Standards › ASF Health Standards V11 › Standard 5
Standard 5 (Digital Care and AI), PDF
ASF-HS11-S05 · 1 November 2026 · 30 pages, A4. The same criteria as this page, for printing and assessors.
Where this standard applies
Standard 5 is a core standard. It applies to every facility type. The table shows each criterion by facility type.
| Criterion | Hospital | Ambulatory clinic | Primary health clinic | Long-term care | Telemedicine | Home care | Fitness and wellness | Dental clinic | Aesthetic and procedure clinic |
|---|---|---|---|---|---|---|---|---|---|
| 5.1 Digital Care Systems Follow an Assessment and Management Process | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| 5.2 Digital Care Never Disadvantages a Patient Who Cannot Use It | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| 5.3 Technical Expertise Supports Digital Care Systems | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| 5.4 AI Systems Are Introduced in Line With WHO AI Governance Guidance | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| 5.5 AI Systems Are Monitored for Unintended Consequences | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| 5.6 Staff Are Consulted Before an AI System Is Introduced | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| 5.7 Accountability for AI-Assisted Care Is Explicitly Defined | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| 5.8 Patients Are Informed When Their Care Involves AI | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
Digital Care and AI
A core standard. It covers how digital care systems are assessed and managed, the alternative route for patients who cannot use them, technical support, and the use of artificial intelligence in care: governance, monitoring, staff consultation, accountability and informing patients.
Each criterion is shown in three settings: Humanitarian settings, Transition and fragile, and Standard. In standard settings, every criterion is Must.
Training colours: red means essential, green recommended, yellow optional. Each course shows its level for each setting: humanitarian, transition and fragile, and standard.
5.1
Digital Care Systems Follow an Assessment and Management Process
Humanitarian settings: MinimumTransition and fragile: MediumStandard: Must
Before any digital care system — telehealth platform, remote monitoring tool, patient-facing app — is introduced, the facility assesses its costs and benefits, manages its implementation, and continues to manage it afterwards — not adopted because a vendor made a compelling pitch, with no internal evaluation.
Rating: Minimum in humanitarian settings; Medium in transition and fragile settings; Must in standard settings.
Why it matters. A digital tool introduced into care without a check can slow care down, expose data, or fail when most needed. Assessing a tool before it is used, and managing it afterwards, protects the patient and the service.
How the surveyor verifies this criterion
Sample. Review the assessment for each digital care system in use.
1. A cost/benefit analysis is done before implementation.
Method. Check the assessment record dated before the go-live date.
Evidence to show. DIG-FRM-002 Digital Tool Assessment Template; DIG-SOP-002 Digital Care System Assessment and Management.
Pass when. The analysis is dated before go-live.
2. Interoperability is assessed.
Method. Check that the assessment covers integration with existing clinical systems.
Evidence to show. DIG-FRM-002 Digital Tool Assessment Template.
Pass when. Interoperability and its risks are recorded.
3. Unintended consequences are considered in advance.
Method. Check the record for an identified risk list and its mitigations.
Evidence to show. DIG-FRM-002 Digital Tool Assessment Template.
Pass when. The risk list is complete and each risk has a mitigation.
Immediate fail (any one means the criterion is not met)
- A digital care system is introduced with no cost/benefit analysis and no interoperability check.
Staff, training and documents
Each occupation carries its ISCO-08 code. Each row is one document: read across to see who does each step. Training evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or your own training certificate kept in the staff file.
Good practice. Each digital care system has a written assessment dated before go-live. It covers costs, benefits and interoperability, and lists the risks with a mitigation for each. The system continues to be managed after it is introduced.
Partial because. An assessment exists, but it was completed after go-live, or interoperability is not recorded, or some risks have no mitigation.
Staff responsible
| Role | Why it applies | Level | Training | Documents |
|---|---|---|---|---|
| Manager ISCO 1342 ASF proposal |
Runs the facility process this criterion measures | Humanitarian: Essential Transition: Essential Standard: Essential |
Technology Platform Reliability & Security (TM-01) Evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or a training certificate in the staff file |
DIG-SOP-002 DIG-FRM-002 |
| Secretary ISCO 3344 ASF proposal |
Front desk and records support | Humanitarian: Recommended Transition: Recommended Standard: Recommended |
Technology Platform Reliability & Security (TM-01) Evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or a training certificate in the staff file |
DIG-SOP-002 DIG-FRM-002 |
E-learning: who must take each course.
