Mental Health API

Mental wellbeing, from what the phone already records.

Sahha scores six daily habits: steps, active hours, extended inactivity, activity regularity, sleep regularity and circadian alignment. The result is a 0 to 100 mental wellbeing score, from models trained against PHQ-9 and DASS-21. No questionnaire to answer and no wearable to own. It is a wellbeing measure, not a diagnosis.

Start free Read the research

Free for 30 days · No credit card

6: daily habits scored
0–100: wellbeing score
0: wearables required
2,000+: study participants

Why Sahha

What a questionnaire cannot tell you.

Carries a smartphone82–97%
Owns a wearable13–46%
Share of adults across surveyed markets. No factor in this score needs the second number.

A signal on the days nobody logs anything

A check-in tells you how someone felt at the moment you asked. Behavior tells you what actually happened on every day in between. All six factors come from a phone in a pocket, so the score keeps reading through weeks when a user never opens your app. It fills the gap between check-ins rather than replacing them.
- Stepsdaily habit
- Active hoursdaily habit
- Sleep regularitydaily habit
mental_wellbeing0.80
Six habits, not a mood guess

The score is not an opaque read on how someone feels. It is six named behaviors: steps, active hours, inactivity, activity regularity, sleep regularity and circadian alignment. Each returns its own sub-score, so you can tell a user which habit moved and point at the one with the most room to improve.
2,000+
participants across 17+ countries in Sahha’s research study

Grounded in PHQ-9 and DASS-21. Not a diagnosis.

The models are trained against clinically used questionnaires, and Sahha publishes the study design, dataset and method openly, so you can read how the score was built. It supports triage toward self-management or professional support. It is not a diagnostic instrument, not peer-reviewed clinical validation, and not a replacement for professional assessment.

Sources

It runs on a phone in a pocket.

Everything the score needs arrives here

Wearables sharpen the inputs. None of them are required.

86 sources deliver step or exercise data. The score asks only for steps, exercise sessions and sleep timing, and the two platforms above supply all three from the handset alone.

Browse all integrations

API reference

Every endpoint, payload and field.

The score with all six factors, each carrying its own sub-score and the goal it was measured against. Three factors is the minimum for a reliable result.

GET /api/v1/profile/score/{externalId}?types=mental_wellbeing

{
  "type": "mental_wellbeing",
  "score": 0.8,
  "state": "high",
  "factors": [
    { "name": "steps", "value": 8000, "goal": 10000, "score": 0.8, "state": "high" },
    { "name": "active_hours", "value": 6, "goal": 12, "score": 0.67, "state": "medium" },
    { "name": "extended_inactivity", "value": 500, "goal": 240, "score": 0.8, "state": "high" },
    { "name": "activity_regularity", "value": 70, "goal": 100, "score": 0.7, "state": "medium" },
    { "name": "sleep_regularity", "value": 85, "goal": 100, "score": 0.85, "state": "high" },
    { "name": "circadian_alignment", "value": 40, "goal": 30, "score": 0.9, "state": "high" }
  ],
  "dataSources": ["age", "activity", "sleep"],
  "scoreDateTime": "2026-08-11T00:00:00+12:00"
}

Score factors 6

Factor	What it measures	From
steps	Total steps taken through the day	Activity
active_hours	Hours with recorded activity or exercise	Activity
extended_inactivity	Time spent sedentary without breaks	Activity
activity_regularity	Consistency of activity patterns across days	Activity
sleep_regularity	Consistency of bed and wake times across days	Sleep
circadian_alignment	Sleep timing against the body clock	Sleep

None of the six requires a wearable. Sub-scores are research-backed curves, not a value divided by its goal.

The behavior the score reads. Ordinary activity and sleep metrics, one typed record per metric per day, reconciled across every device a user has connected.

GET /api/v1/profile/biomarker/{externalId}?categories=activity

{
  "id": "c2d3e4f5-a6b7-4890-cdef-123456789abc",
  "type": "steps",
  "category": "activity",
  "value": "8240",
  "valueType": "long",
  "unit": "count",
  "aggregation": "total",
  "periodicity": "daily",
  "startDateTime": "2026-08-11T00:00:00+12:00",
  "endDateTime": "2026-08-11T23:59:59+12:00",
  "createdAtUtc": "2026-08-12T06:10:00Z"
}

Activity biomarkers 10

Field	Description	Unit
steps	Total steps taken	count
active_hours	Hours with significant activity	hour
active_duration	Total time active	minute
activity_sedentary_duration	Time spent sedentary	minute
activity_low_intensity_duration	Time at low intensity	minute
activity_medium_intensity_duration	Time at medium intensity	minute
activity_high_intensity_duration	Time at high intensity	minute
active_energy_burned	Energy burned through activity	kcal
total_energy_burned	Overall energy burned	kcal
floors_climbed	Flights of stairs climbed	count

Every one of these is available without a wearable. Sleep regularity and circadian alignment come from the sleep category.

