Wearables in clinical trials, with device and time on each reading

A wearable on a participant's wrist reports a heart rate of 72 at 07:30 on day 14. Kymolog codes the reading to FHIR R4 with LOINC and UCUM, and the trial dataset stores it with the device that took it, the time and the setting.

Request a demo The FDA DHT guidance, explained

Device and time per reading · site or home · LOINC · UCUM

Fig. 1. A wearable reading taken at home and filed as a row in the trial dataset.

Wearable data, coded where it is captured

A continuous glucose monitor (CGM) on participant p-017 reads 142 mg/dL at 07:15 on day 14. The Kymolog™ edge agent records it as a FHIR R4 Observation; FHIR, or Fast Healthcare Interoperability Resources, is HL7's standard for coded health records. The Observation carries LOINC 99504-3, the code for glucose in interstitial fluid, and the value in the UCUM unit mg/dL. It also names the device and the time the reading was taken.

In the trial dataset, the row for that reading keeps the Observation's id. A time-in-range endpoint over days 1–14 then counts rows that each lead back to one reading, one device and one time. Digital endpoints in clinical trials depend on that link to the source reading, as do digital biomarkers in clinical trials computed from the same coded series.

A DHT reading to an endpoint, with provenanceFour stages, left to right. 1, capture: a CGM worn at home takes a reading of 142 mg/dL at 07:15:00, shown in red, from Device/c9e2. 2, code: Kymolog files it as a FHIR R4 Observation with LOINC 99504-3, value 142, UCUM unit mg/dL, time 07:15:00, device Device/c9e2 and subject Patient/p-017. 3, dataset: it becomes one row of participant p-017's trial dataset for day 14, between the readings at 07:10 and 07:20, keeping its Observation id 5d0b. 4, endpoint: for p-017 over days 1 to 14, time in range, LOINC 97510-2, counts the readings between 70 and 180 mg/dL out of all readings, as a percentage. A line runs back under the stages from the endpoint to the capture: every endpoint value traces back to its rows, Observations, device and time.1CAPTURECGM, AT HOME142mg/dLtaken  07:15:00by     Device/c9e22CODEFHIR R4 OBSERVATIONcode     LOINC 99504-3value    142unit     mg/dL (UCUM)time     07:15:00device   Device/c9e2subject  Patient/p-0173DATASETTRIAL DATASETp-017, day 14time   mg/dL  obs07:05  138  5d0907:10  140  5d0a07:15  142  5d0b07:20  145  5d0c4ENDPOINTDAYS 1–14p-017time in rangeLOINC 97510-2:readings in 70–180mg/dL out of allreadings, as %Every endpoint value traces back to its rows, Observations, device and time.
A DHT reading to an endpoint, with provenanceFour stages, left to right. 1, capture: a CGM worn at home takes a reading of 142 mg/dL at 07:15:00, shown in red, from Device/c9e2. 2, code: Kymolog files it as a FHIR R4 Observation with LOINC 99504-3, value 142, UCUM unit mg/dL, time 07:15:00, device Device/c9e2 and subject Patient/p-017. 3, dataset: it becomes one row of participant p-017's trial dataset for day 14, between the readings at 07:10 and 07:20, keeping its Observation id 5d0b. 4, endpoint: for p-017 over days 1 to 14, time in range, LOINC 97510-2, counts the readings between 70 and 180 mg/dL out of all readings, as a percentage. A line runs back under the stages from the endpoint to the capture: every endpoint value traces back to its rows, Observations, device and time.1CAPTURECGM, AT HOME142mg/dLtaken  07:15:00by     Device/c9e22CODEFHIR R4 OBSERVATIONcode     LOINC 99504-3value    142unit     mg/dL (UCUM)time     07:15:00device   Device/c9e2subject  Patient/p-0173DATASETTRIAL DATASETp-017, day 14time   mg/dL  obs07:05  138  5d0907:10  140  5d0a07:15  142  5d0b07:20  145  5d0c4ENDPOINTDAYS 1–14p-017time in rangeLOINC 97510-2: readings in 70–180 mg/dL out ofall readings, as %Every endpoint value traces back to its rows,Observations, device and time.
Figure 2 as a table
StageWhat Kymolog holdsProvenance it keeps
1. CaptureCGM reading 142 mg/dL at 07:15:00, worn at homedevice Device/c9e2, device time
2. CodeFHIR R4 Observation, LOINC 99504-3, UCUM mg/dLsubject Patient/p-017, device, effective time
3. Datasetone row of p-017's trial dataset, day 14Observation id 5d0b
4. Endpointp-017, days 1–14: time in range, LOINC 97510-2: readings in 70–180 mg/dL out of all readings (%)the rows it counts
Fig. 2. A digital health technology reading, from capture to a trial endpoint. The CGM represents any wearable onboarded by configuration; the reading keeps its code, unit, device and time at every step, so the endpoint can be traced back to each reading it counts.

