01 PERSONAL SITE
/
← RESEARCHER

Interscalar medicine / long-form research note / January 2025

From the Electronic Medical Record to the Patient’s Digital Trace

An electronic health record is very good at documenting encounters. It records diagnoses, prescriptions, test results, procedures and clinical decisions. Yet recovery, deterioration and adaptation do not happen only while a patient is in a clinic. They unfold at home, at work and in transit: an exercise is completed or skipped, a medicine is taken late, sleep is interrupted, pain changes, a caregiver becomes unavailable, a device loses connection, or an appointment is cancelled because transport failed.

The record therefore contains important points on a trajectory, but often not the trajectory itself.

The idea of a patient’s digital trace is to connect selected events between encounters to the clinical record. The goal is not continuous surveillance and not the indiscriminate storage of everything a phone or wearable can sense. It is a purpose-limited, time-ordered account of observations and actions that may help a patient and a care team answer a defined question and take a defined action.

This distinction matters. A longer timeline can make care more intelligible, but a denser dataset can also create noise, false alarms, inequity and new privacy risks. The transition from a medical record to a digital trace should therefore be understood as an extension of the record—not its replacement—and as a change in clinical method, not merely in storage capacity.

Electronic medical or health record (EMR/EHR). The terms vary by jurisdiction and organization. In this article, “record” means the provider-governed longitudinal information used to document and support care. Modern EHRs can already receive patient-generated and device-generated data, so the contrast is not between an “old” EHR and a separate futuristic system. It is between an encounter-centred implementation and a trajectory-centred use of information.

Patient-generated health data (PGHD). The US Office of the National Coordinator for Health Information Technology defines PGHD as health-related data created, recorded, gathered or inferred by or from patients or their designees to address a health concern. Its distinctive features are that patients or caregivers are primarily responsible for capture and that patients direct sharing [1]. Symptoms entered in an app, a home blood-pressure measurement and a caregiver’s report can all be PGHD.

Digital phenotyping. In a widely used formulation, this is the moment-by-moment quantification of an individual-level phenotype in situ using data from personal digital devices [2]. It may include active data, such as a questionnaire, and passive data, such as mobility or phone-interaction patterns. Ecological momentary assessment is a related, older method that samples a person’s current experience or behaviour in their natural environment and reduces some forms of retrospective recall bias [3].

Digital measure or digital biomarker. A raw accelerometer stream is not a clinical measure. An algorithm may transform it into step count, gait speed or a sleep estimate; only then can the result be evaluated for a specific use. Terminology remains unsettled: a 2024 mapping review found 127 different definitions among 128 papers that defined “digital biomarker” [4]. For this reason, this article reserves “digital biomarker” for a measure with an explicit biological or clinical interpretation and adequate evidence. Most events in a digital trace are better called observations, reports, signals or derived measures.

Digital trace, as used here, is the broadest of the five concepts. It is an operational term for a provenance-preserving timeline of clinically relevant events from multiple sources: the patient, caregivers, clinicians, devices, software and organizational systems. It may contain PGHD and digital measures, but also events that are neither—for example, a delayed equipment delivery, a rejected authorization or a missed transport booking. The trace is not a diagnosis and not a digital twin. It is evidence from which a trajectory can be described and hypotheses can be formed.

From raw signal to clinical action

A useful digital trace preserves the difference between five levels:

  1. Signal: sensor samples, a button press, free text, a portal login or a system log.
  2. Observation: a blood-pressure value, a self-rated pain score, detected movement or an acknowledged dose.
  3. Event: an observation placed in context—who or what produced it, when it occurred, how it was captured and with what uncertainty.
  4. Episode or trajectory: a sequence interpreted against a baseline, care plan and concurrent events.
  5. Decision or action: contact the patient, review a medicine, adjust an exercise plan, repair equipment, schedule an assessment—or deliberately do nothing.

Errors become dangerous when these levels collapse. “No steps recorded” may mean immobility, but it may also mean a discharged battery, a removed device, a changed operating-system permission, an algorithm update or a patient who left the phone at home. “Medication taken” may mean that a smart package was opened, not that the correct dose was swallowed. A falling activity trend may precede deterioration, but it may also reflect rain, a holiday, a broken lift or a change of job.