| E-learning course | Who must take | Who is recommended |
|---|---|---|
| Technology Platform Reliability & Security (TM-01) GMJ Academy: English · Georgian |
Manager (ISCO 1342) | Secretary (ISCO 3344) |
Documents: who does each step
| Document | Writes | Approves | Fills in | Checks and signs | Files and uploads |
|---|---|---|---|---|---|
| DIG-SOP-002 · Digital Care System Assessment and Management Procedure (SOP) · To make sure every digital care system (telehealth, remote monitoring, patient-facing app, clinical software) is assessed, implemented and managed through a defined process, not adopted on a vendor pitch. |
Manager (ISCO 1342) | Chief executive (ISCO 1120) | Manager (ISCO 1342) | Chief executive (ISCO 1120) | Secretary (ISCO 3344) |
| DIG-FRM-002 · Digital Tool Assessment Template Form (record) · Assessment of a digital care system before introduction: costs, benefits, interoperability and unintended consequences. |
Manager (ISCO 1342) | Chief executive (ISCO 1120) | Secretary (ISCO 3344) | Manager (ISCO 1342) | Secretary (ISCO 3344) |
References
- World Health Organization (2021). Recommendations on digital interventions for health system strengthening. Used for: Digital systems assessment. https://news.gmj.ge/guideline/recommendations-on-digital-interventions-for-health-system-strengthening/
5.2
Digital Care Never Disadvantages a Patient Who Cannot Use It
Humanitarian settings: MinimumTransition and fragile: MediumStandard: Must
A process ensures that patients who cannot use digital devices, lack internet access, or are otherwise unable to engage with digital approaches are never disadvantaged by the facility's use of digital care — not digital convenience for most patients purchased at the cost of access for the patients least able to adapt.
Rating: Minimum in humanitarian settings; Medium in transition and fragile settings; Must in standard settings.
Why it matters. Some patients cannot use a phone app, do not have internet, or find video calls hard. A service that moves everything online without an alternative leaves those patients without care.
How the surveyor verifies this criterion
Sample. Select the digital care services in use and the records of patients who used the alternative route.
1. A defined alternative exists.
Method. Check the procedure for the alternative route.
Evidence to show. DIG-SOP-003 Digital Care Access and Alternatives.
Pass when. The alternative is defined and available.
2. The service was pre-tested with people who might struggle.
Method. Check the test record for a group of potential users.
Evidence to show. DIG-FRM-003 Digital Alternative Record.
Pass when. A pre-test with those users is recorded.
3. Patients using the alternative receive equivalent care.
Method. Select records of patients who used the alternative and compare the service they received.
Evidence to show. DIG-FRM-003 Digital Alternative Record.
Pass when. The service received is equivalent.
Immediate fail (any one means the criterion is not met)
- A patient who cannot use digital care has no alternative route, and receives reduced service.
Staff, training and documents
Each occupation carries its ISCO-08 code. Each row is one document: read across to see who does each step. Training evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or your own training certificate kept in the staff file.
Good practice. Every digital service has a defined alternative route that gives equivalent care. The service was tested beforehand with people who might struggle to use it, and each use of the alternative is recorded.
Partial because. An alternative route exists, but the service was not pre-tested with the people who might struggle, or the records do not show that care through the alternative is equivalent.
Staff responsible
| Role | Why it applies | Level | Training | Documents |
|---|---|---|---|---|
| Manager ISCO 1342 ASF proposal |
Runs the facility process this criterion measures | Humanitarian: Essential Transition: Essential Standard: Essential |
Evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or a training certificate in the staff file | DIG-SOP-003 DIG-FRM-003 |
| Secretary ISCO 3344 ASF proposal |
Front desk and records support | Humanitarian: Essential Transition: Essential Standard: Essential |
Evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or a training certificate in the staff file | DIG-SOP-003 DIG-FRM-003 |
E-learning: who must take each course.
No course is required for this criterion. Optional course in the GMJ Academy: Digital Inclusion for Patients Who Cannot Use Digital Tools (English · Georgian).
Documents: who does each step
| Document | Writes | Approves | Fills in | Checks and signs | Files and uploads |
|---|---|---|---|---|---|
| DIG-SOP-003 · Digital Care Access and Alternatives Procedure (SOP) · To make sure patients who cannot or prefer not to use digital care are never disadvantaged by it, and receive the same service through another route. |
Manager (ISCO 1342) | Chief executive (ISCO 1120) | Manager (ISCO 1342) | Chief executive (ISCO 1120) | Secretary (ISCO 3344) |
| DIG-FRM-003 · Digital Alternative Record Form (record) · Records the patient’s preferred route and any alternative route used for a digital service. |
Manager (ISCO 1342) | Chief executive (ISCO 1120) | Secretary (ISCO 3344) | Manager (ISCO 1342) | Secretary (ISCO 3344) |
References
- World Health Organization (2024). Implementation toolkit for accessible telehealth services. Used for: Accessible digital care and alternatives. https://news.gmj.ge/guideline/implementation-toolkit-for-accessible-telehealth-services/
5.3
Technical Expertise Supports Digital Care Systems
Humanitarian settings: MinimumTransition and fragile: MediumStandard: Must
The facility has access to technical expertise — in-house, vendor-provided, or contracted — to support the effective use of its digital care systems, including testing and quality checks before full implementation, not digital systems left running with no technical support behind them.
Rating: Minimum in humanitarian settings; Medium in transition and fragile settings; Must in standard settings.
Why it matters. A digital system that no one can support when it fails leaves staff stuck and patients unseen. Access to technical expertise, in-house or from a vendor, keeps the system working when it matters.
How the surveyor verifies this criterion
Sample. Review the testing records and the support log for each digital care system.
1. Technical expertise is available.
Method. Check the support arrangement, in-house or contracted.
