Mapping — Apple Health / Health Connect

Source: Apple Health export.xml. Pillars: A + B. Mapper: mapAppleHealth — mapAppleHealth(xml) → records[]. Pillar A is mature here: discrete and interval quantity samples, sleep-stage category series, and HKWorkout → Session all map with no model gaps.

Structural correspondence

Apple Health (export.xml)	OpenBody
HKQuantityTypeIdentifierHeartRate / StepCount / BodyMass	Measurement (quantity), type → canonical heart_rate / step_count / body_mass
any other HKQuantityTypeIdentifier…	Measurement (quantity), type → apple:HK… (lazy token)
a daily interval aggregate (startDate ≠ endDate)	quantity Measurement whose window is startTime/endTime (§4.3)
HKCategoryTypeIdentifierSleepAnalysis	multiple category Measurements (sleep_stage) over adjacent intervals (§4.3 — not a sampleArray)
HKWorkout	Session + a continuous WorkUnit (distance + energy + time), disciplines from workoutActivityType
sourceName	provenance.device.model; method: "sensor", sourceApp: "apple"

Output examples

A discrete heart-rate sample → a quantity Measurement (count/min → UCUM /min):

{
  "id": "apple-q-1",
  "recordType": "Measurement",
  "subject": "subj-001",
  "type": "heart_rate",
  "quantity": 64,
  "unit": "/min",
  "startTime": "2026-06-20T07:00:00Z",
  "endTime": "2026-06-20T07:00:00Z",
  "provenance": { "method": "sensor", "sourceApp": "apple", "device": { "manufacturer": "apple", "model": "Apple Watch" } }
}

A sleep stage → a category Measurement over its interval:

{
  "id": "apple-sleep-7",
  "recordType": "Measurement",
  "subject": "subj-001",
  "type": "sleep_stage",
  "category": "asleep_deep",
  "startTime": "2026-06-20T02:10:00Z",
  "endTime": "2026-06-20T02:48:00Z",
  "provenance": { "method": "sensor", "sourceApp": "apple", "device": { "manufacturer": "apple", "model": "Apple Watch" } }
}

An HKWorkout → a Session with a continuous WorkUnit:

{
  "id": "apple-workout-12",
  "recordType": "Session",
  "subject": "subj-001",
  "disciplines": ["running"],
  "intent": "train",
  "startTime": "2026-06-20T06:30:00Z",
  "endTime": "2026-06-20T07:12:00Z",
  "provenance": { "method": "sensor", "sourceApp": "apple", "device": { "manufacturer": "apple", "model": "Apple Watch" } },
  "workUnits": [
    {
      "id": "apple-workout-12-wu",
      "recordType": "WorkUnit",
      "scoring": "continuous",
      "performance": {
        "distance": { "absolute": { "value": 10.0, "unit": "km" } },
        "energy": { "absolute": { "value": 520, "unit": "kcal" } },
        "time": { "absolute": { "value": 2520, "unit": "s" } }
      },
      "links": [{ "type": "measuredBy", "ref": "apple-q-1" }]
    }
  ]
}

Lazy type tokens (§4.4)

Known HealthKit identifiers resolve to canonical tokens (heart_rate, step_count, body_mass); anything unmapped is preserved source-namespaced as apple:HK…. Ingestion never fails for lack of a clinical code — vendor metrics survive, and external-code mapping is deferred to output.

Health Connect parity — not a separate mapper

Android Health Connect maps the same way and needs no new model features, so it is covered by this mapper:

Health Connect	OpenBody
StepsRecord	interval quantity Measurement
HeartRateRecord.samples	sampleArray (or discrete quantity)
SleepSessionRecord stages	category Measurements (like Apple)
ExerciseSessionRecord (+ laps/segments/route)	Session → Blocks/Exercise; route → location sampleArray
record title	the v0.3 name

This shared mapping is itself evidence for the model’s neutrality: two of the largest consumer health platforms reduce to the same canonical shapes.