Each clinically used event should therefore retain at least:

  • the time the event occurred and the time it was recorded;
  • the patient, caregiver, clinician, device or software that supplied it;
  • the measurement method, unit and relevant device/algorithm version;
  • whether the value was reported, observed, calculated or clinician-verified;
  • the intended context of use and the linked care-plan goal;
  • completeness, quality flags, uncertainty and reasons for missingness where known;
  • transformations from source data to derived measures;
  • consent, access and retention status;
  • the threshold crossed, the responsible recipient and the resulting action.

Without provenance, a timeline is only a collage. With provenance, it can be audited, corrected and interpreted.

What the trace can add

1. Frequency and timing

A clinic measurement is a sample. Repeated home observations can reveal variability, delayed responses and short episodes that a scheduled visit misses. This is useful when timing is clinically relevant: symptoms after a dose, fatigue after exercise, nocturnal events, or the interval between a warning sign and a response.

2. Function in the patient’s real environment

A person may perform well in a supervised assessment and struggle at home. Conversely, a low clinic score may not represent everyday adaptation. Activity, gait, use of an affected limb and completion of therapeutic tasks can add information about function in context, provided the measurement has been validated in the relevant population. A step-count algorithm tested in healthy adults cannot automatically be assumed to work for a person with neuropathy, a gait disorder or an assistive device. The V3 framework separates verification of the sensor, analytical validation of the algorithm and clinical validation for a defined population and purpose [5]. The FDA’s guidance on remote digital data acquisition likewise emphasizes that a technology must be fit for purpose in its context of use [6]. A 2025 extension, V3+, adds usability validation: even a technically and clinically valid measure can fail if intended patients and staff cannot use it reliably in their actual environment [35]. Usability is therefore part of measurement validity at scale, not cosmetic interface polish.

3. The patient’s voice

Pain, fatigue, dizziness, confidence, participation and treatment burden are not adequately represented by physiological sensing alone. Brief patient-reported measures can make invisible problems visible and give the patient a role beyond being the object of measurement. More data is not necessarily more patient-centred; a trace becomes patient-centred when it includes what the patient considers important and lets the patient see, correct and govern the information.

4. Coordination and system friction

Some outcomes are shaped less by biology than by organization. A rehabilitation session may be missed because transport was unavailable; a dose may be delayed because a prescription was not filled; a caregiver may be unable to help because of work; an assistive device may arrive after the period when it was most needed. Recording these events prevents a system failure from being mislabelled as “non-adherence.”

5. Earlier, proportional response

The trace can support graded responses: self-management guidance for a small deviation, asynchronous review for a persistent trend, and urgent contact for a validated high-risk signal. The value lies in the response pathway, not in the dashboard. Every monitored signal needs an owner, a review interval, an escalation rule and a safe plan for nights, weekends and technical failure.

Practical scenarios

Home rehabilitation

A rehabilitation trace might combine a prescribed plan, sensor-derived repetitions and range of motion, patient-reported pain and effort, clinician feedback, equipment status and participation goals. This can distinguish several superficially similar cases:

  • exercises were not attempted because the instructions were unclear;
  • exercises were attempted but stopped because pain rose;
  • the required movement was completed but the sensor failed to recognize an atypical gait;
  • adherence was high, while functional improvement remained limited;
  • performance improved in training but did not transfer to daily activity.

A randomized pilot trial of a wearable rehabilitation system after stroke reported potential functional benefit, but it tested a three-week programme in rehabilitation institutions rather than unsupervised home use, and its design limits generalization [7]. Such studies support further evaluation; they do not justify assuming that any wearable, for any rehabilitation population, improves outcomes.

Symptoms during treatment

Electronic patient-reported symptom monitoring during cancer treatment is one of the stronger examples of between-visit data linked to a response workflow. In a randomized trial, symptom reports triggered clinical action; subsequent analysis reported longer overall survival in the intervention group [8]. The lesson is not that questionnaires are universally life-extending. It is that repeated reports can matter when the symptoms are meaningful, the escalation rules are clear and the care team can respond.