Evidence to show. DIG-PLN-001 Technical Support and Testing Arrangement.
Pass when. A named route to expertise exists.
2. Testing is done before full implementation.
Method. Check the test record dated before full use.
Evidence to show. DIG-FRM-004 Digital System Test and Support Record.
Pass when. Testing is recorded before full use.
3. Staff know the support route.
Method. Ask staff how they would get technical help for a digital system.
Evidence to show. DIG-PLN-001 .
Pass when. Staff describe the route correctly.
Immediate fail (any one means the criterion is not met)
- A digital care system has no technical support route and was not tested before full use.
Staff, training and documents
Each occupation carries its ISCO-08 code. Each row is one document: read across to see who does each step. Training evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or your own training certificate kept in the staff file.
Good practice. Each digital care system has a named route to technical expertise and was tested before full use. Staff can say how to get support when the system fails.
Partial because. A support route exists, but testing before full use is not recorded, or staff cannot describe the route.
Staff responsible
| Role | Why it applies | Level | Training | Documents |
|---|---|---|---|---|
| Manager ISCO 1342 ASF proposal |
Runs the facility process this criterion measures | Humanitarian: Essential Transition: Essential Standard: Essential |
Evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or a training certificate in the staff file | DIG-FRM-004 DIG-PLN-001 |
E-learning: who must take each course.
No course is required for this criterion. Optional course in the GMJ Academy: Technical Support for Digital Care Systems (English).
Documents: who does each step
| Document | Writes | Approves | Fills in | Checks and signs | Files and uploads |
|---|---|---|---|---|---|
| DIG-FRM-004 · Digital System Test and Support Record Form (record) · Records the test before full use, after each change, and the support contact for each digital system. |
Manager (ISCO 1342) | Chief executive (ISCO 1120) | Secretary (ISCO 3344) | Manager (ISCO 1342) | Secretary (ISCO 3344) |
| DIG-PLN-001 · Technical Support and Testing Arrangement Plan · To make sure each digital care system has accessible technical expertise and is tested before and after full use. |
Chief executive (ISCO 1120) | Governing body members | Manager (ISCO 1342) | Chief executive (ISCO 1120) | Secretary (ISCO 3344) |
References
- World Health Organization (2022). Consolidated telemedicine implementation guide. Used for: Technical support and testing. https://news.gmj.ge/guideline/consolidated-telemedicine-implementation-guide/
5.4
AI Systems Are Introduced in Line With WHO AI Governance Guidance
Humanitarian settings: MinimumTransition and fragile: MediumStandard: Must
Any artificial intelligence system used to support diagnosis, triage, or care decisions is introduced and managed in accordance with WHO's Ethics and governance of artificial intelligence for health (2021). No such system is adopted without reference to this standard of responsible use.
Rating: Minimum in humanitarian settings; Medium in transition and fragile settings; Must in standard settings.
Why it matters. Artificial intelligence can help with diagnosis and triage, but it can also make errors that look confident. WHO's guidance on AI in health asks for safety, transparency and human oversight, and a facility that introduces AI must follow it.
How the surveyor verifies this criterion
Sample. Review the assessment and policy for each AI system in clinical use.
1. An assessment is made before introduction.
Method. Check the assessment against WHO guidance on the ethics and governance of AI for health.
Evidence to show. AI-FRM-001 AI System Assessment.
Pass when. Each system has an assessment that cites the guidance.
2. A responsible-use policy is in force.
Method. Check that the policy is approved and covers each system in use.
Evidence to show. AI-POL-001 Responsible Use of AI in Care Policy.
Pass when. The policy is approved and covers each system.
3. The facility can name its governance basis.
Method. Ask the responsible person for the governance basis of one named system.
Evidence to show. AI-FRM-001 AI System Assessment.
Pass when. The basis is named and recorded.
Immediate fail (any one means the criterion is not met)
- An AI system is in clinical use with no assessment and no governance basis.
Staff, training and documents
Each occupation carries its ISCO-08 code. Each row is one document: read across to see who does each step. Training evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or your own training certificate kept in the staff file.
Good practice. Each AI system has an assessment made before introduction that cites the WHO guidance. An approved responsible-use policy covers every system, and the facility can name its governance basis.
Partial because. A policy exists, but it does not cover every AI system in use, or an assessment does not cite the guidance.
Staff responsible
| Role | Why it applies | Level | Training | Documents |
|---|---|---|---|---|
| Chief executive ISCO 1120 ASF proposal |
Carries accountability to the governing body | Humanitarian: Essential Transition: Essential Standard: Essential |
Artificial Intelligence in Healthcare Safety (AI-001) Evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or a training certificate in the staff file |
AI-FRM-001 AI-POL-001 |
| Manager ISCO 1342 ASF proposal |
Runs the facility process this criterion measures | Humanitarian: Essential Transition: Essential Standard: Essential |
Artificial Intelligence in Healthcare Safety (AI-001) Artificial Intelligence Systems for Care (TELE-10, Georgian version only) Evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or a training certificate in the staff file |
AI-FRM-001 |
| Nurse ISCO 2221 ASF proposal |
Uses or checks the equipment and falls controls at the bedside | Humanitarian: Recommended Transition: Recommended Standard: Recommended |
Artificial Intelligence Systems for Care (TELE-10, Georgian version only) Evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or a training certificate in the staff file |
— |
E-learning: who must take each course.
