From Concept to Implementation: A Policy and Standards Roadmap for a Global Medical Record System for Travelers with Chronic Disease

Abstract

International travel by people with multimorbidity and polypharmacy is increasing, and clinicians who treat these patients abroad seldom have access to accurate and current information about them. Mechanisms for cross-border exchange do exist: the International Patient Summary (IPS), European cross-border health services, and national patient portals all move summary clinical data between jurisdictions. What none of them yet provides is automated abstraction of decades of longitudinal records into a short summary, and reliable determination of which diagnoses are active and which medications are current. They also do not provide consistent deployment across jurisdictions that record and code data differently. Sørensen (2022) established the clinical case for a Global Medical Record System (GMRS) and outlined its core data elements. We argue that four recent developments provide a window of policy opportunity: COVID-19 vaccination credentials, which showed that portable health data can be deployed at population scale; formalization of the HL7 IPS as an ISO standard with an implementation guide (FHIR IG); regulatory mandates for health data interoperability in the United States and the European Union; and an ongoing transatlantic collaboration aimed at mapping a GMRS algorithm onto those standards. Herein we make a policy case for coordinated international action, describe the standards pathway of mapping the GMRS to the FHIR IPS and United States Core Data for Interoperability (USCDI), and propose a structured maturation pathway for research and validation via a Denmark–United States collaboration. The GMRS would add to the IPS an abstraction layer that decides which of a patient’s accumulated entries belong in the summary. We present this as a proposed roadmap and policy argument as the GMRS algorithm has not yet undergone formal validation which would be required before actual deployment.

Introduction

International travel reached approximately 1.4 billion arrivals annually in 2019 and was projected in 2017 by the United Nations World Tourism Organization to approach 2 billion by 2030.¹ This expansion is occurring alongside a well-documented epidemiological shift: the traveling population is aging, and an increasing proportion of travelers have multiple chronic conditions compounded by polypharmacy regimens of five or more medications.^2,3 Multimorbidity has become the usual pattern of chronic illness in older populations and is consistently associated with higher healthcare use, functional decline, and mortality,^4–8 while the proportion of adults taking five or more medications has risen steadily across health systems, each additional agent raising the probability of a clinically important drug interaction.^9–11 Older travelers with multimorbidity therefore depart carrying pharmacotherapy that is difficult to reconstruct from memory and difficult for a foreign clinician to interpret.¹² In this population, a medical emergency abroad is a patient safety event whose outcome depends on treating clinicians’ timely access to accurate, current clinical information. The same is true for patients going to other countries on medical tourism.¹³ However, few, if any, patients and clinicians have that information readily available.

Health problems during international travel are common, although the reported frequency varies considerably according to destination, traveler characteristics, and study methods. In a systematic review of studies published between 1976 and 2016, Angelo et al reported that between 6% and 87% of travelers experienced a travel-related illness, with the most reliable estimates suggesting that 43% to 79% of travelers visiting low- and middle-income countries developed some form of illness during or shortly after travel; gastrointestinal infections, particularly travelers’ diarrhea, were the most frequently reported conditions.¹⁴ More recently, Potin et al emphasized that although infectious diseases remain an important cause of morbidity, non-communicable conditions, injuries, and accidents account for a substantial proportion of serious health events among travelers. They noted that fever occurs in approximately 2% to 4% of travelers, while injuries, cardiovascular events, thromboembolic disease, and exacerbations of pre-existing conditions contribute importantly to morbidity and mortality during travel.¹⁵ Together these findings indicate that health problems affect a substantial proportion of international travelers, and that a meaningful share of those events involve exacerbation of conditions a clinician abroad would need the patient’s existing record to manage safely. Illness acquired abroad also presents after return, to clinicians who have no documentation of what was done overseas.¹⁶

The consequences of incomplete clinical information are predictable and serious. Without knowledge of established diagnoses, clinicians might administer contraindicated medications, overlook relevant comorbidities, or be unaware of implanted devices or prior surgical history that would influence clinical management. In-flight medical emergencies, estimated to occur in 1 in every 604 commercial flights, illustrate the acute dimensions of this problem.^17,18 The problem is equally present in foreign emergency departments, post-surgical follow-up after medical tourism, and the post-travel care setting for patients who received treatment abroad for which no documentation was provided. We are, however, not aware of published estimates of how often the absence of a patient’s record leads to medication error, duplicated investigation, or delayed treatment in international travelers specifically. Quantifying that burden is itself a research priority, and its absence is one reason we frame what follows as a development and evaluation agenda.