A slower classification than the daily score, recomputed weekly and monthly, for segmenting a population rather than reacting to a day.

GET /api/v1/profile/archetypes/{externalId}

{
  "id": "7f2b91ac-3d55-4e18-9c04-2ab7135ef902",
  "name": "mental_wellness",
  "value": "good_mental_wellness",
  "dataType": "ordinal",
  "ordinality": 2,
  "periodicity": "monthly",
  "startDateTime": "2026-08-01T00:00:00+12:00",
  "endDateTime": "2026-08-31T00:00:00+12:00",
  "createdAtUtc": "2026-09-01T13:08:53.322886Z"
}

Wellbeing archetypes 2

Archetype	Type	Values
mental_wellness	Ordinal	poor · fair · good · optimal
overall_wellness	Ordinal	poor · fair · good · optimal

Two of Sahha’s fourteen archetypes speak to wellbeing directly. The rest describe the activity and sleep behavior underneath it.

A score on its own says little. Trends say whether it is moving, and comparisons say whether it is low for this person or low for everyone like them.

GET /api/v1/profile/insight/comparison/{externalId}

{
  "name": "mental_wellbeing",
  "category": "score",
  "value": 0.8,
  "data": [
    {
      "type": "demographic",
      "value": 0.74,
      "percentile": 0.62,
      "percentageDifference": 0.08,
      "properties": { "sex": "male", "ageRange": "30-35" }
    }
  ],
  "startDateTime": "2026-07-13T00:00:00+12:00",
  "endDateTime": "2026-08-11T00:00:00+12:00"
}

The layer that runs both ways. Reserved symptom tags arrive on their own when a user logs one in their phone’s health app, carrying a five-level severity scale, and you post anything else your product tracks to the same timeline as the score.

POST /api/v1/profile/tag/{externalId}

{
  "type": "event",
  "category": "symptom",
  "name": "mood_changes",
  "value": "moderate",
  "source": "acme.mindapp",
  "startDateTime": "2026-08-13T09:15:00+12:00",
  "additionalProperties": {
    "logged_by": "user"
  }
}

Reserved symptom tags 5

Tag	Type	Severity values
mood_changes	event	unknown, not_present, mild, moderate, severe
sleep_changes	event	unknown, not_present, mild, moderate, severe
appetite_changes	event	unknown, not_present, mild, moderate, severe
fatigue	event	unknown, not_present, mild, moderate, severe
memory_lapse	event	unknown, not_present, mild, moderate, severe

Custom tags have no fixed schema, so anything else your product tracks sits on the same timeline: a therapy session, a medication change, an intervention arm.

The individual activity and sleep samples underneath everything else, each naming the device it came from. This is the one layer delivered by webhook only.

POST your webhook endpoint

[
  {
    "logType": "activity",
    "dataType": "steps",
    "externalId": "ext-412",
    "receivedAtUtc": "2026-08-11T21:05:00+00:00",
    "dataLogs": [
      {
        "id": "8fd21a03-4c19-4f7e-93b2-77c5e0a1d884",
        "parentId": null,
        "value": 412,
        "unit": "count",
        "source": "iPhone 16",
        "recordingMethod": "RECORDING_METHOD_AUTOMATICALLY_RECORDED",
        "deviceType": "iPhone17,1",
        "startDateTime": "2026-08-11T08:40:00+12:00",
        "endDateTime": "2026-08-11T08:55:00+12:00",
        "additionalProperties": {}
      }
    ]
  }
]

The ideas behind the factors, and the instruments the model was trained against. Each links to the research or guide that covers it in full.

PHQ-9 and DASS-21: Two clinically used questionnaires. PHQ-9 screens depression severity; DASS-21 measures depression, anxiety and stress. The models were trained against responses to both. Read the research
Circadian alignment: Whether the sleep window sits where the body clock expects it. Misalignment shows up in the score even when someone is sleeping enough hours. Read the guide
Activity regularity: How consistent movement is from one day to the next, scored separately from how much of it there is. Erratic beats sedentary less often than people assume.
Extended inactivity: Unbroken sedentary time rather than total sedentary time. A long uninterrupted block reads differently from the same hours spread across a day.
Passive digital phenotyping: Inferring behavioral and mental health signal from data a device already collects, instead of asking someone to report it. Read the research
Not a diagnostic: The score supports triage toward self-management, behavioral change or professional advice. It does not diagnose a condition and does not replace clinical assessment. Read the research

REST API
Pull any of it on demand, per profile, whenever your app asks.
Webhooks
Or have it pushed to your endpoint as it arrives, so you never have to poll for it.
Mobile SDK
Read straight from the device on iOS and Android, with no round trip.

