Nutrition quality of food purchases varies by household income: the SHoPPER study – BMC Public Health

Study population and recruitment

The sample was composed of Chicago households enrolled in the Study of Household Purchasing Patterns, Eating, and Recreation [SHOPPER] [15], a cross-sectional study of behavioral and socioeconomic correlates of food purchasing patterns [ClinicalTrials.gov identifier: NCT02073643]. A convenience sample was recruited from the community between 2014 and 2016 through posted flyers, newspaper advertisements, mailings, craigslist.org, word-of-mouth, and other methods. Interested individuals completed a telephone screening to assess eligibility. Adults who reported making ≥75% of their household’s food purchases were eligible to participate. Exclusion criteria included: 1) non-fluent in English, 2) not living in Chicago, 3) major food allergies or sensitivities, 4) religious/spiritual or medical dietary restrictions that could impact food choice, and 5) living in temporary or group housing or living with a roommate with whom food is shared. Of 347 households screened, 300 (86.5%) met eligibility criteria and 209 (69.7%) enrolled. Five participants were withdrawn from the study because of scheduling conflicts that arose during the 14-d assessment period (n = 3) or due to noncompliance with the protocol (n = 2). Two additional participants were not included in the analysis reported here because food receipts were not returned to the research team. The final analysis sample included 202 participants. Participants were compensated $100 for completing all four assessments. Written informed consent was obtained from all participants. Study procedures were approved by the Rush University Medical Center Institutional Review Board.

Measures

Food purchases and receipt collection

Participants were trained by research staff to collected their food purchase receipts and complete annotation procedures throughout the 14-d measurement period. Research staff visited participants’ homes four times during the 14-d measurement period to collect food purchase receipts from participants, with phone calls between to enhance adherence to the food purchase receipt protocol. The receipt data collection protocols are adapted from our previous research studies [5, 7, 15, 17,18,19].

The primary household food shopper was trained to collect and annotate food purchase receipts from all household members on a daily basis (even for purchases without a receipt). Annotation sheets were completed by the participant that included the date, time, source and location, payment methods and foods purchased, including item quantity, size, price, and brief description. Color coded stickers were applied by the participant to both the receipt annotation sheet and the food packages. Food packages were saved for research staff to have direct access to the nutrition information. Details about foods without packaging or nutrition labeling (e.g., fresh produce, deli items, bulk nuts/candy) were recorded by researchers during each of the four data collection home visits. Research staff queried participants about any foods purchased that were consumed immediately and therefore had neither receipts nor food packages with nutrition information (e.g., carry-out or restaurant meals). Ready-to-eat foods that could not be accurately characterized (e.g., a buffet meal purchased and consumed by a household member other than the primary shopper) were deemed “non-codable” and were not subjected to nutrient analysis (< 1% of all purchases).

Food purchase nutrient analysis and diet quality

The Nutrition Data System for Research (NDS-R) [20], was used to compute the nutritional analysis of household food purchases. NDS-R is a database that contains nutrient information for over 18,000 foods and is constantly updated for accuracy and to include newly available foods. The Healthy Eating Index-2010 scoring system [16] was used to compute the diet quality of the food purchase data once entered into the NDS-R software system. The HEI-2010 scores the nutrient densities (kcal/g, per 1000 kcal) for 12 key dietary components on a continuous scale based on conformity to the Department of Health and Human Services’ 2010 Dietary Guidelines for Americans [1]. The 12 component scores are summed to obtain a total score with a maximum of 100 points, with higher scores reflecting better overall diet quality. HEI sub-scores examined here included the following: total fruit; whole fruit; total vegetables; greens and beans; whole grains; dairy; total protein foods; seafood and plant proteins; fatty acids; refined grains; sodium; empty calories. The following food groups created by the NDS-R food coding system were also examined as a second method to describe the quality of the household food purchases: 1) fruits; 2) vegetables; 3) sugar-sweetened beverages (SSBs); 4) sweet baked items; 5) packaged snack foods; 6) frozen desserts; 7) other desserts; 8) candy. The dollars spent on each food category was divided by the total dollars spent from grocery and other stores (excluding restaurants). Of the 2342 receipts collected, 1349 (57%) were from stores and 993 were from eating out or other sources. Only receipts from food stores were included in the analysis of dollars spent.

Demographic and social variables

The primary shopper self-reported age, gender, ethnicity/race, educational attainment, employment, marital status, household size and composition and household income. The income-to-poverty ratio was calculated by dividing annual household income by the current Federal Poverty Threshold [21], which accounts for the number of adult and child family members in the household.

Statistical analyses

The analytic sample includes 202 subjects with complete food purchase, diet recall, and sociodemographic data. Analyses were performed using SAS 9.4 (Cary, NC). Descriptive statistics were calculated to characterize the study sample and food purchasing variables. The food purchase variables derived from the receipt data include the HEI-2010 scores and component scores, and dollar amount spent within pre-specified food categories. These values were calculated for all food purchases combined. However, for the dollars spent variables, purchases from restaurant / eating out sources were excluded due to the inability to determine prices for foods and beverages purchased as a combination (e.g., meals including an entrée, side and beverage with a single price). Models were examined using a three-level category of income-to-poverty ratio as the independent variable. Cutpoints were selected based on values previously used for national data [21]: Low: 0–1.3 (n = 49); Medium: 1.4–3.4 (n = 78); High: > 3.5 (n = 74). High income-to-poverty ratio indicates higher income. Adjusted models were examined that included covariates that might be associated with food spending: education, race and marital status. Unadjusted and adjusted models are shown in the tables below. Results were considered statistically significant where p < .05.