Quantitative bias analysis to adjust for misclassified pediatric diarrhea in a cluster-randomized trial
Investigator: Kristen Aiemjoy, MSc
Sponsor: NIH National Institute of Child Health and Human Development
Diarrhea is a significant cause of death and sickness among children globally. The most common way to measure diarrhea, using caregiver-reported symptoms, may not be very accurate. This study will validate caregiver-reported stool consistency and use quantitative bias analysis to adjust for misclassified diarrhea in a cluster randomized trial.
This mentored Ruth L. Kirschstein National Research Service Award proposes to provide the trainee, a doctoral student in epidemiology at UCSF, training in cluster-randomized trials, validation studies and quantitative bias analysis. Guided by a team of senior mentors, the trainee will gain the necessary skills, practical experience, and knowledge to launch her academic career as an independent researcher dedicated to using the best epidemiologic methods to evaluate interventions aimed at reducing diarrheal disease and improving child health globally. The proposed training plan and research experience are supplemental to the standard PhD curriculum. Building on prior work of the trainee, this F31 focuses on caregiver report of stool frequency and consistency to measure pediatric diarrhea. Diarrhea is the second leading cause of death and illness among children under five globally. Most studies of pediatric diarrhea use caregiver reported symptoms to quantify disease status. Caregiver reported diarrhea symptoms have never been validated against a gold standard measure. Validation studies assess the accuracy of a measurement and generate misclassification probabilities that can be used in quantitative bias analysis to adjust results for measurement error. Using pictures of stool consistency may improve the validity of caregiver reports. One such tool is the Bristol Stool Form Scale, a pictorial scale of five stool consistency types. The proposed project uses data collected from a cluster-randomized trial studying the effect of constructing hand-dug wells on various child health outcomes in rural Ethiopia. We will compare caregiver's report of stool consistency to the consistency of a collected stool sample. Using quantitative bias analysis techniques, we will use the resulting misclassification probabilities to adjust the trial data for misclassified diarrhea. We will also assess if the Bristol Stool Form Scale improves the accuracy of reported stool consistency. The research program consists of three aims:
1) (a) Establish the validity of caregiver-reported stool consistency and (b) Use quantitative bias analysis to account for misclassified stool consistency and
2) Assess the impact of using the Bristol Stool Form Scale on the validity of caregiver-reported stool consistency.
Good outcome measurement is a cornerstone of rigorous quantitative research; the results of this research will provide valuable information on the validity of caregiver-reported stool consistency, generate misclassification probabilities that can be used to correct for measurement error and evaluate a potential low-cost method to improve reporting accuracy.