ASF Standards · Primary Health Clinic · Standard 10

Standard 10 — Digital Care and Artificial Intelligence Systems for Care

7 criteria · 2 core · 5 standard-level · Version 1.0 · Aligned to ISQua EEA Principle 7, 6th Edition

Criteria in this standard

10.1 — Digital Systems Are Genuinely Evaluated Before Adoption
10.2 — Digital Care Never Disadvantages a Patient Who Cannot Use It
10.3 — Genuine Technical Support Is Available for Digital Systems
10.4 — AI Systems Are Introduced in Line With Law and Best-Practice Guidance
10.5 — AI Systems Are Genuinely Monitored for Unintended Consequences, Including Equity Impact
10.6 — Staff Are Genuinely Consulted Before an AI System Is Introduced
10.7 — Accountability for AI-Assisted Care Is Explicitly Defined
10.8 — Patients Are Genuinely Informed When Their Care Involves AI

10.1

Digital Systems Are Genuinely Evaluated Before Adoption

Standard

Before adopting any new digital system — patient records, vaccination tracking, appointment reminders — the clinic genuinely evaluates its cost, benefit, and compatibility with existing systems, not adopted on vendor assurance with no real internal evaluation.

In plain terms: Before the clinic adopts a new digital tool, someone has actually checked it’s worth it and will genuinely work.

Facility category	Crisis	Transition	Small	Standard
Applicability	Adapted	Full	Full	Full

Why this matters

Primary care clinics are increasingly offered digital vaccination tracking and reminder systems, often with genuine public-health benefit — but a real evaluation is what distinguishes a tool that actually fits this clinic’s workflow from one that creates data-entry burden without a matching real benefit.

What good looks like

A genuine cost/benefit evaluation happens before adoption.
Compatibility is genuinely checked.
Unintended consequences are genuinely considered.

Common failure modes

A system is adopted on vendor demonstration alone.

Worked example

In practice

A clinic considering a vaccination tracking system.

BeforeA vendor demo looked promising but no check was made of whether the system could genuinely sync with the national vaccine registry.

ActionA brief evaluation confirmed genuine registry compatibility before full adoption.

AfterThe Monitor reviewed the evaluation notes. Verified.

If you are starting from zero — do this first

Build a simple evaluation checklist before adoption.

The most common mistake: Skipping evaluation because a system promises genuine public-health benefit.

Self-assessment questions

1. Is a genuine cost/benefit evaluation conducted before adoption? — A real, documented evaluation.
Evidence: Evaluation checklist

2. Is compatibility genuinely checked? — Real, verified compatibility.
Evidence: Compatibility check record

3. Are unintended consequences genuinely considered? — Real, proactive consideration.
Evidence: Pre-launch risk review

Common reasons for a PARTIAL answer

Evaluation happens for major systems but is skipped for smaller add-ons.

Implementation plan

When	What
Week 1	Build an evaluation checklist.

How the Monitor verifies this

Method	What	Detail
DOCUMENT	Evaluation review	Reviews the evaluation record for a recently adopted system.

Supervisor tips

Ask about the most recently adopted digital tool.

Evidence base

World Health Organization. Global Strategy on Digital Health 2020-2025. Geneva: WHO; 2021.

ASF training courses on GMJ Academy →

Foundation courses A-00 to A-03 are live. Criterion-specific modules are being developed and will link here when published.

10.2

Digital Care Never Disadvantages a Patient Who Cannot Use It

Core

A genuine process ensures that patients who cannot use digital devices, lack internet access, or face language or literacy barriers are never disadvantaged in booking, reminders, or health communication — a genuinely heightened risk in primary care settings that often serve lower-income, migrant, or less digitally connected populations.

In plain terms: A patient who can’t use the clinic’s app or doesn’t speak the local language still gets exactly the same real access — a genuine alternative, not a quietly worse option.

Facility category	Crisis	Transition	Small	Standard
Applicability	Adapted	Full	Full	Full

Why this matters

This is marked Core because primary health clinics genuinely, frequently serve exactly the populations most vulnerable to digital exclusion — lower-income patients, migrants, those with limited literacy in the local language. A digital-only pathway here doesn’t create a minor inconvenience; it can genuinely cut off access to care for precisely the patients the clinic most exists to serve.

