Abstract
Objectives
A validation study in Burkina Faso compared accuracy of two modes of administration of a 24-hour individual dietary recall (24HR), pen-and-paper (PAPI) and tablet (computer-assisted, CAPI), using weighed food record (WFR) as a benchmark. Sources of divergence between the WFR and each 24HR mode were explored, including number of times items in the WFR were omitted from the 24HR (omissions), and number of times items appeared in the 24HR but not the WFR (intrusions).
Methods
Subjects were 231 rural women aged 18–49 years. For the WFR an enumerator visited the household and recorded all food consumed from early morning until after the evening meal, at home and away. Food prepared at home was weighed before and after cooking; amount taken by the respondent was weighed; leftovers subtracted. For mixed dishes, the enumerator weighed all ingredients prior to cooking, the finished dish, and the amount consumed by the respondent, assuming equal distribution of ingredients. The following day a different enumerator administered a multiple pass 24HR of the previous day. Mixed dishes were recorded in two ways: if the dish was similar to a standard recipe (SR) obtained from previous surveys, the quantity was recorded assuming the ingredients and proportions matched the SR. If the dish was not similar, it was considered a non-standard recipe (NSR), and the enumerator recorded the quantity of each ingredient, the final dish, and amount consumed.
Results
In both modes, omissions outnumbered intrusions. About 1/4 of foods in the WFR were omitted from the 24HR: 26.5% (PAPI); 25.1% (CAPI); about 10% of foods in both 24HR did not match the WFR. For NSRs, omissions and intrusions were less frequent: 15.1% (PAPI) and 16.5% (CAPI) omissions; 2.5% (PAPI) and 5.1% (CAPI) intrusions. SRs were recorded less frequently than individual foods or NSRs. PAPI, 400 foods, 404 NSRs, 91 SRs; CAPI, 386 foods, 411 NSRs, 100 SRs. SRs were more frequently omitted in CAPI (30%) than PAPI (18.7%); and added: CAPI (15.0%); PAPI (7.7%).
Conclusions
Respondents to both 24HR modes were more likely to forget items than recall them erroneously. SRs are intended to reduce respondent burden but were infrequently used. The time saved using SRs should be balanced against possible loss of accuracy. Sources of divergence between 24HR and the WFR need investigation.
Funding Sources
Bill and Melinda Gates Foundation.