The Accuracy of Portion Size Reporting on Self-Administered Online 24-Hour Dietary Recalls Among Women With Low Incomes.

Background: Accurately estimating portion sizes remains a challenge in dietary assessment. Digital images used in online 24-hour dietary recalls may be conducive to accuracy.

Objective: The current analyses were conducted to examine the accuracy of portion size estimation by women with low incomes who completed 24-hour dietary recalls using the online Automated Self-Administered 24-hour Dietary Assessment Tool (ASA24) in the Food and Eating Assessment Study II.

Design: True dietary intake was observed for 3 meals on 1 day through a controlled feeding study conducted from May through July 2016. The following day, participants completed an unannounced 24-hour dietary recall using ASA24, independently or with assistance in a small-group setting.

Participants/setting: Participants included 302 women aged 18 to 82 years living in the Washington, DC, area who met the income thresholds for the Supplemental Nutrition Assistance Program.

Main Outcome Measures: The accuracy of portion size estimation was assessed by comparing the weight truly consumed (observed) and the weight reported for predetermined categories of foods and beverages.

Statistical Analyses Performed: The differences between observed and reported portions were examined and linear regression tested differences by recall condition. Analyses were conducted by condition and repeated with stratification by racial/ethnic identity, education, and body mass index.

Results: On average across foods and beverages, reported portion sizes were 7.4 g (95% CI, 4.3-10.5) and 6.4 g (95% CI, 2.8-10.0) higher than observed portion sizes in the independent and assisted conditions, respectively. Portion sizes were overestimated for small pieces and shaped foods in both conditions, as well as for amorphous/soft foods in the assisted condition and underestimated for single-unit foods in both conditions. Misestimation was fairly consistent by participants' race/ethnicity, education, and body mass index, to varying magnitudes.

Conclusions: Women with low incomes overestimated the amounts of foods and beverages consumed across several categories using online 24-hour dietary recalls with digital images to support portion size estimation. Assistance with ASA24 had little impact on accuracy.