What good looks like

A genuine, equally functional alternative exists.
New services are genuinely pre-tested with the actual patient population, including language barriers.
Real evidence shows equivalent service.

Common failure modes

A digital reminder system assumes smartphone access and local-language literacy.

Worked example

In practice

A clinic that introduced SMS appointment reminders.

BeforeSMS reminders were sent only in the majority local language, leaving a genuine minority of patients without reliable reminders.

ActionReminders were added in the clinic’s second most common patient language, and a phone-call alternative was explicitly offered to any patient who preferred it.

AfterThe Monitor confirmed both the multilingual reminders and the phone alternative were genuinely functioning. Verified.

If you are starting from zero — do this first

Check your digital communication actually reaches your real patient population’s language needs.

The most common mistake: A digital communication system designed around the majority patient profile, quietly excluding a genuine minority.

Self-assessment questions

1. Is there a genuine, equally functional alternative? — A real, equivalent alternative.
Evidence: Documented alternative channel

2. Was a new service genuinely pre-tested with the actual patient population, including language barriers? — Real, prior testing.
Evidence: Pre-launch testing record

3. Is there real evidence the alternative delivers equivalent service? — A real, demonstrated instance.
Evidence: Direct test or comparison data

Common reasons for a PARTIAL answer

An alternative exists but isn’t genuinely promoted to patients who need it.

Implementation plan

When	What
Week 1	Audit digital communication for genuine language and access coverage.

How the Monitor verifies this

Method	What	Detail
ASK	Direct test	Tests the non-digital alternative directly.

Supervisor tips

Ask about the clinic’s actual patient population language mix.

Evidence base

World Health Organization. Ethics and Governance of Artificial Intelligence for Health. Geneva: WHO; 2021.

ASF training courses on GMJ Academy →

Foundation courses A-00 to A-03 are live. Criterion-specific modules are being developed and will link here when published.

10.3

Genuine Technical Support Is Available for Digital Systems

Standard

The clinic has genuine access to technical support — in-house, vendor-provided, or contracted — for its digital systems, with testing before full implementation, not systems left running with no real support behind them.

In plain terms: When a digital system breaks, there’s a real person to call.

Facility category	Crisis	Transition	Small	Standard
Applicability	Adapted	Full	Full	Full

Why this matters

Primary care clinics typically lack dedicated IT staff, making a genuine, accessible vendor support relationship particularly important — without it, a system failure can disrupt vaccination tracking or patient records with no real path to resolution.

What good looks like

Genuine, accessible support exists.
Testing genuinely occurs before implementation.
A real, known escalation path exists.

Common failure modes

Support knowledge rests with one staff member with no documented record.

Worked example

In practice

A clinic with no documented support contact.

BeforeSupport knowledge lived only with the staff member who originally set up the system.

ActionA support contact card was posted and documented for all staff.

AfterThe Monitor asked a staff member who correctly identified the support contact. Verified.

If you are starting from zero — do this first

Document and post support contacts for every digital system.

The most common mistake: Support knowledge living only with whoever originally set up the system.

Self-assessment questions

1. Does the clinic have genuine, accessible technical support? — Real, available support.
Evidence: Support contract

2. Were genuine testing and quality checks performed before implementation? — A real pre-launch check.
Evidence: Testing record

3. Is there a real, known route to support? — A genuine, known path.
Evidence: Staff interview

Common reasons for a PARTIAL answer

Support contact knowledge depends on one staff member.

Implementation plan

When	What
Week 1	Document support contacts for all systems.

How the Monitor verifies this

Method	What	Detail
ASK	Staff interview	Asks a staff member who they’d contact for an issue.

Supervisor tips

Ask a front-desk staff member, not the clinic director.

Evidence base

NHS Digital. Technology Assessment Criteria for Digital Health Products. London: NHS Digital; 2022.

ASF training courses on GMJ Academy →

Foundation courses A-00 to A-03 are live. Criterion-specific modules are being developed and will link here when published.

10.4

AI Systems Are Introduced in Line With Law and Best-Practice Guidance

Standard

Any artificial intelligence tool used to support screening, triage, or risk stratification is introduced in accordance with applicable law, or where none exists, with available best-practice guidance such as WHO’s AI ethics and governance guidance — not adopted with no reference to any external standard of responsible use.

In plain terms: Before using an AI screening tool, the clinic has actually checked what the law requires — not made it up as it went along.

