| name | dietary-assessment |
| description | Methods for assessing what people actually eat — 24-hour recall, food frequency questionnaires, diet diaries, biomarkers, duplicate-plate studies, and controlled feeding. Covers the relative strengths and biases of each instrument, how to choose among them for a given question, and how to interpret results in the presence of measurement error. Use when a user wants to estimate population-level intake, individual-level intake, or evaluate the plausibility of a dietary claim that depends on how intake was measured. |
| type | skill |
| category | nutrition |
| status | stable |
| origin | tibsfox |
| modified | false |
| first_seen | "2026-04-12T00:00:00.000Z" |
| first_path | examples/skills/nutrition/dietary-assessment/SKILL.md |
| superseded_by | null |
Dietary Assessment
"What do people actually eat?" is harder to answer than it looks. Every large claim about diet and health depends on answering it for thousands of people over months or years, and every instrument that attempts to do so has known biases. This skill catalogs the common instruments, documents the biases each one introduces, and gives heuristics for choosing the right tool for a question and reading the resulting numbers with appropriate skepticism.
Agent affinity: lind (controlled-trial design), ancel-keys (population-scale dietary assessment)
Concept IDs: nutrition-assessment-methods, nutrition-study-design, nutrition-measurement-error
The instrument menu
24-hour dietary recall
What it is. A trained interviewer walks the participant through everything eaten and drunk in the previous 24 hours, prompting with portion-size aids and a multi-pass protocol (quick list, meal structure, forgotten-foods probe, detail pass, final review). The USDA Automated Multiple-Pass Method is the reference implementation.
Strengths. Open-ended; captures foods not on any predefined list. Short-term memory is more accurate than long-term. With multiple non-consecutive recalls, can estimate usual intake with statistical modeling (the National Cancer Institute method).
Weaknesses. A single recall does not represent habitual intake; day-to-day variability is large. Portion-size estimation is imprecise without aids. Respondents still underreport energy — roughly 10–15% on average in the AMPM, better than FFQ but not negligible.
When to use. Short-term energy-balance studies, nutrient-intake estimates for large surveys (NHANES uses two recalls per participant), situations where the food list is unknown in advance.
Food frequency questionnaire (FFQ)
What it is. A structured questionnaire asking how often the respondent has eaten each of 100–150 foods over a reference period (typically the past year), with optional portion-size questions.
Strengths. Cheap, rapid, low respondent burden. Well-suited to ranking people for relative-risk epidemiology. Historically the instrument of choice for cohort studies like Nurses' Health, EPIC, and the Harvard cohorts.
Weaknesses. Underreporting of total energy intake is routinely 20–30% and is not random — foods perceived as unhealthy are differentially underreported. Long reference period strains memory. The food list constrains what can be reported. Portion-size estimation over a year is nearly impossible. FFQ-based studies with small effect sizes should be read with substantial skepticism.
When to use. Large cohort studies focused on ranking, not absolute intake. Do not use for clinical or individual-level assessment.