| E-learning course | Who must take | Who is recommended |
|---|---|---|
| Artificial Intelligence in Healthcare Safety (AI-001) GMJ Academy: English · Georgian |
Chief executive (ISCO 1120), Manager (ISCO 1342) | None |
| Artificial Intelligence Systems for Care (TELE-10, Georgian version only) GMJ Academy: Georgian |
Manager (ISCO 1342) | Nurse (ISCO 2221) |
Documents: who does each step
| Document | Writes | Approves | Fills in | Checks and signs | Files and uploads |
|---|---|---|---|---|---|
| AI-FRM-001 · AI System Assessment Form (record) · Assessment of each AI system against WHO principles, with the human oversight rule. |
Manager (ISCO 1342) | Chief executive (ISCO 1120) | Secretary (ISCO 3344) | Manager (ISCO 1342) | Secretary (ISCO 3344) |
| AI-POL-001 · Responsible Use of AI in Care Policy Policy · To make sure AI systems used in diagnosis, triage or care decisions are introduced and governed in line with WHO guidance on the ethics and governance of AI for health. |
Chief executive (ISCO 1120) | Governing body members | N/A | Chief executive (ISCO 1120) | Secretary (ISCO 3344) |
References
- World Health Organization (2021). Ethics and governance of artificial intelligence for health. Used for: WHO guidance on AI governance. https://news.gmj.ge/guideline/ethics-and-governance-of-artificial-intelligence-for-health/
5.5
AI Systems Are Monitored for Unintended Consequences
Humanitarian settings: MinimumTransition and fragile: MediumStandard: Must
The facility monitors and evaluates its AI systems — auditing results for over-diagnosis or missed diagnosis, tracking patient complaints related to AI-assisted care, and checking for other unintended consequences — not deploying an AI tool and assuming it works as intended with no ongoing scrutiny.
Rating: Minimum in humanitarian settings; Medium in transition and fragile settings; Must in standard settings.
Why it matters. An AI system can work well at first and then drift, missing diagnoses or over-diagnosing as the patients change. Auditing its results regularly is the only way to notice.
How the surveyor verifies this criterion
Sample. Review the monitoring register for each AI system in use over the last 12 months.
1. AI output is audited for over- and under-diagnosis.
Method. Check the audit record for a sample of AI-assisted decisions.
Evidence to show. AI-SOP-001 AI Performance and Safety Monitoring.
Pass when. Audits are recorded for each system.
2. AI-related complaints are tracked.
Method. Check the complaints log for AI-related entries and their review.
Evidence to show. AI-REG-001 AI Monitoring Register.
Pass when. AI-related complaints are logged and reviewed.
3. A monitoring finding led to a response.
Method. Ask for one recorded case where monitoring found a problem and check the response.
Evidence to show. AI-REG-001 AI Monitoring Register.
Pass when. A recorded case shows a finding and a response.
Immediate fail (any one means the criterion is not met)
- An AI system's output is not checked, and a known performance problem has no recorded response.
Staff, training and documents
Each occupation carries its ISCO-08 code. Each row is one document: read across to see who does each step. Training evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or your own training certificate kept in the staff file.
Good practice. Each AI system is audited for over-diagnosis and missed diagnosis. Complaints related to AI are logged and reviewed, and at least one recorded finding shows the response that followed.
Partial because. Audits are recorded for some AI systems only, or AI-related complaints are not logged, or findings have no recorded response.
Staff responsible
| Role | Why it applies | Level | Training | Documents |
|---|---|---|---|---|
| Manager ISCO 1342 ASF proposal |
Runs the facility process this criterion measures | Humanitarian: Essential Transition: Essential Standard: Essential |
Artificial Intelligence in Healthcare Safety (AI-001) Evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or a training certificate in the staff file |
AI-SOP-001 AI-REG-001 |
| Nurse ISCO 2221 ASF proposal |
Uses or checks the equipment and falls controls at the bedside | Humanitarian: Essential Transition: Essential Standard: Essential |
Artificial Intelligence in Healthcare Safety (AI-001) Evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or a training certificate in the staff file |
— |
E-learning: who must take each course.
