Grocery purchase data in the study of alcohol use - A validity study

Area of Science:

• Epidemiology
• Consumer Behavior
• Data Science

Background:

• Traditional alcohol use epidemiology faces declining survey response rates and high costs.
• Population surveys often underestimate total alcohol consumption.
• This study validates grocery purchase data as a complementary epidemiological tool.

Purpose of the Study:

• To validate the use of grocery purchase data for population alcohol use research.
• To assess the accuracy of using loyalty card data to estimate alcohol consumption.

Main Methods:

• The LoCard study analyzed loyalty card data from a Finnish grocery co-operative (over 45% market share).
• 11,818 participants completed a web-based Food Frequency Questionnaire.
• Poisson regression and crosstabulations examined the association and agreement between beer purchase and self-reported drinking frequency.

Main Results:

• A good association (Gamma = .556) was found between beer purchase and drinking frequency.
• Initial agreement was fair (Kappa = .189), but improved to good (Kappa = .463) when data was limited to households making over 61% of purchases from the retailer and categories were collapsed.
• Accuracy of estimation varied based on the proportion of grocery purchases made at the retailer.

Conclusions:

• Loyalty card purchase data can effectively rank individuals by alcohol drinking frequency.
• Estimating beer drinking frequency using purchase data offers fair to good accuracy.
• The accuracy is dependent on the proportion of an individual's total grocery purchases made at the specific retailer.