Personalized azithromycin treatment rules for children with watery diarrhea using machine learning

Area of Science:

• Pediatric infectious diseases
• Clinical trial methodology
• Machine learning in healthcare

Background:

• Watery diarrhea is a significant cause of childhood morbidity and mortality globally.
• Azithromycin has shown efficacy in treating certain types of diarrhea, but optimal targeting is needed.
• Personalized medicine approaches can improve antibiotic effectiveness and reduce resistance.

Purpose of the Study:

• To develop and validate personalized treatment rules for azithromycin in children with watery diarrhea using machine learning.
• To identify child and clinical characteristics that predict benefit from azithromycin treatment.
• To assess the impact of targeted azithromycin therapy on diarrhea duration and severe outcomes.

Main Methods:

• Utilized data from a randomized trial of azithromycin for watery diarrhea.
• Employed an ensemble machine learning procedure to develop personalized treatment rules based on diagnostic, child, and clinical characteristics.
• Estimated treatment proportions and average benefits for each derived rule.

Main Results:

• Personalized treatment rules recommended azithromycin for approximately one-third of 6,692 children.
• Targeted azithromycin reduced diarrhea risk by 10.1% and re-hospitalization/death risk by 2.4% compared to placebo.
• Host characteristics were more predictive of benefits for severe outcomes than pathogen diagnostics.

Conclusions:

• Machine learning can identify children most likely to benefit from azithromycin for watery diarrhea.
• Targeting antibiotic treatment based on host characteristics may be feasible without pathogen diagnostics.
• Personalized azithromycin strategies hold promise for improving pediatric diarrhea management.