Cardiovascular monitoring

The Apple Heart Study enrolled 419,297 participants in a prospective, single-group study. Only 0.52% received an irregular-pulse notification; among those who returned an ECG patch after an initial notification, subsequent notifications had a positive predictive value of 84% for atrial fibrillation. The study demonstrated reach and the feasibility of a digital-to-telemedicine pathway, but it was not a randomized outcomes trial, the cohort reflected smartwatch ownership, and the diagnostic yield depended on who completed confirmatory testing [9].

Blood-pressure monitoring illustrates the opposite result. In a 246-person randomized trial, remote monitoring, text prompts, EHR nudges and optional social-support feedback did not significantly improve blood-pressure control over usual care at four months [10]. A functioning data channel therefore does not guarantee a clinical effect. The bottleneck may be treatment intensification, medication access, behaviour, workload or the intervention design itself.

Heart-failure telemonitoring makes the contradiction sharper. In the 1,653-participant TELE-HF randomized trial, an automated telephone monitoring programme after hospitalization did not reduce readmission or death compared with usual care [36]. In the 1,571-participant TIM-HF2 trial, a structured remote patient-management programme for a more specifically defined population reduced the proportion of days lost to unplanned cardiovascular hospitalization or all-cause death [37]. These trials are not a head-to-head comparison: populations, measurements, adherence mechanisms, clinical staffing and response pathways differed. Together they show that “remote monitoring” is not a single intervention. The outcome depends on the care system wrapped around the signal.

Medication and adverse effects

A trace can align prescription, dispensing, self-report, packaging events, symptoms and relevant physiology. It should not silently equate these sources. A refill indicates availability, package opening indicates access, and self-report indicates a patient account; none alone proves ingestion or pharmacological effect. The most useful output may be a conversation about barriers rather than a binary adherence score.

Organizational events and caregivers

Care is often distributed across relatives, home-care staff, therapists, pharmacies and transport services. Their contributions can explain outcomes, but they also create rights and privacy questions. A caregiver’s identity and activity are not merely “patient data.” The minimum necessary event may be “planned assistance unavailable,” not the caregiver’s location, employer or private calendar.

What the evidence does—and does not—show

Research on integrating PGHD with EHRs remains less mature than the volume of digital-health marketing suggests. A scoping review found only 19 eligible studies through 2019; most were pilots, and important themes included resources, delivery into the EHR and preferences for review [11]. A systematic review of remote symptom monitoring integrated with EHRs found technical feasibility but limited robust evidence of improved care outside a small number of settings [12].

Four conclusions follow.

First, technical integration is not clinical effectiveness. A FHIR endpoint, a synchronized wearable and a polished dashboard can transmit an invalid or irrelevant measure perfectly.

Second, prediction is not causation. A symptom may follow reduced activity, but both may be caused by an infection, weather, caregiving burden or treatment change. Longitudinal data add temporal resolution, not automatic causal identification. When causal claims are needed, investigators must define the intervention, comparator, eligibility, time zero, outcome and analysis—ideally by designing a trial or explicitly emulating a target trial [13].

Third, absence of data is itself ambiguous. In mobile-health research, data may be missing not at random: people can stop responding when symptoms, motivation or circumstances change [14]. A retrospective analysis of 1,178 digital-phenotyping participants found that sampling frequency, engagement, phone type and protocol characteristics affected missingness [15]. Missingness should be modelled and displayed, not converted into a normal value or ignored.

Fourth, a measure that adds detail may still be unfit to replace a clinical outcome. A 2026 synthesis covering 111 studies, 19,945 participants and 69 ambulatory-assessment or mood-monitoring protocols in depression and bipolar disorder found inconsistent performance for passive measures, variable engagement and underreported adverse effects. The authors concluded that these measures could add granularity and confirmation but were not robust enough to replace established outcomes [38]. This conclusion is specific to mood disorders and should not be generalized to every sensor or disease, but it demonstrates why claims must remain tied to a population and use case.

Counterarguments and limits

“The EHR is already longitudinal”

Correct. Many records contain years of encounters, messages, remote measurements and patient reports. The title describes a shift in emphasis, not a clean technological break. The proposed trace should use and improve the EHR where possible. Building a separate, ungoverned shadow record would reproduce fragmentation.

