Trust
Answers for clinical safety, IG, security and accessibility teams.
Organised around the questions you have to answer internally. Every item names its scope, owner and evidence route.
Send this page to your clinical safety, IG, cyber and accessibility teams. It sets out our core assurance position and identifies the evidence route for supporting documents shared through the appropriate due-diligence process.
What we already hold
-
NHS DTAC Full evidence pack -
DCB0129 Clinical safety case -
UK GDPR DPIA support -
Cyber Essentials Current certification - Penetration tested CREST and NCSC CHECK provider How we assure security
-
NHS DSPT Toolkit submission -
G-Cloud 15 Direct NHS purchase View listing - Insurance Framework-level cover What we hold
-
WCAG 2.2 AA Tested release gate -
NHS Innovation Accelerator Alumni and mentor -
NHS Clinical Entrepreneur Programme Fellow and mentor -
SBRI Healthcare NHS England programme -
Texas Medical Center Supported innovator -
Mayo Clinic Innovation Exchange Supported innovator
Marks shown are the frameworks, certifications and programmes CardMedic holds, submits to or is supported by. Certificate numbers, scopes and current validity are supplied in the assurance pack. NHS DTAC is an assessment framework rather than a certification – we hold and share every piece of evidence it asks for.
Clinical content and translation governance
Healthcare organisations can usually see interpreting activity and spend, but often cannot see the complete communication pathway across content, accessible formats, live translation, interpreter escalation, delays, repeat contacts and outcomes.
That is not a criticism of interpreting teams or providers; it is a measurement gap created by fragmented workflows and contract boundaries. We help organisations define the pathway, measure appropriate use and improve access without removing the option of a professional interpreter, and we invite Trusts, providers and framework owners to contribute privacy-safe measures. Internal prompts, models, translation memory, thresholds and taxonomies stay private.
-
Clinically validated content
Authored and reviewed by clinicians, versioned, dated and owned. Translation is assessed against the approved clinical source through a controlled process combining linguistic expertise, approved terminology, technology-assisted analysis and targeted human and clinical review.
-
Live Translate
AI-assisted live translation inside a governed workflow, with visible limits and explicit escalation guidance for consent, safeguarding and breaking bad news.
-
Professional interpretation
Human language skill, cultural understanding and judgement. Required for high-risk conversations and available from the same workflow.
CardMedic provides clear signposting to professional interpreting where the conversation, local policy or clinical judgement makes that appropriate. CardMedic supports the decision; it does not make the clinical decision.
Communication assurance
How CardMedic assures clinical communication
CardMedic is a complete clinical communication system designed to help healthcare professionals communicate safely and effectively whenever communication is difficult. Curated clinical content, multilingual communication, Live Translate, Easy Read and Read Aloud, BSL and direct access to professional interpreters sit inside one governed platform.
It is live across NHS Trusts and hospital and community settings in the UK and the US, supported by relevant information-governance and NHS assurance controls including GDPR, DSPT and DTAC-aligned requirements, and evaluated through real-world NHS and academic studies.
- Clinician authored Every source is written or verified by clinicians before anything is translated.
- Clinically governed One framework covers content, translation, accessibility, Live Translate and interpreter access.
- Independently evaluated Assessed in real-world NHS use and in academic study, with methodology published.
- Continually improved Feedback is triaged by clinical consequence, corrected under version control and monitored.
Our Clinical Communication Governance Framework
Effective clinical communication extends beyond translation alone. The framework supports the quality, safety and continual improvement of communication across the whole platform – not only the words on a card.
It is applied proportionately to the communication method, its intended use, the clinical context and the potential consequence of an error. A pre-operative fasting question and a safeguarding conversation are not governed the same way, and should not be.
What the framework governs
A clinical baseline for every communication
Assurance begins with a clinically authored or clinically verified source. That source establishes the intended meaning, purpose, audience, context, risk and acceptable communication outcome before anything is translated or adapted.
Translation and accessible communication are then assessed against that clinical baseline – not solely against literal equivalence between two languages.
This is materially different from translating arbitrary text.