Facility category	Crisis	Transition	Small	Standard
Applicability	Adapted	Full	Full	Full

Why this matters

AI screening tools are increasingly deployed in community and primary care settings for population-level risk stratification, and a real, grounded governance basis is what separates responsible adoption from an ungoverned experiment affecting real community health decisions.

What good looks like

AI systems are genuinely introduced in line with applicable law.
Where no law exists, introduction is genuinely informed by recognized guidance.
The clinic can identify the specific basis.

Common failure modes

No reference to law or guidance was made before adoption.

Worked example

In practice

A clinic using an AI-assisted diabetes risk screening tool.

BeforeThe tool had been adopted with no check of applicable regulation or any formal reference to external guidance.

ActionA governance review confirmed applicable regulation and formally adopted WHO’s AI guidance.

AfterThe Monitor reviewed the governance note. Verified.

If you are starting from zero — do this first

Check for applicable AI regulation and document a governance basis.

The most common mistake: Adopting an AI tool on clinical recommendation alone with no separate governance check.

Self-assessment questions

1. Is any AI system genuinely introduced in line with applicable law? — Real, verified compliance.
Evidence: Regulatory check record

2. Where no law exists, is introduction genuinely informed by recognized guidance? — A real, referenced source.
Evidence: Documented guidance reference

3. Can the clinic identify the specific basis for AI governance? — A real, specific answer.
Evidence: Staff interview

Common reasons for a PARTIAL answer

A guidance source is named but hasn’t actually been reviewed for its application.

Implementation plan

When	What
Week 1-2	Document a governance basis for each AI tool.

How the Monitor verifies this

Method	What	Detail
DOCUMENT	Governance note review	Reviews the documented governance basis.

Supervisor tips

Ask for the name of the specific guidance document.

Evidence base

World Health Organization. Ethics and Governance of Artificial Intelligence for Health: Guidance on Large Multi-Modal Models. Geneva: WHO; 2024.

ASF training courses on GMJ Academy →

Foundation courses A-00 to A-03 are live. Criterion-specific modules are being developed and will link here when published.

10.5

AI Systems Are Genuinely Monitored for Unintended Consequences, Including Equity Impact

Core

The clinic genuinely monitors any AI-assisted screening or risk tool for over-diagnosis, missed diagnosis, and specifically for differential accuracy across the clinic’s actual patient population — a genuine, documented risk where AI tools trained on unrepresentative data can perform less reliably for exactly the populations primary care most often serves.

In plain terms: The clinic actually keeps checking whether its AI screening tool works equally well for everyone it actually serves — not just assuming it’s accurate across the board.

Facility category	Crisis	Transition	Small	Standard
Applicability	Adapted	Full	Full	Full

Why this matters

This is marked Core because of a genuine, well-documented risk specific to this setting: AI diagnostic and screening tools are frequently trained on data that underrepresents the exact populations — lower-income, certain ethnic groups, older adults — that primary health clinics most often serve, and this can produce genuinely lower accuracy for exactly the patients the clinic exists to protect, without that gap ever being visible unless someone specifically checks for it.

What good looks like

AI output is genuinely audited.
Accuracy is genuinely checked for consistency across the real patient population.
A documented instance exists of monitoring catching a real issue.

Common failure modes

Accuracy is assumed uniform with no equity-specific check.

Worked example

In practice

A clinic using an AI-assisted skin condition screening tool.

BeforeThe tool’s accuracy had never been checked specifically across the clinic’s genuinely diverse patient population, despite known industry concerns about such tools’ accuracy on darker skin tones.

ActionA quarterly audit was introduced specifically comparing tool accuracy against independent clinical review across the clinic’s actual patient demographic range.

AfterThe Monitor reviewed the audit methodology confirming genuine, demographic-specific comparison. Verified.

If you are starting from zero — do this first

Establish an audit that genuinely checks accuracy across your actual patient population’s diversity.

The most common mistake: Auditing overall accuracy without ever checking whether it’s genuinely consistent across the clinic’s actual patient population.

Self-assessment questions

1. Is AI-assisted output genuinely audited? — Real, ongoing audit.
Evidence: Audit record

2. Is accuracy genuinely checked for consistency across the clinic’s population? — A real, specific equity check.
Evidence: Demographic-specific audit data

3. Is there a documented instance of monitoring identifying a genuine issue? — A real, concrete example.
Evidence: Issue response record

Common reasons for a PARTIAL answer

Overall accuracy is checked but not broken down by patient demographic.