“Continuous data are more objective”

They are often more frequent, not more objective. Sensors are affected by placement, motion, physiology, firmware and proprietary algorithms. Patient reports have recall and response biases, but they may measure experiences no sensor can observe. Clinician notes contain expertise and context, but also selection and documentation biases. A defensible trace combines sources without pretending they are interchangeable.

“More monitoring means safer care”

Only if detection leads to a timely, effective and proportional response. Low-specificity alerts can create unnecessary consultations and testing. Repeated alerts also reduce clinician responsiveness: workload and repeated exposure have been associated with alert fatigue [16]. Safety evaluation must therefore count false positives, false negatives, time to review, downstream procedures, patient anxiety, workload and outcomes—not only detections.

“Passive collection reduces burden”

It can reduce data-entry burden, but it can increase surveillance burden. People may alter behaviour because they feel watched, worry about an unexplained risk score, or lose the ability to have an unrecorded day. Consent is not meaningful if monitoring is effectively required to receive care, or if refusal changes insurance, employment or access. Ethical analyses of digital phenotyping consistently emphasize transparency, consent, privacy and accountability [17].

“The trace represents everyone equally”

It does not. In a US survey of 23,974 respondents, wearable ownership was associated with age, income, education, insurance and rural residence [18]. Requiring consumer devices can select a healthier, wealthier or more digitally confident population. Device provision, connectivity support, accessible interfaces and non-digital alternatives are therefore part of data quality—not optional social programmes.

“De-identification solves privacy”

Dense behavioural data are difficult to anonymize. In a mobility dataset of 1.5 million people, four random spatiotemporal points were sufficient to uniquely characterize 95% of individual traces [19]. Linkage risk increases when location, time, routines and health events are combined. Raw location, contact and communication data should not be collected merely because they may become useful later.

“Removing protected attributes makes the algorithm fair”

Not necessarily. Proxy variables can reproduce structural inequities. A widely used population-health algorithm showed racial bias because cost was used as a proxy for health need; changing the target reduced the disparity [20]. Digital-trace systems require subgroup validation, error analysis and scrutiny of the outcome being optimized, not only removal of sensitive fields.

A standards-based architecture

Interoperability is necessary but not sufficient. A practical architecture can use:

  • HL7 FHIR Observation for measurements and simple assertions [21];
  • FHIR QuestionnaireResponse for patient- or caregiver-supplied answers, including capture time and source [22];
  • FHIR CarePlan and Task to connect data to intended work rather than leave it as an isolated stream;
  • FHIR Provenance to record how a resource was created, transformed or signed [23];
  • FHIR AuditEvent for security-relevant access and system events, distinct from clinical provenance [24];
  • FHIR Consent for permissions or denials tied to purposes, recipients and periods [25];
  • IEEE 1752.1-2021 for standardized mobile-health metadata, sleep and physical-activity measures [26];
  • ISO 13606-1:2019 for communicating EHR information through a reference model [27];
  • ISO/TS 82304-2:2021 for life-cycle quality and reliability requirements and a quality label for health and wellness apps; ISO confirmed this edition as current in 2025 [39];
  • W3C PROV-O when provenance must be exchanged beyond FHIR systems [28].

The most important architectural decision is what not to copy into the EHR. High-frequency raw streams may remain in a governed source repository. The record may store a validated summary, the relevant interval, quality metadata, a link to provenance and the clinical action. This reduces clutter while preserving the ability to audit a conclusion.

Interoperability also requires semantic discipline. “Activity,” “adherence” and “sleep” are not self-explanatory variables. Implementers must state the unit, window, algorithm, wear-time rule, baseline, acceptable missingness and clinical meaning. A software or firmware update can change a longitudinal measure even when its label does not change.