Our multilingual content starts from communication that has already been considered for clinical clarity, appropriate register and patient understanding. The starting point is the assurance.
A clinically anchored, risk-weighted assurance cycle
Not a human-versus-technology model. Clinical authorship, linguistic expertise, technology – including machine translation and automated checking – and targeted human and clinical review each perform a defined role, and the cycle does not stop at release.
- Clinical baseline A clinically authored or verified source sets intended meaning, purpose, audience, context, risk and acceptable outcome.
- Linguistic and technology-enabled checks How meaning, completeness and context have been carried across languages and formats is analysed consistently at library scale.
- Clinical and linguistic review Human expertise assesses suitability, register, cultural fit and potential clinical risk against the baseline.
- Structured sampling and re-assessment Ongoing sampling weighted to risk identifies areas for improvement rather than waiting for a complaint.
- Version control and release Changes are versioned, dated and owned by a named reviewer, so what is in use is always identifiable.
- Monitoring and corrective action Concerns and confirmed errors are triaged by clinical consequence, corrected, released and carried forward as learning.
What quality actually means
Quality is more than grammatical or literal accuracy. For CardMedic it means communication that is clinically appropriate for its intended use, assessed across seven dimensions.
- Clinical meaning The intended clinical message survives the change of language or format.
- Completeness Necessary information is present – nothing safety-relevant omitted or added.
- Terminology Medical terms are appropriate to the setting and the audience.
- Language variation Dialect, regional variation and register suit the person in front of the clinician.
- Cultural suitability Framing is appropriate to the patient’s context, not only literally correct.
- Patient understanding Clarity is judged by whether a patient can act on it, not by reading level alone.
- Potential consequence Findings are classified by the clinical harm an error could cause.
Being straight about accuracy
A universal 100% accuracy claim is not a meaningful or achievable measure for any communication provider – including us.
Communication is affected by clinical context, dialect, language variation, cultural framing, ambiguity, omissions, additions, interpretation and patient understanding. That is true of curated content, of live translation and of human interpreting alike.
The appropriate standard is to define intended use, establish the clinical baseline, assess the relevant quality dimensions, identify clinically significant issues, and ensure confirmed concerns are investigated, corrected and learned from. That is the standard we hold ourselves to, and the standard we would ask you to apply to every provider you assess.
One system, three communication pathways
CardMedic is not a standalone interpreting company or a single translation tool. Staff select the most appropriate pathway for the patient and the situation, inside one clinically governed system.
- Curated clinical content Clear, consistent and immediately available communication, authored and reviewed before it ever reaches a bedside. Best for: planned, recurring, complex or safety-sensitive scenarios
- Live Translate Dynamic two-way communication generated during an interaction, inside a governed workflow with its limits visible. Best for: unplanned conversations where no content exists yet
- Professional interpreters Direct access to your approved provider, one tap from the clinical workflow rather than a separate process, wherever the organisation has enabled connection. Best for: live, open-ended or individualised conversations, consent, safeguarding and breaking bad news
A shared basis for communication assurance
Trusts should be able to assess every communication method with the same questions – curated content, professional interpreting, digital translation or a blended model. These are the questions we answer about ourselves, and we would encourage you to put them to anyone bidding for this work.
- Intended use and risk
- What is the intended use of the communication method?
- Which patients, interactions and clinical settings is it designed to support?
- What risks does it address?
- What uses fall outside its intended scope?
- Clinical quality and accuracy
- What is the approved clinical baseline or reference standard?
- How are meaning, completeness, terminology, language variation, cultural suitability and patient understanding assessed?
- What constitutes a clinically significant error?
- Are findings classified by potential clinical consequence?
- How are complex, sensitive or safety-critical communications handled?
- Evidence and measurement
- What evidence supports the reported results?
- Which languages, content, settings and versions were assessed?
- What was the sample and the denominator?
- Is quality assessed before release, in real-world use, or both?
- How are staff and patient experience incorporated?
- Governance and improvement
- Who is accountable for clinical and linguistic quality?
- How are concerns, near misses and confirmed errors recorded and escalated?
