Development, validation, and application of a machine learning model to estimate salt consumption in 54 countries

Area of Science:

• Public Health
• Epidemiology
• Machine Learning

Background:

• Global initiatives aim to reduce population salt intake.
• Monitoring salt consumption is hindered by a lack of comprehensive population-based data.
• Existing methods for assessing salt intake are often resource-intensive.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for predicting population-level salt consumption.
• To apply the ML model to estimate mean salt intake across diverse countries.
• To provide a scalable tool for monitoring salt consumption in the absence of direct measurement.

Main Methods:

• Developed a supervised ML regression model using data from 21 national surveys (49,776 individuals).
• Predictors included sex, age, weight, height, systolic, and diastolic blood pressure.
• Validated the model against observed salt intake and applied it to 54 additional surveys (166,677 individuals).

Main Results:

• The ML model demonstrated high accuracy in predicting mean salt intake, with no substantial differences from observed values (p<0.001).
• Predicted mean salt consumption varied globally, from 6.8 g/day in Eritrea to 10.0 g/day in American Samoa.
• Highest predicted intakes were observed in the Western Pacific region, with the lowest in Africa.

Conclusions:

• A machine learning model utilizing readily available predictors can accurately estimate population-level daily salt consumption.
• This ML approach offers a viable solution for monitoring salt intake in populations lacking direct urine sample data.
• The model supports global efforts to track and manage salt consumption for public health improvement.