| E-learning course | Who must take | Who is recommended |
|---|---|---|
| Artificial Intelligence in Healthcare Safety (AI-001) GMJ Academy: English · Georgian |
Manager (ISCO 1342), Nurse (ISCO 2221) | None |
Documents: who does each step
| Document | Writes | Approves | Fills in | Checks and signs | Files and uploads |
|---|---|---|---|---|---|
| AI-SOP-001 · AI Performance and Safety Monitoring Procedure (SOP) · To monitor AI systems after introduction for over-diagnosis, missed diagnosis, complaints and other unintended effects, and to act on findings. |
Manager (ISCO 1342) | Chief executive (ISCO 1120) | Manager (ISCO 1342) | Chief executive (ISCO 1120) | Secretary (ISCO 3344) |
| AI-REG-001 · AI Monitoring Register Register · Records each AI audit, complaint, finding and action, for the last four quarters and onward. |
Secretary (ISCO 3344) | Chief executive (ISCO 1120) | Secretary (ISCO 3344) | Manager (ISCO 1342) | Secretary (ISCO 3344) |
References
- World Health Organization (2021). Ethics and governance of artificial intelligence for health. Used for: Monitoring AI for unintended consequences. https://news.gmj.ge/guideline/ethics-and-governance-of-artificial-intelligence-for-health/
5.6
Staff Are Consulted Before an AI System Is Introduced
Humanitarian settings: MinimumTransition and fragile: MediumStandard: Must
Staff who will deliver care using a new AI system are consulted before it is introduced, to understand the practical implications and identify training needs — not an AI tool rolled out to clinical staff with no engagement beyond a brief notification that it now exists.
Rating: Minimum in humanitarian settings; Medium in transition and fragile settings; Must in standard settings.
Why it matters. Staff who will use an AI system every day know the practical problems that a manager may not. Consulting them before it is introduced finds those problems early.
How the surveyor verifies this criterion
Sample. Review the consultation record for each AI system introduced in the last 12 months.
1. Staff are consulted before introduction.
Method. Check the consultation record dated before the decision was final.
Evidence to show. AI-FRM-002 Staff Consultation Record.
Pass when. Consultation is recorded before the decision.
2. Training needs are identified and met before go-live.
Method. Check that the consultation identified training needs and that training was recorded before go-live.
Evidence to show. AI-SOP-002 Staff Consultation Before AI Introduction.
Pass when. Identified training was completed before go-live.
3. Staff can describe the consultation.
Method. Ask staff who use a recent AI system whether they were consulted.
Evidence to show. AI-FRM-002 Staff Consultation Record.
Pass when. Staff confirm the consultation took place.
Immediate fail (any one means the criterion is not met)
- An AI system goes live with no staff consultation before the decision was final.
Staff, training and documents
Each occupation carries its ISCO-08 code. Each row is one document: read across to see who does each step. Training evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or your own training certificate kept in the staff file.
Good practice. Staff who will use the AI system are consulted before the decision is final. The training needs they raise are met before go-live, and staff confirm that the consultation took place.
Partial because. Staff were consulted before the decision, but the training identified was not completed before go-live, or the consultation is not recorded.
Staff responsible
| Role | Why it applies | Level | Training | Documents |
|---|---|---|---|---|
| Manager ISCO 1342 ASF proposal |
Runs the facility process this criterion measures | Humanitarian: Essential Transition: Essential Standard: Essential |
Artificial Intelligence in Healthcare Safety (AI-001) Evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or a training certificate in the staff file |
AI-SOP-002 AI-FRM-002 |
| Nurse ISCO 2221 ASF proposal |
Uses or checks the equipment and falls controls at the bedside | Humanitarian: Recommended Transition: Recommended Standard: Recommended |
Artificial Intelligence in Healthcare Safety (AI-001) Evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or a training certificate in the staff file |
— |
E-learning: who must take each course.
| E-learning course | Who must take | Who is recommended |
|---|---|---|
| Artificial Intelligence in Healthcare Safety (AI-001) GMJ Academy: English · Georgian |
Manager (ISCO 1342) | Nurse (ISCO 2221) |
Documents: who does each step
| Document | Writes | Approves | Fills in | Checks and signs | Files and uploads |
|---|---|---|---|---|---|
| AI-SOP-002 · Staff Consultation Before AI Introduction Procedure (SOP) · To make sure staff who will deliver care with an AI system understand its practical effect, raise their concerns, and are trained before go-live. |
Manager (ISCO 1342) | Chief executive (ISCO 1120) | Manager (ISCO 1342) | Chief executive (ISCO 1120) | Secretary (ISCO 3344) |
| AI-FRM-002 · Staff Consultation Record Form (record) · Records the consultation with staff before an AI system is introduced, with concerns and responses. |
Manager (ISCO 1342) | Chief executive (ISCO 1120) | Secretary (ISCO 3344) | Manager (ISCO 1342) | Secretary (ISCO 3344) |
References
- World Health Organization (2021). Ethics and governance of artificial intelligence for health. Used for: Staff consultation and stakeholder involvement. https://news.gmj.ge/guideline/ethics-and-governance-of-artificial-intelligence-for-health/
5.7
Accountability for AI-Assisted Care Is Explicitly Defined
Humanitarian settings: MinimumTransition and fragile: MediumStandard: Must
The facility has considered and documented accountability arrangements for care and treatment delivered with AI support — who is responsible when an AI-assisted decision is wrong — not leaving this as an unexamined question until the first time it matters.
Rating: Minimum in humanitarian settings; Medium in transition and fragile settings; Must in standard settings.
Why it matters. When an AI system contributes to a wrong decision, it is not clear who is responsible: the clinician, the vendor or the facility. A documented accountability arrangement answers that question before there is an incident.
How the surveyor verifies this criterion
Sample. Review the accountability policy and the records of AI-assisted decisions in the last 12 months.