The term traveler may cover groups with different clinical needs, data-access patterns, and legal circumstances. Hence, we define our primary use case as the conventional international traveler with established chronic disease who requires unplanned or emergency care abroad, where the treating clinician has no prior relationship with the patient, little time, and no practical route to the home record. Two secondary applications follow from the same infrastructure: planned care obtained abroad, or medical tourism, where records can in principle be arranged in advance and the characteristic failure is absence of documentation on return;¹⁹ and post-travel care in the home system, which depends on documentation generated elsewhere. Migrants, displaced people, and long-term expatriates raise further questions of legal status and record custody that we do not address here. Travel medicine has developed largely around pre-travel risk assessment and immunization rather than around portability of a traveler’s existing record,^20–26 and uptake of the pre-travel consultation is itself incomplete,²⁷ although that consultation is an obvious point at which a portable summary could be generated and checked.^28,29

In a 2022 commentary in Clinical Epidemiology, Sørensen described the scale of this problem and proposed a Global Medical Record System (GMRS) as an algorithmic approach to distilling large longitudinal health records into a concise, mobile-accessible, privacy-preserving, multilingual clinical summary for use across borders.³⁰ That article clearly identified the core challenge in which patients with multiple chronic diseases accumulate decades of diagnoses that create unnecessary noise for treating physicians; consequently, algorithms must strike an appropriate balance between sensitivity and specificity.³⁰ Sørensen concluded by calling for action to develop and propagate such a GMRS system. To clarify, GMRS here refers to the abstraction algorithm together with the filtered clinical summary it produces; patient summary means the standardized document format, such as the IPS, in which such content is carried; and portable clinical summary is the generic term for any patient-accessible artifact of this kind. The GMRS output is a filtered summary rather than a complete longitudinal medical record, and we retain the established name for continuity with the 2022 proposal while noting that its content is deliberately partial.

Proposals for portable travel health records are not new; a global travelers’ electronic health record template was described more than a decade ago.³¹ Several classes of initiatives now operate. Standardized summary specifications, principally the IPS, define the sections, resource profiles, and terminology bindings a cross-border summary should use, and they do not determine which of a patient’s accumulated entries should populate those sections.³² Regional exchange infrastructure, notably the European eHealth Digital Service Infrastructure, moves patient summaries and ePrescriptions between member states but does not extend to transatlantic or global exchange, where cross-jurisdictional patient identity resolution also remains unresolved.³³ National portals and personal health records give patients access to their own data within a jurisdiction, and a right of access is not the same as an abstracted, clinically prioritized summary. Cross-border personal health record pilots have demonstrated FHIR-based exchange outside Europe at a scale short of routine service,³⁴ and evaluations of the European Patient Summary have found its content usable for defined purposes, with completeness and consistency varying considerably between contributing systems.³⁵ The gap that remains, and the one the GMRS is intended to fill, is the automated reduction of a long record to a short one whose currency a clinician can trust: which diagnoses are still active, which medications are actually being taken, and which historical entries can be set aside. FHIR-based patient summaries, terminology harmonization, synthetic data testing, clinician panel review, and standards body submission are established components of health information system development, and what we propose as new is their combination in an abstraction layer operating within an existing summary standard.

Although our eventual objective is global implementation, most of the technical and policy detail below concerns Denmark, the European Union, and the United States, because those are the settings we consider now and in which mature electronic records, standardized terminology, reliable patient identity infrastructure, and FHIR-based exchange currently largely coexist. Many countries lack one or more of these. What we propose should therefore be read as a transatlantic proof of concept designed with global extension in mind rather than as a global implementation plan.

We argue that a convergence of recent independent developments has provided a genuine window of policy opportunity, and that a concrete maturation pathway has potential to advance the GMRS from a concept toward a validated, internationally interoperable system. Figure 1 illustrates this pathway at a high level, and Table 1 gives the detail. Figure 2 depicts the relations among the policy drivers, the standards stack, the GMRS algorithm layer, and the clinical deployment contexts.

