Leveraging data from a longitudinal birth cohort to improve attribution of diarrhea etiology among children in low-resource settings.
Maria Garcia QuesadaJames A Platts-MillsJie LiuEric R HouptElizabeth T Rogawski McQuadePublished in: The Journal of infectious diseases (2024)
Attributing infectious causes of diarrhea is critical to inform treatment and burden estimates. The attributable fraction (AF) approach based on the association between pathogen quantity and diarrhea has been frequently used but may underestimate incidence. We leveraged data from the multisite birth-cohort Malnutrition and Enteric Disease (MAL-ED) Study, where diarrheal and non-diarrheal stools were collected from 1,715 children from 0-2 years. We compared attribution using a longitudinal AF (LAF) method that considers the temporal association between pathogen quantity and diarrhea symptoms to previously-published AF estimates. For rotavirus and Shigella, attribution did not meaningfully change. For others like adenovirus 40 & 41, astrovirus, norovirus GII, sapovirus, Campylobacter jejuni or C coli, ST ETEC, typical EPEC, and Cryptosporidium, attribution increased, demonstrating longitudinal data may be informative for pathogens with weak associations between quantity and diarrhea. We further derived accuracy-based, pathogen-specific quantity cut-offs that may improve attribution in the absence of longitudinal data.
Keyphrases
- irritable bowel syndrome
- electronic health record
- atrial fibrillation
- big data
- clostridium difficile
- emergency department
- risk factors
- candida albicans
- escherichia coli
- randomized controlled trial
- cross sectional
- artificial intelligence
- pseudomonas aeruginosa
- depressive symptoms
- data analysis
- physical activity
- biofilm formation
- smoking cessation
- gram negative