1. Accountability is explicitly documented.
Method. Check the policy and a sample of AI-assisted decision records.
Evidence to show. AI-POL-002 Accountability for AI-Assisted Care.
Pass when. Clinical accountability is stated in the policy and the records.
2. Clinicians understand their accountability.
Method. Ask clinicians who use AI-supported tools to describe their accountability for the final decision.
Evidence to show. AI-POL-002 Accountability for AI-Assisted Care.
Pass when. Clinicians describe it correctly.
3. Accountability is reviewed after an incident.
Method. Check the process for any AI-related safety incident.
Evidence to show. AI-POL-002 Accountability for AI-Assisted Care; INC-SOP-001 Root Cause Analysis of Serious Incidents.
Pass when. A review of accountability follows any such incident.
Immediate fail (any one means the criterion is not met)
- Accountability for an AI-assisted decision is not documented, and no review follows an AI-related safety incident.
Staff, training and documents
Each occupation carries its ISCO-08 code. Each row is one document: read across to see who does each step. Training evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or your own training certificate kept in the staff file.
Good practice. A written policy states who is accountable for an AI-assisted decision, and the care records reflect it. Clinicians can describe their accountability, and it is reviewed after any AI-related incident.
Partial because. Accountability is written in the policy, but clinicians cannot describe it, or it is not reflected in the records of AI-assisted decisions.
Staff responsible
| Role | Why it applies | Level | Training | Documents |
|---|---|---|---|---|
| Chief executive ISCO 1120 ASF proposal |
Carries accountability to the governing body | Humanitarian: Essential Transition: Essential Standard: Essential |
Artificial Intelligence in Healthcare Safety (AI-001) Evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or a training certificate in the staff file |
AI-POL-002 |
| Manager ISCO 1342 ASF proposal |
Runs the facility process this criterion measures | Humanitarian: Essential Transition: Essential Standard: Essential |
Artificial Intelligence in Healthcare Safety (AI-001) Evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or a training certificate in the staff file |
— |
| Nurse ISCO 2221 ASF proposal |
Uses or checks the equipment and falls controls at the bedside | Humanitarian: Essential Transition: Essential Standard: Essential |
Artificial Intelligence in Healthcare Safety (AI-001) Evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or a training certificate in the staff file |
— |
E-learning: who must take each course.
| E-learning course | Who must take | Who is recommended |
|---|---|---|
| Artificial Intelligence in Healthcare Safety (AI-001) GMJ Academy: English · Georgian |
Chief executive (ISCO 1120), Manager (ISCO 1342), Nurse (ISCO 2221) | None |
Documents: who does each step
| Document | Writes | Approves | Fills in | Checks and signs | Files and uploads |
|---|---|---|---|---|---|
| AI-POL-002 · Accountability for AI-Assisted Care Policy · To define, in writing and in advance, who is accountable when an AI-assisted care decision is wrong, so that the question is answered before it matters. |
Chief executive (ISCO 1120) | Governing body members | N/A | Chief executive (ISCO 1120) | Secretary (ISCO 3344) |
References
- World Health Organization (2021). Ethics and governance of artificial intelligence for health. Used for: Accountability. https://news.gmj.ge/guideline/ethics-and-governance-of-artificial-intelligence-for-health/
5.8
Patients Are Informed When Their Care Involves AI
Humanitarian settings: MinimumTransition and fragile: MediumStandard: Must
A patient whose care involves an aspect delivered with AI system support is informed of this — not left to assume every decision was made by a human clinician alone, with disclosure treated as optional rather than a standard part of informed care.
Rating: Minimum in humanitarian settings; Medium in transition and fragile settings; Must in standard settings.
Why it matters. A patient who is treated with the help of AI has a right to know. Telling them, rather than leaving them to assume, is part of informed consent and trust.
How the surveyor verifies this criterion
Sample. Review the patient information and the records of AI-supported care for the last 12 months.
1. Patients are informed when AI supports their care.
Method. Check the information given and the record of disclosure.
Evidence to show. AI-SOP-003 Informing Patients About AI-Supported Care.
Pass when. Disclosure is recorded for AI-supported care.
2. The information is understandable.
Method. Read the patient information for plain language and check it is not buried in a general consent form.
Evidence to show. AI-FRM-003 Patient Information on AI-Supported Care (template).
Pass when. The information is written in plain language and given separately.
3. Patients confirm they were told.
Method. Ask a sample of patients whether they were told AI was involved.
Evidence to show. AI-FRM-003 Patient Information on AI-Supported Care (template).
Pass when. Patients confirm they were informed.
Immediate fail (any one means the criterion is not met)
- A patient is not informed that AI supported their care, and no information is available.
Staff, training and documents
Each occupation carries its ISCO-08 code. Each row is one document: read across to see who does each step. Training evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or your own training certificate kept in the staff file.
Good practice. Patients are told when AI supports their care, in plain language and as separate information. The disclosure is recorded, and patients confirm that they were told.
Partial because. Patient information exists, but disclosure is not recorded for every case of AI-supported care, or the information is not written in plain language.