Figure 1.

High-level view of the seven-stage maturation pathway proposed for the Global Medical Record System (GMRS). Stages 1–3 constitute the standards and mapping phase; stages 4 and 5 constitute the validation phase; and stages 6 and 7 constitute the translation-to-practice phase. The figure is intended as an orientation to the sequence and its phase structure; see more details in Table 1.

Table 1.

Seven-Stage GMRS Maturation Pathway

Maturation Stage	Key Activities and Deliverables	Indicative Duration, Dependencies, and Responsible Parties	Decision Required Before Proceeding
Stages 1 and 2: standards and mapping	Element-level mapping of GMRS data elements to FHIR IPS profiles and USCDI data classes; terminology harmonization (ICD-10 to ICD-10-CM, ATC to RxNorm, NOMESCO to CPT) expressed as FHIR ConceptMap resources; evaluation of the Danish OMOP mappings against a clinical reference standard; legal review of HIPAA, GDPR, and EHDS requirements for cross-border data sharing	About 12 to 18 months. Depends on terminology licensing and access to Danish and United States terminology services. Led by the collaborating academic groups with terminology and legal specialists	Mapping coverage and residual ambiguity are acceptable, and a lawful basis for the planned validation work has been established
Stage 3: FHIR computability	FHIR StructureMap implementation of GMRS abstraction logic; publication of the abstraction rules in inspectable form; publication of a GMRS-on-IPS FHIR IG with versioning policy and conformance test scripts	About 12 months, overlapping stage 2. Depends on completion of core mappings. Led by the development team with HL7 working group input	The IG builds and passes conformance tests, and the rules are documented sufficiently for independent implementation
Stage 4: technical validation	Synthetic cohort testing (Synthea); automated conformance validation; regression testing as mappings and profiles evolve	About 6 to 9 months. Depends on stage 3 deliverables. Led by the development team	Technical error rates are within prespecified limits on synthetic data, with no unresolved safety-critical failure mode
Stage 5: empirical validation	Application of the GMRS algorithm to US EHR and HIE data (adults ≥30 years old, ≥12 months’ observation), sampled with stratification by number of chronic conditions and by medication count; sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) stratified by disease category and data source type; error rates reported separately for safety-critical and lower-risk elements; derived population-level descriptive statistics	About 18 to 24 months. Depends on institutional review board approval, data use agreements, and reference standard construction, which are the principal schedule risks.	Performance meets prespecified thresholds, with the threshold for safety-critical omissions met separately
Stage 6: clinical utility	Clinician and informaticist panel review of generated summaries vs full records; classification of errors as omissions, commissions, or ambiguities; focused assessment of safety-critical content (insulin, opioids, implants, anticoagulants); assessment of usability, time to retrieve information, and effect on simulated prescribing and diagnostic decisions; assessment of patient understanding	About 12 months, following stage 5. Requires recruitment of emergency, travel, and primary care clinicians in both countries, and of patient participants	Clinicians can use the summary reliably under time pressure, and patients find the content acceptable and comprehensible
Stage 7: governance and deployment	Engagement of HL7 International, ISO TC 215, and European Health Data Space technical committees; open-source publication of the FHIR IG; definition of deployment patterns (patient-held, HIE on-demand, travel clinic integration); consent, identity, and audit specification; equity-focused implementation research in at least one lower-resource setting	Continuous from stage 3, intensifying after stage 6. Requires standards development organizations, regulators, patient organizations, and implementation partners	Standards bodies accept the mappings, and a governance and consent model is agreed before any deployment

Figure 2.

Policy convergence and technical architecture of the proposed GMRS–IPS harmonization framework. The converging policy drivers enable the standards stack (FHIR IPS, USCDI, ATC/RxNorm, ICD-10/SNOMED CT, LOINC, and UCUM), over which the GMRS algorithm layer would operate as a clinical abstraction filter, delivering actionable summaries in five categories relevant to the cross-border care context. Data flow from source health systems on the left to the point of care on the right. Consent capture, identity verification, and audit logging apply at the boundary between the algorithm layer and the deployment contexts, where data leave the custody of the source health system. The layers shown are logical rather than physical.

