Intersections of educational attainment, Indigenous identity and race/ethnicity best predicted diet quality among adults in Canada: A conditional random forests analysis.

Background: While it is well-known that diet quality varies according to multiple dimensions of socioeconomic position (SEP), much remains unknown about how these dimensions together shape diet quality. Given that diet quality associated with one SEP dimension (e.g., income) can systematically differ across another dimension (e.g., race/ethnicity), it is necessary to investigate diet quality across SEP intersections.

Objectives: To identify SEP intersections that best predicted lower and higher diet quality among adults in Canada.

Design: Population-based data from the cross-sectional 2015 Canadian Community Health Survey - Nutrition. Data were collected by interviewers who visited selected dwellings to collect household information and administer a general health questionnaire, along with a 24-hour dietary recall.

Participants/setting: Data from 13,617 adults (≥18 years) living in Canada's ten provinces.

Main outcome measures: 24-hour dietary recall data were used to assess diet quality, based on the Healthy Eating Index-2015 (HEI-2015; range 0-100).

Statistical analyses performed: Conditional random forests (CRF), a supervised machine learning technique, were used to identify the four (out of 12) SEP indicators that best individually predicted HEI-2015 scores. The resulting four most important predictors were used to predict diet quality using all possible two-way intersections.

Results: The four most important intersectional predictors of HEI-2015 scores based on CRF variable importance measures were: (1) educational attainment and Indigenous identity and race/ethnicity, (2) educational attainment and household food insecurity, (3) educational attainment and sex/gender, and (4) household food insecurity and sex/gender. Among these four SEP intersections, individuals without a high school diploma living in a severely food insecure household had the lowest (55.7), while individuals without a high school diploma identifying as Middle Eastern had the highest (64.5) predicted HEI-2015 scores.

Conclusion: The SEP intersection defined by educational attainment and Indigenous identity and race/ethnicity was the most important predictor of diet quality among adults in Canada.