- How quickly can a confirmed issue be corrected?
- How is learning applied to future content and interactions?
- How is professional interpretation provided when live, open-ended communication is required?
These questions assess the actual quality, safety, transparency and sustainability of a communication service – not simply whether a human or a technology was involved at one stage.
And how we will conduct ourselves
Asking an organisation to assess every provider by the same standard only carries weight if we are willing to say how we intend to behave.
Quality at NHS scale
The NHS needs communication support that is clinically governed, operationally usable and financially sustainable. CardMedic combines clinical authorship, linguistic expertise, technology-enabled assurance, structured governance and professional interpreter access in a repeatable system that deploys across services, languages and patient needs.
The NHS should not have to choose between quality and scale.
Findings from evaluated services
Every figure below comes from a service that has been evaluated. Each is drawn from an independent evaluation or case study of a single deployment, and measured against that service’s own users and activity. The first two figures show the scale those findings rest on.
Closely examined samples, assessed by people who do not work for us. That is what makes them worth reading.
- 30,000+ patient communication interactions in the services evaluated
- 5,000+ clinical staff using CardMedic in those services
- 100% / 96% of surveyed urgent-care staff reported improved / faster communication
- 50%+ reduction in blended cost per interaction while those measures were maintained
- p<0.001 improvement in patient satisfaction in a US doctoral study – a result this strong would occur by chance less than once in a thousand times
- 88 / 86 / 81% staff awareness, conversion to registered users and rated useful in a regional maternity, neonatal and community evaluation
Sample, denominator and setting for every figure are set out on the Evidence page and in the individual case studies. Evaluated settings span urgent and emergency care, maternity and neonatal services, inpatient and outpatient care, district and school nursing and community services.
Continuous improvement
Feedback from healthcare professionals, patients, partner organisations, language specialists and accessibility experts is recorded and assessed through structured governance. Issues are triaged by nature and potential clinical consequence; corrections are version-controlled, released and monitored.
The V5 platform and expanded clinical phrasebook continue this process, incorporating learning from clinical use, evaluation and colleague feedback.
Clinical use and scope
CardMedic supports – but does not replace – professional judgement or local clinical policy. Professional interpreter access is one of its pathways, alongside curated content, Live Translate and accessibility tools.
Healthcare professionals should select the method most appropriate for the patient and situation, confirm understanding where necessary, and use professional interpretation for live, open-ended or individualised communication when required by law, local policy or clinical judgement.
Accessible Information Standard
The standard is your duty. We cover part of it, and we will tell you which part.
The Accessible Information Standard applies to NHS and publicly funded adult social care organisations in England. Its reference changed from DCB1605 to DAPB1605 when the updated version was published in July 2025, and it now has six stages rather than five: identify, record, flag, share, meet and review. Organisations are expected to assess how well they meet it using NHS England’s self-assessment tool, with first assessments due by March 2027.
No software makes an organisation compliant, and we will not tell you otherwise. Compliance is a set of processes you own. What we can do is the “meet” stage, for the formats most services struggle to produce.
Where we help
- Meet. Easy Read, Read Aloud, British and American Sign Language, large-print-friendly display and 45 languages, available at the point of care rather than ordered in advance.
- Review. Our reporting shows which formats and languages were actually used, which is evidence a recorded need was met in practice and not just recorded.
- Self-assessment. That usage data is the kind of evidence the assessment asks for. Healthwatch found that most trusts describing themselves as compliant could not evidence a formal audit.
Where we do not
- Identify, record, flag and share. These happen in your patient record, not in our app. Your records system owns them.
- Certifying you. Nobody can sell you compliance. Any supplier claiming their product makes you AIS-compliant is overstating what software can do.
- Replacing professional interpreting or BSL interpreters. A registered sign language interpreter is often what the standard requires, and we route to yours.
The standard is not assured by the CQC, which is a commissioning responsibility – but the CQC does consider how well organisations meet it as part of assessment. Our own product accessibility is covered in our accessibility statement.
Insurance and accountability
CardMedic maintains professional indemnity, public and products liability, and cyber and data insurance appropriate to its business activities and use in healthcare environments. Card Medic Ltd is a named insured subsidiary under these arrangements.