The Case for Acting Now: Converging Policy Developments

The clinical case for a globally portable patient summary has existed for more than a decade.³¹ However, the convergence of structural developments now makes the GMRS more achievable than it has previously been.

COVID-19 Vaccination Credentials

The COVID-19 pandemic was the most consequential natural experiment in health data portability in history. Within 18 months after the first vaccine approvals, dozens of countries had deployed nationally scaled, mobile-accessible health credentials that were internationally recognizable without real-time database queries.³⁶ Denmark made vaccination and immunization records available in three languages on personal mobile devices for its entire population of 6 million people.³⁰ This system was operational at the population scale, thus demonstrating that technically and legally credible health data can be made globally portable within a compressed timeframe, when a policy mandate exists. In addition, the system established public familiarity with patient-held health credentials.³⁷ Though public trust challenges persisted in some countries,³⁸ and data standardization across countries was lacking,³⁹ most of the public was educated regarding the need to hold mobile-accessible health credentials.

While a vaccination certificate carries a small number of standardized data elements with a simple structure and an unambiguous temporal status, a GMRS summary requires extraction, reconciliation, prioritization, and translation of decades of heterogeneous clinical information. The vaccine credential experience does not demonstrate the feasibility, safety, or public acceptability of the GMRS but it does demonstrate that portable digital health credentials can be deployed and accepted at scale.

Formalization of the HL7 International Patient Summary

The International Patient Summary (IPS), published as ISO standard 27269 and implemented through Health Level Seven, Fast Healthcare Interoperability Resources (HL7 FHIR) and Clinical Document Architecture (CDA) implementation guides (IGs), defines a minimal dataset for unplanned, cross-border care.³² Part of the structure of IPS, including active problems, current medications, known allergies, immunization history, procedures, medical devices, and key results, maps directly onto the data elements proposed by Sørensen for the GMRS.³⁰ IPS is now formally endorsed by ISO, HL7 International, and the World Health Organization; moreover, it underlies the technical architecture of the European Health Data Space (EHDS), the most ambitious cross-border health data initiative attempted to date.^40,41

A critical policy implication is that the GMRS no longer needs to create its own transport and representation standard, because a globally governed scaffold now exists. What IPS provides and what the GMRS must supply are complementary: IPS provides the exchange container and terminology bindings, and the GMRS algorithm provides the abstraction logic that makes the container clinically useful. Many electronic health record (EHR) problem lists are over-inclusive and inconsistently maintained. For example, medication lists may include discontinued therapies without clear temporal status or information about non-adherence. In emergencies, such summaries can be counterproductive.⁴² The GMRS’s intended distinctive contribution is the signal-to-noise filter and temporal rules identifying which conditions are plausibly active, which therapies are current, and which historical entries should be suppressed. Demonstrating that a filtered summary performs better clinically than a well-implemented IPS is a task for the validation stages described below.

United States and European Interoperability Mandates

The 21st Century Cures Act, in 2016, and its implementing regulations, administered by the Office of the National Coordinator for Health Information Technology in the United States, established a federal mandate for health data interoperability with unique specificity and enforcement mechanisms.⁴³ Every certified EHR system must support data exchange by using FHIR-based APIs conforming to the United States Core Data for Interoperability (USCDI).

The USCDI data classes—edications, diagnoses, immunizations, laboratory results, allergies, and vital signs—substantially overlap with the GMRS data element inventory. Because USCDI conformance is implemented through the United States Core FHIR IG, which is itself designed to align with the international IPS profile, a dual mapping between GMRS elements and both USCDI and IPS is achievable in principle. This alignment is expressed at the level of profiles and value sets, but the equivalence at the level of individual instances has to be demonstrated case by case rather than assumed.^44,45 In the European Union, the EHDS Regulation creates a parallel obligation with a defined timetable.⁴⁶ The United States is therefore a natural first site for cross-national GMRS validation, and the Denmark–United States collaboration described herein is designed to leverage this opportunity.