Staff responsible
| Role | Why it applies | Level | Training | Documents |
|---|---|---|---|---|
| Manager ISCO 1342 ASF proposal |
Runs the facility process this criterion measures | Humanitarian: Essential Transition: Essential Standard: Essential |
Artificial Intelligence in Healthcare Safety (AI-001) Evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or a training certificate in the staff file |
AI-SOP-003 AI-FRM-003 |
| Nurse ISCO 2221 ASF proposal |
Uses or checks the equipment and falls controls at the bedside | Humanitarian: Essential Transition: Essential Standard: Essential |
Artificial Intelligence in Healthcare Safety (AI-001) Evidence: a verified micro-credential in your GMJ Academy micro-credential portfolio, or a training certificate in the staff file |
— |
E-learning: who must take each course.
| E-learning course | Who must take | Who is recommended |
|---|---|---|
| Artificial Intelligence in Healthcare Safety (AI-001) GMJ Academy: English · Georgian |
Manager (ISCO 1342), Nurse (ISCO 2221) | None |
Documents: who does each step
| Document | Writes | Approves | Fills in | Checks and signs | Files and uploads |
|---|---|---|---|---|---|
| AI-SOP-003 · Informing Patients About AI-Supported Care Procedure (SOP) · To make sure patients know when AI is part of their care, in words they understand, and can ask questions about it. |
Manager (ISCO 1342) | Chief executive (ISCO 1120) | Manager (ISCO 1342) | Chief executive (ISCO 1120) | Secretary (ISCO 3344) |
| AI-FRM-003 · Patient Information on AI-Supported Care (template) Form (record) · Plain-language notice telling patients when an AI system supported their care, and what that means. |
Manager (ISCO 1342) | Chief executive (ISCO 1120) | Secretary (ISCO 3344) | Manager (ISCO 1342) | Secretary (ISCO 3344) |
References
- World Health Organization (2021). Ethics and governance of artificial intelligence for health. Used for: Informing patients. https://news.gmj.ge/guideline/ethics-and-governance-of-artificial-intelligence-for-health/
Role details: reporting, supervision and training
Each role links to its own terms of reference, which holds the full reporting line, supervision, training and competency record.
| Role | Reports to | Supervises | Core training |
|---|---|---|---|
| Chief executive ISCO 1120 | The board of directors. | The executive team (COO, CMO, CNO, CFO, CQO, CHRO, CIO, CSO), and the legal, data protection, compliance and communications leads. | Artificial Intelligence in Healthcare Safety (AI-001) |
| Manager ISCO 1342 | The chief information officer (CIO). | Systems, database, network and help desk teams. | Technology Platform Reliability & Security (TM-01) Artificial Intelligence in Healthcare Safety (AI-001) Artificial Intelligence Systems for Care (TELE-10, Georgian version only) |
| Nurse ISCO 2221 | The head nurse. | Nursing associates and health care assistants. | Artificial Intelligence Systems for Care (TELE-10, Georgian version only) Artificial Intelligence in Healthcare Safety (AI-001) |
| Secretary ISCO 3344 | The head of department (per specialty). | No direct reports. | Technology Platform Reliability & Security (TM-01) |
Who is responsible for Standard 5
Each criterion below names the positions responsible for it.
| Criterion | Positions responsible or involved | |
|---|---|---|
| 5.1 | Digital Care Systems Follow an Assessment and Management Process | Manager · Secretary |
| 5.2 | Digital Care Never Disadvantages a Patient Who Cannot Use It | Manager · Secretary |
| 5.3 | Technical Expertise Supports Digital Care Systems | Manager |
| 5.4 | AI Systems Are Introduced in Line With WHO AI Governance Guidance | Chief executive · Manager · Nurse |
| 5.5 | AI Systems Are Monitored for Unintended Consequences | Manager · Nurse |
| 5.6 | Staff Are Consulted Before an AI System Is Introduced | Manager · Nurse |
| 5.7 | Accountability for AI-Assisted Care Is Explicitly Defined | Chief executive · Manager · Nurse |
| 5.8 | Patients Are Informed When Their Care Involves AI | Manager · Nurse |
Quality oversight
The quality chain keeps this standard under review, from the governing body down to the teams that run each process.
Questions and answers
The criterion text is the authority. These answers summarise it.
Do I need to meet every criterion in Standard 5?
Yes. Standard 5 is a core standard, and all eight criteria apply to all nine facility types. Each criterion has its own pass rule and its own immediate-fail rule, so each one must be met on its own. Criteria 5.1 to 5.3 are about digital care systems, and criteria 5.4 to 5.8 are about artificial intelligence. The samples for 5.4 to 5.8 are the AI systems in clinical use. If your facility uses no AI system at all, tell ASF before the survey so that the scope can be confirmed.
What is an immediate fail?
An immediate fail is one event or finding that means a criterion is not met, even if everything else in that criterion is in place. For example, if an AI system is in clinical use with no assessment and no governance basis, criterion 5.4 fails outright. Each criterion on this page lists its own immediate fail under “Immediate fail”. Facilities should read these before the survey, because they are the points on which the outcome is decided most quickly.
Which documents do I submit?