Governance: the trace belongs in a care relationship

A defensible programme should begin with a question, not a device:

  1. Define the decision. What uncertainty will the data reduce, and what action could change?
  2. Choose the minimum signal. Collect the least intrusive data with sufficient validity.
  3. Validate the chain. Verify the device, analytically validate the derived measure and clinically validate it for the population and purpose [5]; then validate usability with the intended patients and staff rather than inferring it from laboratory success [35].
  4. Design the response. Set thresholds, owners, service hours, escalation and fallback procedures before collection starts.
  5. Preserve provenance and uncertainty. Keep source, timestamps, transformations, versions, missingness and confidence visible.
  6. Give the patient agency. Explain what is collected and inferred; provide access, correction, pause and withdrawal mechanisms; offer a non-digital route where feasible.
  7. Limit purpose and retention. GDPR Article 5 articulates purpose limitation and data minimization as core principles [29]. Comparable obligations differ by jurisdiction, so legal review must follow the actual deployment context.
  8. Separate care from secondary use. Research, product improvement, marketing, insurance and employment are distinct purposes requiring distinct authority and safeguards.
  9. Measure harms and workload. Evaluate subgroup error, false alerts, downstream utilization, patient burden, clinician time, security incidents and opportunity costs.
  10. Create an exit. Define what happens when a device is discontinued, a vendor closes, consent changes or the programme shows no benefit.

In the United States, data can also cross regulatory boundaries: HHS notes that information sent at a patient’s direction to an app that is neither a HIPAA covered entity nor a business associate is no longer protected by HIPAA Rules in the same way [30]. The FTC’s Health Breach Notification Rule covers many health apps and connected technologies outside HIPAA [31]. These examples should not be generalized as global legal advice; they show why “health data” does not have one uniform legal container.

In the European Union, Regulation (EU) 2025/327 establishes the European Health Data Space (EHDS), distinguishing primary use for care from secondary use such as research and policymaking. It also introduces interoperability and logging requirements for EHR systems and certain interoperable wellness applications [40]. Its application is staged, with the Regulation generally applying from 26 March 2027, and it does not displace the GDPR or other sectoral rules. A digital trace must therefore record not only who accessed data, but for which legally and clinically defined purpose.

What patents tell us

Patents document proposed technical claims and commercial directions, not validated clinical benefit. Three granted US patents and one granted European patent illustrate the design space:

  • US10276188B2 describes deriving mental-health estimates from mobile-device sensor and usage data, including GPS, calls, messages and voice features [32].
  • US11043139B1 describes digital-health interventions that can adapt measurement goals using sensor data and digital-phenotyping markers [33].
  • US12011291B2 describes wearable biomechanical and physiological monitoring for injury prevention and rehabilitation, including movement, temperature and heart-rate signals [34].
  • EP3956907B1 describes active tests and passive smartphone or wearable sensing to assess cognitive, motor and functional symptoms associated with Alzheimer’s disease [41].

They demonstrate that the proposed trace is already an active field of intellectual property. They also make the governance problem concrete: the same technical capacity can support rehabilitation, early warning, behavioural inference or intrusive surveillance. Patent status and freedom to operate must be checked separately for each jurisdiction and date.

Conclusion

The electronic record answers: What was documented when the patient met the healthcare system?

A well-designed digital trace can add: What happened between those meetings, what evidence supports that account, what remains uncertain, and what should happen next?

The word “trace” should not imply that everything traceable ought to be traced. The clinically useful trace is selective, explainable, provenance-preserving and linked to a care decision. It treats missing data as information, predictions as hypotheses, patient reports as essential evidence and organizational barriers as part of the causal context. It is validated for the people who will actually use it and governed by a right to understand, correct, pause and refuse.

The future is therefore not a medical record replaced by an infinite stream. It is a record made more temporally complete—and more honest about the limits of what data can explain.