Standards Pathway

We suggest that the GMRS be built as a computable abstraction layer operating over IPS-conformant data. A standalone GMRS would require every participating health system to use a new data standard and create new data sharing agreements. A GMRS implemented as an IPS-conformant abstraction layer would require only that health systems already meeting their IPS or USCDI obligations, as is increasingly mandatory in the European Union and the United States, activate the GMRS algorithm against data that are already required to be made available. This approach should reduce the marginal cost of participation and facilitate technical implementation, although the size of that reduction is an empirical question that will depend on local implementation quality.

Table 2 summarizes the complete standards stack, distinguishing what the specifications require from what is an implementation choice and noting the known limitations of each layer. The data elements proposed by Sørensen must be systematically mapped to the corresponding FHIR resource profiles and terminologies in the IPS specification.³⁰ These mappings must then be expressed as FHIR ConceptMap and StructureMap artifacts, which are computable, versioned, and published in an HL7 FHIR IG, so that the transformation logic is inspectable, testable, and replicable across institutions and need not be recreated by each implementer.^44,45

Table 2.

Standards Stack for GMRS Implementation, with Requirement Status and Known Limitations

Layer	Standard	Role in GMRS	Requirement Status and Known Limitations
Exchange structure	HL7 FHIR IPS (ISO 27269)	Document scaffold: section structure, resource profiles, terminology bindings for unplanned cross-border care	Normative for the structure of the summary. Section-level requirements are defined; which of a patient’s entries populate each section is left to the implementer
Exchange structure	USCDI and the United States Core FHIR IG	US federal minimum data standard; dual conformance with IPS is a stated design goal of US Core	Normative in the United States. Alignment with IPS is asserted at profile and value-set level and requires instance-level testing
Medications	ATC classification (WHO)	Language-neutral hierarchical drug coding used in the IPS Medication Summary section	One of several permitted code systems rather than a sole requirement. Encodes class; does not identify product, strength, or dose form
Medications	RxNorm (NLM)	US normalized drug nomenclature	Implementation choice in the US context. ATC to RxNorm mapping is one-to-many; combination, compounded, and nationally licensed products map poorly
Diagnoses	ICD-10 and ICD-10-CM	International disease classification in near-universal administrative use	Permitted where SNOMED CT is unavailable. National modifications are not fully interoperable; mapping between ICD-10 and ICD-10-CM is incomplete in both directions
Diagnoses	SNOMED CT	Higher clinical granularity; preferred for the IPS Problems section; mandated in several national implementations	Preferred rather than universal. Licensing and national extensions limit uniform availability
Laboratory results	LOINC (Regenstrief)	Universal codes for laboratory observations; used in the IPS Results section and in USCDI	Widely adopted. Local test codes with no assigned LOINC remain common in practice
Procedures	CPT, HCPCS, NOMESCO, and other national systems	National systems requiring explicit crosswalk in a Danish to US adaptation	No authoritative international crosswalk exists. Mapping is context dependent and needs case-by-case validation
Mapping infrastructure	FHIR ConceptMap and StructureMap	Computable, versioned terminology mappings and transformation logic; packaged in a FHIR IG	The mechanism is normative; the mappings themselves must be authored, versioned, and validated. Existing crosswalks reduce this work but do not remove it

For medication coding, a particular challenge relates to the use of the Anatomical Therapeutic Chemical (ATC) system in Danish and most European records, and RxNorm in United States EHRs. ATC encodes pharmacological class, whereas RxNorm differentiates ingredients, dose forms, and branded products. The two are not isomorphic: mapping from RxNorm to ATC discards dose form and strength; mapping from ATC to RxNorm is one-to-many and cannot recover a specific product without additional data; and combination products, nationally licensed products, compounded preparations, and biosimilars are recurrent failure points. Because a mistranslated medication entry can cause direct harm, we propose specific mitigations rather than reliance on crosswalk availability: a GMRS medication entry should carry the original national code alongside any mapped code, with its source terminology, version, and mapping provenance; uncertain or one-to-many mappings should be flagged and should display the original term rather than being silently normalized; dose, route, and frequency should travel as recorded; and safety-critical classes should be subject to a stricter error threshold and to human review before a summary is released. The crosswalks maintained by the National Library of Medicine, and more broadly the Observational Health Data Sciences and Informatics (OHDSI/OMOP) vocabularies and Unified Medical Language System (UMLS) crosswalks, provide the working translation infrastructure, and an OMOP node maintained by the OHDSI Denmark community provides a practical starting point for Danish medication and vaccine translation.⁴⁷