Implementation plan

When	What
Week 1-2	Establish a demographic-aware audit process.

How the Monitor verifies this

Method	What	Detail
DOCUMENT	Audit review	Reviews the audit for genuine, demographic-specific comparison.

Supervisor tips

Ask specifically whether accuracy has ever been checked across patient demographics.

Evidence base

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.

ASF training courses on GMJ Academy →

Foundation courses A-00 to A-03 are live. Criterion-specific modules are being developed and will link here when published.

10.6

Staff Are Genuinely Consulted Before an AI System Is Introduced

Standard

Clinical staff who will actually use a new AI tool are genuinely consulted before its introduction, with real training needs identified and addressed — not a tool rolled out with no engagement beyond a brief notification.

In plain terms: Before a new AI tool goes live, the staff who will actually use it have had a real say.

Facility category	Crisis	Transition	Small	Standard
Applicability	Adapted	Full	Full	Full

Why this matters

Frontline clinicians, nurses, and community health workers in a primary care setting often have the clearest, most grounded sense of how a new AI screening tool will actually function in practice with their real patient population — genuine consultation surfaces this knowledge before it becomes a real problem.

What good looks like

Staff are genuinely consulted before introduction.
Consultation genuinely identifies training needs.
Staff can describe genuine consultation.

Common failure modes

Staff learn about a new tool only once the decision is final.

Worked example

In practice

A clinic introducing an AI-assisted screening tool.

BeforeNurses received an email announcing the tool with no prior input sought.

ActionThe launch was delayed to hold a genuine consultation session, surfacing a real workflow concern addressed before go-live.

AfterThe Monitor interviewed a nurse who described the consultation. Verified.

If you are starting from zero — do this first

Hold a consultation session before finalizing any AI system decision.

The most common mistake: Treating a notification email as equivalent to genuine consultation.

Self-assessment questions

1. Are staff genuinely consulted before introduction? — Real, prior consultation.
Evidence: Consultation record

2. Does consultation genuinely identify training needs, actually addressed? — Real needs matched by real training.
Evidence: Training delivery record

3. Can staff describe having been genuinely consulted? — Tests whether consultation actually registered.
Evidence: Staff interview

Common reasons for a PARTIAL answer

Consultation happened with one lead clinician but not the wider team.

Implementation plan

When	What
Before any launch	Hold genuine consultation with actual users.

How the Monitor verifies this

Method	What	Detail
ASK	Staff interview	Asks staff whether they felt genuinely consulted.

Supervisor tips

Ask for a specific example of a change made because of staff input.

Evidence base

Topol EJ. High-Performance Medicine: The Convergence of Human and Artificial Intelligence. Nature Medicine. 2019;25(1):44-56.

ASF training courses on GMJ Academy →

Foundation courses A-00 to A-03 are live. Criterion-specific modules are being developed and will link here when published.

10.7

Accountability for AI-Assisted Care Is Explicitly Defined

Standard

The clinic has genuinely considered and documented accountability for care decisions made with AI support — who is responsible when an AI-assisted decision is wrong — not leaving this as an unexamined question until it actually matters.

In plain terms: Everyone knows, in advance, who is actually responsible if an AI-supported decision turns out to be wrong.

Facility category	Crisis	Transition	Small	Standard
Applicability	Adapted	Full	Full	Full

Why this matters

Without clear, documented accountability, a safety incident involving AI-assisted screening risks becoming a confused dispute over responsibility exactly when clarity matters most — for the patient, the clinician, and the clinic’s own ability to learn from what happened.

What good looks like

Clinical accountability is explicitly documented.
Clinicians genuinely understand their own accountability.
A defined process exists for reviewing accountability.

Common failure modes

Accountability has never been explicitly addressed.

Worked example

In practice

A clinic using an AI-assisted risk stratification tool.

BeforeStaff were unclear whether responsibility for a missed high-risk flag sat with the nurse or the AI system.

ActionA clear policy was documented: the tool is decision support only, and the reviewing clinician retains full accountability.

AfterThe Monitor interviewed a nurse who could clearly state her own accountability. Verified.

