PGHDonFHIR

PGHDonFHIR: An Open, Real-World FHIR R4 Dataset of Patient-Generated Health Data This repository contains an open, real-world dataset of patient-generated health data (PGHD) encoded natively in HL7 FHIR R4. The data originates from a ten-month sleep-health study (20 participants, March 2024 to January 2025) combining consumer-wearable, under-mattress, environmental-sensor, and questionnaire sources. All resources were validated with the official FHIR validator CLI, including terminology-code validation. Contents and structure The dataset is distributed as FHIR R4 resources in JSON. It comprises 107,787 resources: Resource type Count Description Patient 20 Study participants (synthetic demographic profiles, see Privacy) Practitioner 2 Synthetic clinicians for the role-based access demonstration CareTeam 20 One per patient; encodes the patient–practitioner assignment Device 4 Data sources (see device list below) Observation 97,597 All physiological and environmental measurements Questionnaire 3 Morning, evening, and weekly protocols QuestionnaireResponse 10,143 Completed subjective protocols Devices Fitbit Charge 6 (wrist-worn activity tracker): 72,734 Observations Withings Sleep Analyzer (under-mattress sleep sensor; underlying hardware: Withings Aura Sensor V2): 21,168 Observations Milesight WS302 (LoRaWAN noise-level sensor): 1,849 Observations Milesight AM307 (LoRaWAN indoor air-quality sensor): 1,846 Observations FHIR resource to PGHD source mapping The table below summarizes what each FHIR resource type represents in terms of its originating PGHD source. FHIR Resource PGHD Source Patient Anonymized participants Practitioner Anonymized clinicians Observation Fitbit and Withings sleep and activity data (e.g., heart rate, step count, sleep stages) Questionnaire Questionnaire definitions and structure QuestionnaireResponse Questionnaire responses Device PGHD source (e.g., Fitbit) How the resources link Every Observation references the originating Device (provenance at the resource level) and the Patient it belongs to. Each measurement is an Observation with a valueQuantity in UCUM units. Standard terminologies (e.g. LOINC 8867-4 for heart rate) are applied where available; where no suitable code exists (e.g. sound level), code.text is used as a documented plain-text fallback. Interval-bearing physiology (e.g. sleep stages) uses effectivePeriod; point-in-time measurements use effectiveDateTime. Co-measured values that share one physiological event (e.g. the stages within a sleep period) are grouped in a single Observation via component[]. The patient-clinician boundary is expressed as one CareTeam per patient with the assigned Practitioner as a participant, queryable via CareTeam?participant=Practitioner/{id}. A full account of these encoding decisions is given in the accompanying paper. Loading the data The resources can be loaded into any FHIR R4 server. For a quick local setup with Medplum: # Example: POST each resource (or a Bundle) to a running FHIR R4 endpoint curl -X POST https://<your-fhir-server>/fhir/R4 \ -H "Content-Type: application/fhir+json" \ -H "Authorization: Bearer <token>" \ --data-binary @<bundle-or-resource>.json Please refer to the Medplum documentation for further details on loading and querying FHIR resources. If you would rather explore the data without hosting your own server, the live prototype and FHIR API let you browse and query it directly (see Prototype and infrastructure). Privacy, ethics, and data governance This dataset was prepared for open release under the research provisions of Art. 89 GDPR. The study was approved by [ethics committee, hidden until de-anonymization]. To enable publication while protecting participants: Participant demographics in Patient resources were replaced with synthetic profiles; patient, practitioner, and resource identifiers were randomized. Free-text fields containing participant-authored content were removed. Source metadata that could reveal the collection infrastructure was stripped. Every physiological measurement and structured questionnaire response is otherwise preserved unchanged from the source data. The residual re-identification risk of longitudinal physiological patterns was assessed in a formal data protection impact assessment. Note that, because demographics are synthetic, the dataset does not support analyses relating physiological patterns to real demographic attributes. It is intended for researching PGHD integration Prototype and infrastructure A SMART on FHIR prototype demonstrating secure, role-specific clinical access to this dataset is available: Prototype application: https://pghdonfhir.com FHIR API: https://fhir.pghdonfhir.com EHR (Medplum): https://medplum.pghdonfhir.com For testing, the prototype accepts the credentials of any of the 20 synthetic patients or 2 practitioners in this dataset (see Test credentials). The prototype provides a patient-facing dashboard and a clinician-facing dashboard. Everyone is welcome to log in, test the workflows, and look around the data. Please note that the deployment is intended as a showcase only; it is not a production service, and excessive usage may be rate limited. The hosted instances are provided for demonstration and may change or be withdrawn over time. The dataset in this repository is the stable, citable artifact. To work with the data at scale, load it into your own FHIR server as described in Loading the data. Test credentials All accounts below have read-only access. They exist solely to demonstrate role-specific access on the showcase deployment. Sign in at https://pghdonfhir.com or the Medplum app at https://medplum.pghdonfhir.com/signin?project=5637407a-a59a-4ed1-a5cb-25c1072605ae. Please add the ?project=5637407a-a59a-4ed1-a5cb-25c1072605ae query parameter to the sign-in URL for Medplum to ensure you are logging into the correct project. Logging in without the project parameter will result in a User not found error. Practitioners Email Password [e-mail] AHKr&bgF3P$o6ARCqBf! [e-mail] *gG2IR2Li5pSa119BZFM Patients Email Password [e-mail] *U%Kd4*deB6cfqOTw8Gy [e-mail] YyCwK7TJOo@shvl4m!X* [e-mail] 6CHaNh8%I8$P!0veE@A# [e-mail] fz%D$zL%M&D#lL23XHJn [e-mail] vMLNeN&cup!igonZfLW3 [e-mail] xwRvsk0@99POyftJa5h# [e-mail] F01&ccjk89Ohy2pC!iRx [e-mail] LjMmFj*y$Gw&0HRCT4Hc [e-mail] qGL3JIcvLveGvHCmsDyQ [e-mail] nZiS8$9HOlZM%98AsgQf [e-mail] %S8myUNI#mc#gw7228MA [e-mail] AydtIBa3ala1CsQUfc6g [e-mail] TS6CH@WOWeZP1#XYZLt* [e-mail] pyUI6ik2n&reR4*JwmkL [e-mail] dKzPA#@H7lXyPWyjWb3N [e-mail] o@@uLy8zikZj$6FkkAq3 [e-mail] mrDzlsEmLqfw2fFryk@c [e-mail] QUT^lX7YtCF6lrQjbouT [e-mail] kEK8VUg8JFTsfKkbdM1v [e-mail] 56Ut3ewlxX4SEULJzsmu How to cite [the paper is currently under review; citation will be updated upon acceptance] License Creative Commons Attribution 4.0 International (CC BY 4.0)