Legal review must accompany the technical work. The General Data Protection Regulation (GDPR) in the European Union, the Health Insurance Portability and Accountability Act (HIPAA) in the United States, and analogous instruments elsewhere impose requirements on cross-border health data sharing that a deployable GMRS must simultaneously satisfy, and the EHDS Regulation answers several of these questions for exchange within the European Union.^46,48 A patient’s right to obtain a copy of their own record, which exists in almost all jurisdictions, does not by itself establish a lawful basis for automated cross-border processing, for transformation of that record into a derived summary, for its storage, or for access by a clinician in another country. Consent and its withdrawal, residency, identity verification, emergency access, authentication and provenance, cybersecurity, and liability for decisions taken on an inaccurate summary are distinct questions requiring distinct answers. This review should therefore produce jurisdiction-specific data sharing frameworks as formal outputs of the project rather than treat them as a downstream consideration.

The Seven-Stage Maturation Pathway

Table 1 summarizes the proposed GMRS-on-IPS maturation pathway in full. The durations are estimates, offered so that the scale of the undertaking can be assessed, rather than commitments; they assume a funded core team and are contingent on data access approvals. The stages were derived by author synthesis, drawing on the sequence customarily used in HL7 implementation guide development and on established practice in validating EHR-based phenotyping algorithms. The GMRS as proposed in 2022 is a specification of intent rather than released software, so the validation described below would test a newly developed algorithm rather than an existing Danish deployment. The abstraction rules themselves, including temporal definitions, precedence rules for conflicting sources, and a versioning policy, are deliverables of stages 1 to 3 and would be published in inspectable form in the FHIR IG. The three validation stages are described in greater detail below, because they represent not only the primary scientific contribution of the work but also the area in which published evidence is sparse.

The GMRS algorithm’s distinctive function of filtering active from historical clinical status must be validated against external data before clinical deployment. The well-established methodological framework⁴⁹ requires defining a reference standard for an accurate GMRS summary constructed by physician review; applying the algorithm to generate candidate summaries from structured EHR or claims data; and computing sensitivity, specificity, positive predictive value, and negative predictive value, stratified by disease category, medication class, and data source type.

United States data sources include structured EHR data from academic medical centers or integrated delivery networks, administrative claims data, and data from health information exchanges (HIEs) aggregating multi-provider records. Adults 30 years or older with at least 12 months of continuous observation will constitute the study population; the age threshold is intended to enrich the sample for accumulated chronic disease while retaining working-age travelers, and the observation requirement is intended to ensure that the temporal rules have sufficient data to operate on.¹² Sampling should be stratified by number of chronic conditions and by medication count rather than drawn at random, because algorithm performance is expected to degrade as record complexity increases. The most robust validation will compare algorithm performance across multiple source types, because the GMRS must function acceptably regardless of the data architecture from which a patient’s record is drawn. The OMOP Common Data Model, successfully applied to cross-national EHR validation studies including Danish data, may provide a standardized representation layer that facilitates this comparison.^50,51

The validation design provides for independent review of each sampled record by at least two clinicians with relevant expertise, blinded to the algorithm output, with disagreements resolved by a third adjudicator, inter-rater agreement reported, and sample size determined by the precision required on the least common stratum. In addition, because predictive values depend on prevalence and the consequences of different errors can be asymmetric, acceptable thresholds will be prespecified separately for safety-critical elements and for lower-risk discrepancies. Safety-critical elements, where an omission or false entry could plausibly cause immediate harm, include anticoagulants, insulin and other hypoglycemic agents, opioids, immunosuppressants, antiepileptics, severe drug allergy, implanted cardiac devices, transplant status, and adrenal insufficiency; for these the acceptable false negative rate should approach zero and be stated before analysis.