References / Список литературы

Definitions, measurement and integration

  1. Shapiro M, Johnston D, Wald J, Mon D. Patient-Generated Health Data: White Paper. Prepared for the Office of the National Coordinator for Health Information Technology, 2012. Official PDF.
  2. Onnela J-P, Rauch SL. Harnessing Smartphone-Based Digital Phenotyping to Enhance Behavioral and Mental Health. Neuropsychopharmacology. 2016;41(7):1691–1696. DOI 10.1038/npp.2016.7; PMID 26818126.
  3. Shiffman S, Stone AA, Hufford MR. Ecological Momentary Assessment. Annual Review of Clinical Psychology. 2008;4:1–32. DOI 10.1146/annurev.clinpsy.3.022806.091415; PMID 18509902.
  4. Macias Alonso AK, Hirt J, Woelfle T, Janiaud P, Hemkens LG. Definitions of digital biomarkers: a systematic mapping of the biomedical literature. BMJ Health & Care Informatics. 2024;31:e100914. DOI 10.1136/bmjhci-2023-100914; PMID 38589213.
  5. Goldsack JC, Coravos A, Bakker JP, et al. Verification, analytical validation, and clinical validation (V3): the foundation of determining fit-for-purpose for Biometric Monitoring Technologies. npj Digital Medicine. 2020;3:55. DOI 10.1038/s41746-020-0260-4; PMID 32337371.
  6. US Food and Drug Administration. Digital Health Technologies for Remote Data Acquisition in Clinical Investigations: Guidance for Industry, Investigators, and Other Stakeholders. December 2023, docket FDA-2021-D-1128. Official guidance page.

Clinical scenarios and counterevidence

  1. Guo L, Wang J, Wu Q, et al. Clinical Study of a Wearable Remote Rehabilitation Training System for Patients With Stroke: Randomized Controlled Pilot Trial. JMIR mHealth and uHealth. 2023;11:e40416. DOI 10.2196/40416; PMID 36821348.
  2. Basch E, Deal AM, Dueck AC, et al. Overall Survival Results of a Trial Assessing Patient-Reported Outcomes for Symptom Monitoring During Routine Cancer Treatment. JAMA. 2017;318(2):197–198. DOI 10.1001/jama.2017.7156; PMID 28586821.
  3. Perez MV, Mahaffey KW, Hedlin H, et al. Large-Scale Assessment of a Smartwatch to Identify Atrial Fibrillation. New England Journal of Medicine. 2019;381:1909–1917. DOI 10.1056/NEJMoa1901183; PMID 31722151; trial NCT03335800.
  4. Mehta SJ, Volpp KG, Troxel AB, et al. Remote Blood Pressure Monitoring With Social Support for Patients With Hypertension: A Randomized Clinical Trial. JAMA Network Open. 2024;7(6):e2413515. DOI 10.1001/jamanetworkopen.2024.13515; PMID 38829618.

EHR integration evidence

  1. Tiase VL, Hull W, McFarland MM, et al. Patient-generated health data and electronic health record integration: a scoping review. JAMIA Open. 2020;3(4):619–627. DOI 10.1093/jamiaopen/ooaa052; PMID 33758798.
  2. Gandrup J, Ali SM, McBeth J, van der Veer SN, Dixon WG. Remote symptom monitoring integrated into electronic health records: a systematic review. Journal of the American Medical Informatics Association. 2020;27(11):1752–1763. DOI 10.1093/jamia/ocaa177; PMID 32968785.

Causality, missingness, workload, equity and ethics

  1. Hernán MA, Robins JM. Using Big Data to Emulate a Target Trial When a Randomized Trial Is Not Available. American Journal of Epidemiology. 2016;183(8):758–764. DOI 10.1093/aje/kwv254; PMID 26994063.
  2. Goldberg SB, Bolt DM, Davidson RJ. Data Missing Not at Random in Mobile Health Research: Assessment of the Problem and a Case for Sensitivity Analyses. Journal of Medical Internet Research. 2021;23(6):e26749. DOI 10.2196/26749; PMID 34128810.
  3. Currey D, Torous J. Increasing the value of digital phenotyping through reducing missingness: a retrospective review and analysis of prior studies. BMJ Mental Health. 2023;26:e300718. DOI 10.1136/bmjment-2023-300718; PMID 37197799.
  4. Ancker JS, Edwards A, Nosal S, Hauser D, Mauer E, Kaushal R. Effects of workload, work complexity, and repeated alerts on alert fatigue in a clinical decision support system. BMC Medical Informatics and Decision Making. 2017;17:36. DOI 10.1186/s12911-017-0430-8; PMID 28395667.
  5. Martinez-Martin N, Insel TR, Dagum P, Greely HT, Cho MK. Data mining for health: staking out the ethical territory of digital phenotyping. npj Digital Medicine. 2018;1:68. DOI 10.1038/s41746-018-0075-8; PMID 31211249.
  6. Nagappan A, Krasniansky A, Knowles M. Patterns of Ownership and Usage of Wearable Devices in the United States, 2020–2022: Survey Study. Journal of Medical Internet Research. 2024;26:e56504. DOI 10.2196/56504; PMID 39058548.
  7. de Montjoye Y-A, Hidalgo CA, Verleysen M, Blondel VD. Unique in the Crowd: The privacy bounds of human mobility. Scientific Reports. 2013;3:1376. DOI 10.1038/srep01376.
  8. Obermeyer Z, Powers B, Vogeli C, Mullainathan S. Dissecting racial bias in an algorithm used to manage the health of populations. Science. 2019;366(6464):447–453. DOI 10.1126/science.aax2342; PMID 31649194.