What you build

What a continuous wellbeing signal lets you ship.

Show the six

Render the score with the habits behind it, so someone sees which behavior moved rather than a number that arrived from nowhere.

Sahha Scores

Fill the gaps between check-ins

Keep a reading on the weeks a user skips their survey, and bring context to the session where they finally do answer.

Sahha Tags

Notice the drift early

Behavior tends to shift before anyone says something is wrong. Watch activity regularity and sleep timing move against a person’s own baseline, weeks before a questionnaire would pick it up.

Sahha Insights

Recommend one habit

A sub-score points at the specific behavior with the most headroom, which is a far better prompt than telling someone to feel better.

Sahha Biomarkers

Find who to reach first

Surface the cohort most likely to benefit from a program, without screening an entire population one questionnaire at a time.

Sahha Archetypes

Measure an intervention

Compare a cohort before and after a program against a population benchmark, aggregated, without surfacing any individual.

Workplace wellness

How it compares

Three ways to get a wellbeing signal.

	Sahha Mental Wellbeing	Questionnaires alone	Other health APIs
Reads on days users disengage
Works with no wearable
Explains which behavior moved	Six factors
Clinical grounding	Trained on PHQ-9, DASS-21	Is the instrument
Method published openly	Sahha research	Peer-reviewed
Measurement type	Observed behavior	Self-reported	Observed
Effort asked of the user	None	Every check-in	None

Questionnaires are not a competitor. PHQ-9 and DASS-21 are the reference instruments, carry decades of peer-reviewed validation, and this score is trained against them. The column is here because most teams already run one and want to know what a passive signal adds between administrations. Sahha’s own studies are published on this site, not in journals. Other health APIs = wearable and health-data APIs that also offer scores.

FAQ

Questions that come up before you integrate.

Is this a diagnostic tool?

No. It is a wellbeing measure built to support triage toward self-management, behavioral change or professional advice. It does not diagnose any condition, it is not a medical device, and it does not replace clinical assessment by a qualified professional. Products built on it should present it that way to end users.

Can it replace PHQ-9 or a clinical assessment?

No, and it is not designed to. PHQ-9 and DASS-21 remain the reference instruments; the model is trained against them. What the score adds is continuity: a reading on the weeks between administrations, when a questionnaire tells you nothing because nobody filled it in. Most teams run both and cross one against the other.

What data does the score actually need?

Steps, exercise sessions, and bed and wake times. A minimum of three of the six contributing factors is required for a reliable score, so a user who only ever supplies step data will still be scored, on fewer factors.

Does it need a wearable?

No. Every one of the six factors is available from Apple Health or Health Connect without one, which means the phone in someone’s pocket is enough. A wearable improves the precision of the sleep inputs but is never required.

Does it read messages, calls or screen time?

No. The score is computed from activity and sleep data plus demographics. It does not use message content, call logs, keyboard input, location or app usage, and Sahha does not collect them.

What was the model trained on?

Responses to clinically used rating scales, principally PHQ-9 and DASS-21, paired with passively collected activity and sleep data. The research study behind it runs to over 2,000 participants across 17 or more countries, reviewed by a university human ethics committee, and Sahha publishes the design, data collection and ethical framework on its research pages. These are Sahha’s own studies rather than peer-reviewed journal publications.

How often does the score update?

Daily, as new activity and sleep data arrives. Trends and comparisons run over longer windows, and the wellbeing archetypes are recomputed weekly and monthly for a slower read.

Can users log how they actually feel?

Yes, and that is the point of pairing the two. Reserved symptom tags including mood_changes, sleep_changes, appetite_changes, fatigue and memory_lapse carry a severity scale, and custom tags cover anything else your product tracks. Both sit on the same timeline as the score, so you can cut one by the other.

Is Sahha HIPAA and GDPR compliant?

Yes, and SOC 2. Health data is handled under all three, and the end-user consent screens are configurable to carry your brand rather than Sahha’s.

Add a wellbeing signal in days.

One integration reading behavior your users already generate, with the modeling and the validation already done.

Start building Book a demo

Free for 30 days · No credit card

A daily score for every user, no questionnaire and no wearable
Six named habits, each with its own sub-score
Trained against PHQ-9 and DASS-21, with the method published openly
HIPAA, GDPR and SOC 2, with your brand on the consent screen

Mental wellbeing, from what the phone already records.

The day, sample by sample

The behavior underneath

A 0–100 mental wellbeing score

Where someone sits, long term

Moving, or just low?

Your context, on the same timeline