CardMedic is an approved supplier on G-Cloud 15 and meets the framework’s applicable insurance and supplier-assurance requirements.
Our role and accountability
CardMedic is a communication-support tool, not a diagnostic or treatment system. It supports communication between healthcare professionals and patients.
CardMedic remains accountable for the content it authors and maintains. Curated content is subject to controlled translation and assurance, version control, monitoring and corrective action.
Clinical judgement, including the decision about which communication method is appropriate for a particular patient, remains with the treating clinician and their organisation.
Artificial intelligence
How we use AI, and where we do not
Written for assurance, procurement and clinical safety teams. This is what we would say in a tender response, and we will stand behind it in a contract.
-
Curated clinical content is clinically anchored, and assured before release
The source is written or verified by a clinician, and that source sets the meaning every translation is measured against. Translation into the curated library is prepared and quality-assured before release, through a controlled process combining linguistic expertise, approved terminology resources, technology-assisted analysis – including machine translation – and targeted human and clinical review. Versions are controlled, use is monitored, and confirmed issues are corrected and re-released.
Scrutiny is weighted to the potential clinical consequence of an error rather than applied evenly to every string, which is what makes the model both rigorous and scalable. All current curated content has passed through our defined translation and quality-assurance process. The method, its scope and its limits are set out in full in our translation safety and quality policy, which we share with customers and assurance teams on request.
-
Live Translate is machine translation, and is described as such
It handles open, unplanned conversation at the bedside. It is not a professional interpreting service and we do not call it one. Its limits are stated in the product and in our terms, not buried in a footnote.
-
No AI makes or influences a clinical decision
CardMedic does not diagnose, triage or recommend treatment. It helps a clinician and a patient understand each other. The judgement stays with the clinician.
-
A human route is never disincentivised, and never hidden
What you pay us does not change with the route. The licence is fixed, not per minute, so a conversation going to a human costs you no more than one that does not, and nothing in your bill pushes a conversation towards the machine.
Where an organisation has enabled interpreter connection, escalation is one tap from any pathway. Where it has not, CardMedic signposts professional interpreting and points to the organisation’s own booking route – see below. We would rather state that than imply a connection we do not have.
Two rules sit behind that, and they hold however our commercial model develops: how we are paid is set out in full in the contract, and our organisational licence is fixed and independent of interpreter minutes. Any referral, reseller or partner arrangement is transparent, separately agreed and subordinate to the organisation’s patient-appropriate route and governance. Those are the questions worth putting to any supplier in this market, including us. A price list tells you about today. How a supplier is paid tells you what they will do when nobody is looking.
-
When the interpreter connection is not in place
Enabling it is your organisation’s decision, and it is configured against the interpreting contract you already hold. The set-up itself is routine and we coordinate it. Where a connection is not yet live, that is usually a question of timing or local commercial arrangements rather than anything technical.
If it is not connected, nothing quietly degrades. Content and accessible formats work as normal. The workflow still tells staff when a conversation needs an interpreter and sends them to your existing booking process instead. We will tell you which routes are live in your set-up at implementation – and in the bid, not after go-live. If enabling it raises a contractual question, we have set out our answer in full.
-
Where we say it should not be used
Consent, safeguarding, mental capacity, breaking bad news, complaints and any complex individual discussion should use a professional interpreter. We say this in commercial conversations as readily as in governance documents, and you can configure the routing to enforce it.
-
No claim of 100% accuracy – from us or from anyone
We do not claim 100% accuracy, because no communication method achieves it. Meaning can shift with dialect and regional variation, gender and register, cultural framing, idiom, and the clinical situation the conversation happens in – and that is as true of a live human interpretation as it is of prepared content. So we do not publish a single accuracy percentage, and we would treat any supplier who does with caution. Ask what the number measures, across which content and languages, on what sample, and what was counted as an error. Ask us the same.
Data handling for each pathway, including what is processed, where and for how long, is set out in our privacy notice and in the terms covering Live Translate and use of artificial intelligence.