Any device, onboarded by configuration

Your protocol names the devices; Kymolog takes readings from any device via config-driven onboarding. A manifest lists what the device measures, with each LOINC code and UCUM unit, and the edge agent reads the device's own output from then on. Wearables, home blood pressure cuffs, pulse oximeters that read oxygen saturation (SpO₂), electrocardiogram (ECG) patches and CGMs all join this way. DHT clinical trials can then mix device classes in one dataset.

When a protocol allows more than one model, each reading names its model and unique device identifier (UDI), and you can show that the models agree. Kymolog also keeps battery level, signal strength and last-seen time for each device. A gap in a participant's series can then be checked against the device's own record.

Readings from the site visit and from home, in one dataset

At a site visit on 2 October, a monitor in exam room 2 records p-017's heart rate as 68 /min at 09:40. Kymolog ties the reading to p-017 by the room and the time, and links it to the visit encounter. The next morning, a wearable assigned to p-017 sends 74 /min from home over HTTPS, and Kymolog matches it through that device assignment.

Both rows carry LOINC 8867-4 in /min, so they sit in one column. Each row also keeps its setting, its device and how it was matched to the participant. FDA's final guidance on remote data acquisition (December 2023) asks sponsors to evaluate differences between remote and in-clinic measurements. Those three fields support that comparison. Remote patient monitoring in clinical trials and site capture end up in the same table.

Site and remote capture, one datasetTwo sources feed one trial dataset for participant p-017. At the site visit, a monitor in exam room 2, Device/vs-02, takes a reading that Kymolog matches to p-017 by the room and the time, 09:40, and links to the visit, Encounter/e-5120. At home, a wearable, Device/w-31, is assigned to p-017 and sends its readings over HTTPS, so each one is matched by that device assignment. The dataset table holds both as heart rate, LOINC 8867-4, UCUM /min: the site row, 10-02 09:40, 68 /min from vs-02 linked by location, and the remote row, 10-03 07:15, 74 /min from w-31 linked by assignment, highlighted in red. Each row keeps its setting, so remote and site readings can be compared.AT THE SITE VISITmonitor, exam room 2device   Device/vs-02matched  p-017 by room, 09:40visit    Encounter/e-5120AT HOMEwearable, worn by p-017device   Device/w-31matched  p-017 by assignmentsends    over HTTPSp-017, heart rate LOINC 8867-4, /minsettingtakenvaluedevicematched bysite10-02 09:4068 /minvs-02locationremote10-03 07:1574 /minw-31assignmentThe same code and unit from both settings; the setting rides onevery row.
Site and remote capture, one datasetTwo sources feed one trial dataset for participant p-017. At the site visit, a monitor in exam room 2, Device/vs-02, takes a reading that Kymolog matches to p-017 by the room and the time, 09:40, and links to the visit, Encounter/e-5120. At home, a wearable, Device/w-31, is assigned to p-017 and sends its readings over HTTPS, so each one is matched by that device assignment. The dataset table holds both as heart rate, LOINC 8867-4, UCUM /min: the site row, 10-02 09:40, 68 /min from vs-02 linked by location, and the remote row, 10-03 07:15, 74 /min from w-31 linked by assignment, highlighted in red. Each row keeps its setting, so remote and site readings can be compared.AT THE SITE VISITmonitor, exam room 2device   Device/vs-02matched  p-017 by room, 09:40visit    Encounter/e-5120AT HOMEwearable, worn by p-017device   Device/w-31matched  p-017 by assignmentsends    over HTTPSp-017, heart rate LOINC 8867-4, /minsettingtakenvaluedevicesite10-02 09:4068 /minvs-02remote10-03 07:1574 /minw-31The same code and unit from both settings; the settingrides on every row.
Figure 3 as a table
SettingReadingMatched to p-017 by
site visit, exam room 2heart rate, LOINC 8867-4: 68 /min at 2026-10-02 09:40 from Device/vs-02location and time; visit Encounter/e-5120
remote, at homeheart rate, LOINC 8867-4: 74 /min at 2026-10-03 07:15 from Device/w-31device assignment; sent over HTTPS
Fig. 3. Site and remote capture in one dataset. Both settings use the same code and unit, and every row records the setting, the device and how the reading was matched to the participant.

Our walk-through of the DHT guidance maps its other data asks to fields the same way.

The site keeps participant identities

The edge agent carries device values and nothing that names a participant, and Kymolog matches readings to participants inside the install. The site chooses which fields reach the sponsor's dataset, such as a participant number in place of a name.

Wearable data collection runs through the same review queue the site uses for bedside devices. A coordinator can approve, reject or reassign a reading, and Kymolog logs each action with the person and the time.

Show us the endpoints in your protocol

Tell us the endpoint and the device class it depends on. The demo follows one reading from that class into a trial dataset row, then back from the row to the device that took it. We reply within one business day with times to talk.

Request a demo