Shrinkage estimation of long-term water ingestion rates.

Abstract

Water consumption is a necessity for human life, though it also presents an opportunity for exposure to harmful chemicals and toxins. In order to gain a better understanding of the potential levels of chronic exposure, accurate estimates of long-term water consumption are needed. The objective of this study is to estimate long-term water consumption using a nationally representative sample of the US population. In this study, we use a random effects model to obtain shrinkage estimates of average daily water consumption for National Health and Nutrition Examination Survey (NHANES) participants from 2005 to 2010, and compare to their empirical 2-day averages. Our results demonstrate that the shrinkage estimates yielded a reduction in estimated mean water consumption. The 95th percentile was reduced from 3292 to 2529 ml/day. In addition, standard deviation of water consumption for this group decreased from 1052 to 688 ml/day. Similar reductions in the mean and variance were observed stratifying by age and race. Random effects models may provide a more accurate measure of daily water consumption and could be utilized for future exposure and risk assessments.