Standards and official governance sources

  1. HL7 International. FHIR R5 Observation. Official specification.
  2. HL7 International. FHIR R5 QuestionnaireResponse. Official specification.
  3. HL7 International. FHIR R5 Provenance. Official specification.
  4. HL7 International. FHIR R5 AuditEvent. Official specification.
  5. HL7 International. FHIR R5 Consent. Official specification.
  6. IEEE Standards Association. IEEE 1752.1-2021: IEEE Standard for Open Mobile Health Data—Representation of Metadata, Sleep, and Physical Activity Measures. Official standard record.
  7. International Organization for Standardization. ISO 13606-1:2019 Health informatics—Electronic health record communication—Part 1: Reference model. Official standard record.
  8. World Wide Web Consortium. PROV-O: The PROV Ontology. W3C Recommendation, 30 April 2013. Official recommendation.
  9. European Union. Regulation (EU) 2016/679 (General Data Protection Regulation), Article 5. Official EUR-Lex text.
  10. US Department of Health and Human Services. The access right, health apps, & APIs. Official guidance.
  11. US Federal Trade Commission. Complying with FTC’s Health Breach Notification Rule. July 2024. Official guidance.

Patents

  1. Systems and methods for identifying human emotions and/or mental health states based on analyses of audio inputs and/or behavioral data collected from computing devices. US Patent US10276188B2. Patent document.
  2. Providing digital health interventions. US Patent US11043139B1. Patent document.
  3. Non-invasive wearable biomechanical and physiology monitor for injury prevention and rehabilitation. US Patent US12011291B2. Patent document.

Sources added in the 29 July 2026 research cycle

  1. Bakker JP, Barge R, Centra J, et al. V3+ extends the V3 framework to ensure user-centricity and scalability of sensor-based digital health technologies. npj Digital Medicine. 2025;8:51. DOI 10.1038/s41746-024-01322-2; PMID 39856145.
  2. Chaudhry SI, Mattera JA, Curtis JP, et al. Telemonitoring in Patients with Heart Failure. New England Journal of Medicine. 2010;363(24):2301–2309. DOI 10.1056/NEJMoa1010029; PMID 21080835; trial NCT00303212.
  3. Koehler F, Koehler K, Deckwart O, et al. Efficacy of telemedical interventional management in patients with heart failure (TIM-HF2): a randomised, controlled, parallel-group, unmasked trial. The Lancet. 2018;392(10152):1047–1057. DOI 10.1016/S0140-6736(18)31880-4; PMID 30153985; trial NCT01878630.
  4. Astill Wright L, Rawsthorne M, Nixon N, Guo B, Morriss R. Recommendations for Research and Clinical Implementation of Ambulatory Assessment, Mood Monitoring, Digital Phenotyping, and Remote Measurement Technology in Mood Disorders: Synthesis of Systematic Review Findings. JMIR Mental Health. 2026;13:e79501. DOI 10.2196/79501; PMID 42228842.
  5. International Organization for Standardization. ISO/TS 82304-2:2021 Health software—Part 2: Health and wellness apps—Quality and reliability. Confirmed current in 2025. Official standard record.
  6. European Union. Regulation (EU) 2025/327 on the European Health Data Space. Official Journal of the European Union, 5 March 2025. Official EUR-Lex text.
  7. Digital biomarker. European Patent EP3956907B1. Grant document published 20 August 2025. Patent document.