Current EHR state

The EHR succeeds as an institutional record and fails as a cognitive interface.

The clinical case for replacement is not that EHRs are bad systems. It is that they were optimized for institutional operations and have been asked to do something they were not designed for — serve as the primary cognitive interface for clinical decision-making in a fragmented, multi-institution, AI-mediated care environment.

~⅔ U.S. hospitals on one vendor EHR market is dominated by a small number of incumbents, with one vendor approaching majority share of inpatient beds.[1]
EHR time per patient time Multiple time-and-motion studies show clinicians spend roughly twice as long on EHR documentation as on direct patient contact.[2]
>50% Clinician burnout National surveys consistently show clinician burnout above 50%, with EHR usability among the most frequently cited drivers.[3]

The EHR is the system the C-suite loves and the bedside clinician dislikes. That asymmetry is not anecdote — it is well-measured, durable, and structural.

Patient access

For the first time, patients have a legal right to computable health data.

2016

21st Century Cures Act

Established patient right to electronic access without special effort, with information-blocking prohibitions implemented through subsequent ONC rulemaking.[4]

2020–

ONC Final Rule and information-blocking enforcement

Operationalized Cures requirements: standardized API access, USCDI as the minimum data set, penalties for information-blocking behavior.[5]

2018–

Apple Health Records on FHIR

Made consumer-grade patient access tangible at scale. Hundreds of health systems now expose patient data through the same SMART-on-FHIR endpoints that any compliant application can use.[6]

2024–

TEFCA and national exchange

The Trusted Exchange Framework and Common Agreement now provides a national backbone for cross-network exchange, including individual access services.[7]

Platform shift

The standards layer is real and broadly adopted.

The infrastructure for a patient-controlled longitudinal record is no longer hypothetical. It is the baseline.

6 of 7 Top U.S. EHR vendors SMART on FHIR is supported by six of the seven largest U.S. EHR vendors.[8]
1,000+ SMART-enabled hospitals More than a thousand U.S. hospitals now expose SMART on FHIR endpoints, providing a real distribution channel for patient-facing applications.[8]
~75% U.S. adults with a smartphone The personal compute layer Guardian Angel assumed in 1994 is now near-universal in high-income countries and rising elsewhere.[9]

Prior waves

What earlier personally controlled record efforts taught us.

The relevant question is not whether the idea is right. It is what specifically blocked prior implementations, and whether those blocks are still in place.

2007–2011

Microsoft HealthVault, Google Health, Dossia

Validated the thesis at scale. Failed primarily because EHR data extraction was manual, fragmented, and not legally guaranteed; the patient-facing AI interface did not exist; and the business model assumed institutional cooperation that did not materialize.[10]

2009

HITECH-era EHR expansion reinforced incumbents

Federal investment accelerated deployment of existing institutional EHRs and entrenched their position. Patient-centered architectures were not the priority.[11]

2010s

Substitutable apps and SMART on FHIR changed the trajectory

The shift from process-heavy interoperability to app-level substitutability on a shared API created the first real path for patient-controlled applications to scale.[12]

The 2007–2011 wave failed for reasons that no longer hold: no patient-side AI, no legal access right, no ubiquitous personal compute, no information-blocking enforcement. The blocks are gone; the architecture is what is new.

Validation pathway

How we will know it is working.

Cohort

Consented longitudinal cohorts

Pilots run on consented cohorts where Guardian Angel is enabled alongside existing care, with prespecified clinical and operational outcomes.

Observational

Real-world outcome comparisons

Comparison of patients with and without Guardian Angel on trajectory-aware diagnostic capture, time-to-recognition for slowly evolving conditions, and shared-decision-making quality measures.

RCT

Targeted randomized trials

For specific conditions where trajectory reasoning has measurable clinical impact (e.g., pediatric growth trajectories, oncology response, chronic disease control), randomized comparisons against standard of care.

Operational

Operational metrics

Clinician documentation burden, time-to-information, escalation rates, missed-finding rates, patient-reported understanding of care plans.

References

Sources cited.

References are provided for the audience to verify and extend. Statistics marked with the year of the underlying study; figures rounded for readability. Where ranges are well-established (clinician burnout, EHR time-on-task), the cited reviews summarize multiple primary sources.

  1. U.S. EHR market concentration — KLAS Research and ONC Health IT Dashboard, recent reporting periods. Verify exact share figures against current ONC data.
  2. Time-and-motion studies of EHR use, including Sinsky et al. (2016), Annals of Internal Medicine; and ambulatory documentation studies summarized in subsequent reviews.
  3. Burnout prevalence and EHR drivers — Shanafelt et al., Mayo Clinic Proceedings; AMA-Hennepin Healthcare studies; ongoing AMA Physician Practice Benchmark surveys.
  4. 21st Century Cures Act (2016), Pub. L. 114-255, Title IV (Delivery), including patient access and information-blocking provisions.
  5. ONC 21st Century Cures Act Final Rule (2020), 45 CFR Part 170; subsequent enforcement and disincentives rulemaking.
  6. Mandl & Kohane, "A 21st-Century Health IT System — Creating a Real-World Information Economy," NEJM (2017); Apple Health Records launch and FHIR-based architecture documentation.
  7. Trusted Exchange Framework and Common Agreement (TEFCA), Office of the National Coordinator; Recognized Coordinating Entity reporting.
  8. SMART on FHIR adoption among major EHR vendors and hospitals — Mandl et al., subsequent updates from the SMART Health IT project at Boston Children's / Harvard DBMI.
  9. Pew Research Center, Mobile Fact Sheet (most recent year); ITU mobile penetration statistics.
  10. Mandl & Kohane, retrospective analyses of HealthVault, Google Health, and Dossia adoption failures; Halamka commentary on personally controlled health record history.
  11. Blumenthal D, "Wiring the Health System — Origins and Provisions of a New Federal Program," NEJM (2011); ONC HITECH meaningful use program documentation.
  12. Mandl & Kohane, "No Small Change for the Health Information Economy," NEJM (2009); Mandl, Mandel, et al., SMART platform papers, JAMIA.

Citation system follows the structure proposed in the project's internal site plan: each marker maps to a single source record. The reference list will be migrated to a versioned, citable artifact as the project scales.