If you are starting from zero — do this first

Explicitly document where clinical accountability sits for each AI tool.

The most common mistake: Leaving accountability ambiguous until an actual incident forces clarity.

Self-assessment questions

1. Is clinical accountability explicitly documented? — A real, clear answer.
Evidence: Documented accountability policy

2. Do clinicians genuinely understand their own accountability? — Real, demonstrated understanding.
Evidence: Staff interview

3. Is there a defined process for reviewing accountability in a safety incident? — A real, usable process.
Evidence: Incident review protocol

Common reasons for a PARTIAL answer

A policy exists but hasn’t been communicated to frontline staff.

Implementation plan

When	What
Week 1	Document accountability for each AI tool in use.

How the Monitor verifies this

Method	What	Detail
ASK	Staff interview	Asks a clinician to describe their accountability.

Supervisor tips

Ask a clinician directly: if the AI flagged something wrong, who is responsible?

Evidence base

European Commission. Ethics Guidelines for Trustworthy Artificial Intelligence. Brussels: European Commission; 2019.

ASF training courses on GMJ Academy →

Foundation courses A-00 to A-03 are live. Criterion-specific modules are being developed and will link here when published.

10.8

Patients Are Genuinely Informed When Their Care Involves AI

Standard

A patient whose care involves an aspect delivered with AI system support is genuinely informed of this — not left to assume every decision was made by a human clinician alone, with disclosure treated as optional rather than a genuine, standard part of informed care.

In plain terms: If AI is actually involved in part of a patient’s care, they’re genuinely told — not left to assume a human made every decision alone.

Facility category	Crisis	Transition	Small	Standard
Applicability	Adapted	Full	Full	Full

Why this matters

A patient’s genuine right to understand their own care includes knowing, in plain terms, when a decision affecting them was shaped by an AI system rather than a clinician’s judgement alone. This is especially relevant in a primary health clinic setting, where AI-assisted screening tools may genuinely flag risk factors a patient would want to know were AI-generated. Disclosure buried in a general consent form a patient never actually reads does not meet this bar.

What good looks like

Patients are genuinely informed when AI is involved in their care.
Disclosure is genuinely understandable, in plain language.
A patient can genuinely confirm they were told.

Common failure modes

AI disclosure is buried in a general consent form, technically present but never genuinely read or understood.

Worked example

In practice

A clinic using an AI-assisted screening tool for a chronic disease risk check.

BeforeAI’s role was mentioned only in a lengthy general consent document, which patients signed without genuine awareness that an AI tool had actually been involved in their specific screening.

ActionA brief, plain-language note was added to the results conversation: “An AI-assisted tool helped flag your risk level; your clinician reviewed it directly.”

AfterThe Monitor interviewed a patient who could genuinely confirm they understood AI had been involved. Verified.

If you are starting from zero — do this first

Add a brief, plain-language disclosure point at the moment results or care decisions involving AI are actually discussed with the patient.

The most common mistake: Treating a line buried in a general consent form as genuine disclosure, when the patient never actually registers it.

Self-assessment questions

1. Is a patient genuinely informed when AI is involved in their care? — Real, standard disclosure.
Evidence: Disclosure protocol

2. Is this disclosure genuinely understandable, not buried in technical language? — A real, plain-language explanation.
Evidence: Disclosure wording sample

3. Can a patient asked directly confirm they were genuinely told? — A real, concrete confirmation.
Evidence: Patient interview

Common reasons for a PARTIAL answer

Disclosure exists in a consent form but patients genuinely cannot recall or explain it when asked directly.

Implementation plan

When	What
Week 1-2	Draft plain-language disclosure wording for each AI-assisted care point and train staff to deliver it.

How the Monitor verifies this

Method	What	Detail
ASK	Patient interview	Asks a patient whether they were genuinely told AI was involved in their care.

Supervisor tips

Ask a patient directly rather than relying on the consent form’s existence alone.

Evidence base

World Health Organization. Ethics and Governance of Artificial Intelligence for Health: Guidance on Large Multi-Modal Models. Geneva: WHO; 2024.

ASF training courses on GMJ Academy →

Foundation courses A-00 to A-03 are live. Criterion-specific modules are being developed and will link here when published.

← PHC Standards — all standards
Self-assess against these criteria →
← Standard 9 — Sustainable Care
Standard 11 — Supporting the Care Workforce →