monitored by reviewing the electronic medical record for 3 months following enrollment.

Duration:

1. The app will only be used once, before the clinician exam.
2. The ear symptom score will be done at the enrollment visit and then for the next 10-11 days in electronic diaries.
3. Text message reminders to complete the diaries will go for 10-11 days after the enrollment visit.
4. Medical record review will proceed for 90 days after enrollment.