Before application to real patient data, algorithm logic and terminology mappings should be tested against synthetic cohorts generated with any open-source software package that produces FHIR-formatted synthetic EHRs with realistic clinical trajectories.⁵² These synthetic cohorts would provide a controlled ground-truth environment for testing transformation logic and terminology translations, thus enabling systematic identification of mapping gaps before exposure to identifiable patient data. Studies have confirmed that synthetically generated data produce plausible distributions of chronic conditions and medications suitable for algorithm development and testing, while also showing that synthetic data do not reproduce the documentation gaps and inconsistencies of real records, which is why synthetic testing cannot substitute for empirical validation.⁵³

Clinical Utility Review

Technical conformance and clinical utility are different properties, and demonstrating the first does not demonstrate the second: a summary can validate cleanly against the IPS profile and still be incomplete, out of date, poorly translated, or unreadable in the two minutes available during a resuscitation. Clinical validation is therefore an essential step, in addition to technical testing, to ensure that the system performs in accordance with its purpose. A panel of clinicians and health informatics specialists from the United States and Denmark must review GMRS-derived summaries against full longitudinal records for a sample of patients and classify errors as omissions, commissions, or ambiguities, with particular attention to the safety-critical content listed above. Beyond accuracy, the evaluation should include usability and cognitive burden, time taken to retrieve a specific item, the effect on diagnostic and prescribing decisions in simulated encounters, patient understanding, and performance across languages and levels of health literacy. Outputs of this review would drive algorithm refinement, define minimum data requirements for clinically acceptable GMRS performance, and establish whether clinicians will trust and use the summary. Whether any of this translates into improved clinical outcomes is a question for later prospective study, and we make no claim about it here.

It is worth mentioning that algorithmic filtering carries its own risk. Suppression of a diagnosis, procedure, or medication can cause greater harm than an over-inclusive summary, and relevance is context dependent: a condition that appears inactive for routine outpatient care may be decisive in evaluating infection, in pregnancy, in trauma, or in selecting an anesthetic agent. Ideally, summaries should therefore be capable of being generated for a stated use case rather than only as a single universal artifact; elements retained or suppressed on uncertain evidence should carry a confidence indicator and their provenance rather than appearing as plain assertions; and a clinician should be able to reach the underlying record where one is accessible, with patients able to review and correct the summary generated about them.

Governance, Equity, and International Scalability

The GMRS initiative requires formal engagement with existing international governance bodies. HL7 International, ISO Technical Committee 215, and the EHDS technical committees maintain the standards on which the GMRS relies. Therefore, submitting GMRS mapping results for community review is both a scientific peer review process and an implementation prerequisite for use.^33,41 The EHDS, which entered into force in March 2025 and requires patient summaries and ePrescriptions to be exchanged across EU member states by 2029 in a common European format, is the most concrete near-term policy environment for GMRS deployment.⁴⁶ Similar patient protections and legal frameworks exist in other parts of the world that will require such engagement for oversight and governance.

Although the United States and Europe may lead this effort, equity considerations require architectural attention from the outset with a view to global use. The clinical benefit of a portable health summary is greatest in low- and middle-income travel destinations where local health information infrastructure might be less developed. There is also a real possibility that a standards-intensive system would widen disparities rather than narrow them, by working best for people whose care is delivered in highly digitized health systems and who own capable devices. The architecture should therefore accommodate people whose records are fragmented across unconnected providers, exist only on paper or in narrative text, or sit in health systems with no FHIR capability, and people without a smartphone or reliable connectivity, with low health literacy, or who do not read the languages in which the summary is rendered. The GMRS must be readable without specialized equipment and functional in low-bandwidth environments; a summary carried by the patient as a signed and encrypted machine-readable code with a human-readable printed counterpart would let a clinician working offline both read it and verify its origin, provided it carries a visible generation date and a stated validity period. A minimum viable summary should degrade gracefully when source data are thin rather than fail to generate. Whether these measures suffice is an empirical question, and equity-focused implementation research should be built into the pathway rather than added at the end.

Artificial intelligence may have a role here, but it remains unproven for this purpose. Large language models can extract structured clinical concepts from narrative text, which is precisely the gap that limits GMRS coverage in systems with poorly structured data, and they could in principle generate IPS-conformant content from unstructured records, propose candidate terminology mappings for human review, and produce translations for the destination language. They also fabricate content, perform unevenly across languages and populations, and produce outputs that are difficult to audit, so any such component would require the same two-tier error evaluation described above, with human verification for safety-critical content. We regard this as a defined line of investigation for a later phase rather than a necessary component of the present effort.

