By PulsHealth. Updated
Research reviewed .
What does an Apple Health food record contain?
A food correlation groups related nutritional quantity samples into a single object. An energy sample, a protein sample and a sodium sample can describe the same food while keeping their own units. Apple documents food correlations as a way to preserve that grouping.
A source may save an ingredient, a packaged item or a whole meal. The identifier does not establish a serving size, meal category or complete nutrient panel. The optional food-type metadata is a string, rather than a universal food classification.
How is a food object different from its nutrients?
| Record | What it represents | What to retain |
|---|---|---|
| Food correlation | A group of related samples | Object identity and explicit member links |
| Dietary energy | Recorded food energy | Amount, energy unit and source |
| Protein, carbohydrate or fat | Separate recorded nutrient amounts | Component identity and mass unit |
| Food description | Optional descriptive context | Original string and writing source |
There is no useful single numeric range for a food object. Sum compatible quantities within the same nutrient type, after checking source overlap and serving conventions. Adding calories to grams or counting grouped objects as nutritional intake has no coherent meaning. A missing component is unknown, rather than zero.
Can the database reconstruct meals?
The reviewed database schema stores nutrient quantities but has no food-correlation object or membership table. The coverage review checked schema support and component availability in the 365-day example window. It cannot recover meal membership, correlation counts, component completeness or meal-level cadence.
Matching timestamps and sources does not restore the missing links: several foods can be logged together, and a source can backdate or batch imports. Keep standalone nutrient analyses separate and describe grouping as unavailable. A nutrition chart would imply evidence that this snapshot cannot provide, so none is shown here.
This type is also absent from the reviewed PulsHealth sync catalog. Database availability therefore does not establish what is present in the phone’s HealthKit store.
Do food timestamps identify meal times?
Food entries follow logging and synchronization behavior. Their start and end times need not mark the beginning and end of eating. A timestamp alone cannot establish meal timing or a response to a particular food.
A useful meal-level comparison needs verified ingestion times, explicit grouping and adequate recording coverage. Blood glucose, body mass and activity energy also have their own measurement schedules. Nearby values do not prove that a meal caused a change, and activity energy alone is not total energy expenditure.
What should a food analysis check first?
Check whether the export retains object identities and member links, then inspect the writing source, serving convention and component units. Compare source-specific recording days and duplicate-import candidates before adding totals. A logged diet can omit meals, beverages or individual nutrients; sparse logging cannot establish dietary deficiency or a complete intake history.
See Dietary Energy, Carbohydrates and Protein for quantity-specific interpretation.
Understanding your records
Recording frequency, source and coverage affect what you can conclude from this data. Read the recording guide.
- One record
- A grouped object with explicit nutritional sample membership, not a cumulative quantity.
- Frequency
- Source-specific food logging and synchronization; no automatic continuous measurement cadence.
- Context
- Use explicit food-object membership and documented serving context; timestamps alone cannot identify a meal or when its components were consumed.
- Coverage and gaps
- Missing observations are unknown, not zero intake or complete capture.
- Summaries
- Query food objects as groups. Sum only compatible component quantities; timestamp coincidence never establishes membership.
- Metadata
- Optional food-description and user-entered metadata provide context only when present. Missing metadata does not prove automatic sensing or complete logging.
Recording guidance reviewed 2026-10-14