Predicting habitual water intake from lifestyle questions

Area of Science:

• Nutritional Science
• Biostatistics
• Computational Biology

Background:

• Traditional methods for assessing water intake, such as selective recall and descriptive dietary records, are labor-intensive.
• Accurate and efficient methods for determining habitual water intake are crucial for health management.

Purpose of the Study:

• To develop and validate a simplified lifestyle questionnaire to predict habitual water intake (SQW).
• To evaluate the SQW's accuracy and efficiency compared to established dietary assessment methods.

Main Methods:

• Utilized crowdsourcing and machine learning (random forests) to develop the SQW from 305 lifestyle questions.
• Employed synthetic minority oversampling and regression analysis for model training.
• Validated the SQW against selective recall and descriptive dietary records in a cross-sectional study, with data stratified by sex.

Main Results:

• The SQW demonstrated a Spearman's correlation coefficient of 0.60 for total water intake against the selective recall method.
• A Pearson's correlation coefficient of 0.50 was observed for total water intake.
• Validations confirmed correlations across various sources (non-alcoholic beverages, alcohol, food) and demographic factors (sex, season).

Conclusions:

• The simplified questionnaire (SQW) provides a quick and accurate method for assessing habitual water intake.
• Dissemination of the SQW is expected to enhance health management through easier determination of individual water consumption.