Estimating the Dynamic Diarrhea Effects on Childhood Growth with Statistical Models

Diarrhea effects on children health have been a popular topic in biomedical research. Previous studies have focused on short term effect on mortality and morbidity. Recent evidence has raised interests on the long term effect on childhood growth. Instead of treating diarrhea effect as constant as what previous studies do, we propose a dynamic nonparametric model to estimate diarrhea effect as a function, studying how long the diarrhea effect lasts and its changes over time. Simulation study shows that our model can capture the length of diarrhea effect and quantify the effect curve simultaneously. In addition to the original proposed model, we also develop four extended models to estimate curves leveling off to a nonzero constant, to take into account additional covariates, to estimate multiple curves, and to model curves with more than one dimension. The proposed models are applied to the data from NIH cohort study collected from children in Bangladesh. Results of the original model show that the diarrhea effect on children's HAZ score starts to show up at 3 months, becomes most significant at around 9 months with a decrease of HAZ -0.013, and levels off to zero after 15 months. Overall, our models provide new statistical tools to quantify the relationship between diarrhea and childhood growth in a dynamic fashion, which gives us insights on the changing pattern and the effect window of diarrhea effect.