The sustainability of the GMRS will also require engagement with clinical professional organizations in travel medicine, emergency medicine, and primary care, to ensure that the generated summaries reflect the actual information needs of clinicians under time pressure. The most technically sophisticated summary without clinical relevance to acute encounters has little value. Patients themselves, and organizations representing them, should be involved from the design stage rather than consulted after specification, because decisions about what a patient-held summary discloses to a stranger, and about what a patient may withhold, cannot be settled by clinicians and standards bodies alone. In the ideal scenario, GMRS would be available to all, regardless of purchasing power, insurance status, or geography. A broad partnership of government agencies, technology companies, philanthropic foundations, and health payers must be established to ensure the future scalability and sustainability of GMRS internationally.

Improving the clinical quality of care for people who fall ill while traveling is thus of broad public interest. Better clinical decisions will lead to immediate improvements in treatment, resulting in fewer complications and a lower risk of long-term consequences. It will likely also be possible to avoid unnecessary and expensive diagnostic tests. All of this together could result in lower costs for patients, insurance companies, and health and social systems, although we are not aware of data that would allow the size of that effect to be estimated.

Limitations

This Commentary has several limitations. It is a policy argument and a proposed research agenda, and it presents no empirical results; the algorithm it describes has not been built in released form, and the seven stages were derived by author synthesis rather than by a formal consensus process. Terminology mappings between national coding systems will remain imperfect when clinical concepts are not the same across jurisdictions. Some information loss will be inevitable in the mapping process and must be quantified and disclosed. Real-world EHR data contain unstructured clinical text that may capture clinically relevant information not represented in structured fields. The current GMRS framework does not address unstructured data extraction, and this limitation is more pronounced in some health systems than others. The GMRS captures documented clinical status but cannot capture data never recorded; this limitation disproportionately applies to patients in health systems with lower documentation completeness.

Further risks concern the system rather than the algorithm. Cross-border patient identity matching has no agreed solution. A patient-held summary is a target for interception and forgery. A summary that is out of date may create false reassurance, which can be more dangerous than an acknowledged absence of information. Consent obtained once may not reflect what a patient would choose later, and liability for clinical decisions taken on an inaccurate summary is unallocated. Algorithms tuned on data from one health system may perform less well for populations under-represented in it, and translation errors may introduce clinical error invisible to both the sending and the receiving clinician. Finally, emergency clinicians may not trust or use a summary generated elsewhere.

Conclusion

In 2022, Sørensen raised awareness about the clinical need for a global medical record for traveling patients and called for efforts to start immediately.³⁰ Three years later, the standards environment is more favorable and several of the regulatory barriers are lower.

A missing component has been the framework in which recent developments have the potential to lead to an actionable agenda. At its core, a global medical record for travelers is a patient safety intervention, a public health imperative, and a policy choice. The seven-stage maturation pathway described herein provides a roadmap from conceptual alignment to deployment. The proposed work is designed to produce validated performance benchmarks, published FHIR IGs, and a replicable method that international partners can apply to their own data and health systems. What current standards create is an opportunity to build and test a GMRS-on-IPS approach under conditions that did not exist five years ago. Patients who collapse in a foreign city deserve clinicians who can know, within minutes, what conditions they carry, what medications they take, and what allergies they have. Reaching that point requires staged prototype development and prospective validation, with rigorous technical, clinical, legal, and implementation evaluation before international deployment.

Disclosure

The authors report no conflicts of interest in this work.

The Department of Clinical Epidemiology, Aarhus University, receives funding for other studies in the form of institutional research grants to (and administered by) Aarhus University. The Department of Clinical Epidemiology, Aarhus University, confirms that none of these studies have any relation to the present study. Henrik Toft Sørensen has made paid evaluations for University of Oslo, the Norwegian Research Council, the Independent Research Fund Denmark, the European Research Council, and EpidStrategies.

Anjum Khurshid is co-founder of a start-up, Pulsar Health, LLC, in the United States and received consulting fees from Aarhus University in 2025–2026.