The documents are the procedures, policies, forms and registers listed under each criterion in the “Documents: who does each step” table. Records, such as the Digital Tool Assessment Template (DIG-FRM-002) completed for each system, or the AI Monitoring Register (AI-REG-001), are kept as working files. Policies and procedures are signed documents that show the version, date, owner and approver. Upload or have them ready before the survey, and check that each one covers the period stated in its criterion, usually the last 12 months. You may use your own document if it meets the same check, as long as it is clearly the same record.
What counts as a digital care system?
Any digital system used to deliver or support care: a telehealth platform, a remote monitoring tool, a patient-facing app, or clinical software. The criterion is about how the system came into use. Each one needs an assessment made by the facility itself before it is introduced, and it must be managed afterwards. A system adopted because a vendor presented it well, with no internal evaluation, does not meet criterion 5.1.
What must the assessment of a digital system cover?
Three things, all recorded on the Digital Tool Assessment Template (DIG-FRM-002) under the assessment procedure (DIG-SOP-002). First, a cost and benefit analysis dated before go-live. Second, interoperability, meaning how the system works with the systems you already use, and the risks of that. Third, the unintended consequences considered in advance, as a list of risks with a mitigation for each one. The surveyor reviews the assessment for each digital care system in use. A system introduced with no cost and benefit analysis and no interoperability check is an immediate fail.
What do we owe patients who cannot use digital care?
An alternative route that gives equivalent care. Some patients cannot use an app, have no internet access, or find video calls hard, and they must not receive a reduced service. The procedure is DIG-SOP-003, and the Digital Alternative Record (DIG-FRM-003) records the route each patient prefers and any alternative used. The surveyor also checks that the digital service was tested beforehand with people who might struggle to use it. A patient with no alternative route who receives reduced service is an immediate fail.
What technical support is expected?
Each digital care system needs a named route to technical expertise, which may be in-house, provided by the vendor or contracted. The arrangement is written in the Technical Support and Testing Arrangement (DIG-PLN-001), and tests are recorded on the Digital System Test and Support Record (DIG-FRM-004). Testing must be done before full implementation. The surveyor reviews the testing records and the support log, and asks staff how they get support, so staff should be able to describe the route.
Which AI systems does the standard cover, and what does it ask?
It covers any artificial intelligence system used to support diagnosis, triage or care decisions. Such a system must be introduced and managed in line with the World Health Organization guidance, Ethics and governance of artificial intelligence for health (2021), which asks for safety, transparency and human oversight. Each system needs an assessment made before introduction that cites the guidance (AI-FRM-001), and an approved Responsible Use of AI in Care Policy (AI-POL-001) that covers it. The surveyor reviews the assessment and the policy for each AI system in clinical use.
How do we monitor an AI system after it is introduced?
An AI system can work well at first and then change as the patients change, so its results must be checked regularly. Under the monitoring procedure (AI-SOP-001), the output of each system is audited for over-diagnosis and missed diagnosis, and complaints related to AI-assisted care are logged and reviewed. Audits, complaints, findings and actions are recorded on the AI Monitoring Register (AI-REG-001). The surveyor reviews the register for the last 12 months and looks for at least one case where a finding led to a response. An AI system whose output is not checked, with a known problem left unanswered, is an immediate fail.
Why must staff be consulted before an AI system is introduced?
Staff who will use the system every day know practical problems that a manager may not see. They must be consulted before the decision is final, not informed afterwards. The procedure is AI-SOP-002, and the consultation, with the concerns raised and the responses, is recorded on the Staff Consultation Record (AI-FRM-002). Training needs found in the consultation must be met before go-live. The surveyor reviews the record for each AI system introduced in the last 12 months and asks staff whether the consultation took place.
Who is accountable when an AI-assisted decision is wrong?
The standard does not give one answer for every facility. It requires the facility to decide and to write the answer down in advance, in the Accountability for AI-Assisted Care policy (AI-POL-002), so that the question is settled before an incident and not after it. Clinical accountability must be stated in the policy and reflected in the records of AI-assisted decisions, and clinicians must be able to describe it. After any AI-related safety incident, accountability is reviewed, using the root cause analysis procedure (INC-SOP-001).
Do we have to tell patients that AI was used?
Yes. A patient whose care involves AI support must be told, as a normal part of informed care. The procedure is AI-SOP-003, and the patient information template (AI-FRM-003) gives a plain-language notice that is handed over as separate information, not hidden in other paperwork. The disclosure is recorded. The surveyor reviews the patient information and the records of AI-supported care for the last 12 months, and asks patients to confirm that they were told.
What do the colours mean?
Red means Must: the criterion is required in every setting, and criteria that protect life are always red. Yellow means Medium: the criterion is required, but adapted to what a transition setting can realistically do. Green means Minimum: the lowest acceptable level in a humanitarian setting, where resources are very limited. In standard settings every criterion is Must. Each criterion shows its rating for each of the three settings, so a facility can see exactly what applies to it.
Do we have to use your forms and policies?
No. Every policy, procedure, form and course on this page is a reference draft to help you meet the criterion. You can adapt it to your facility, write your own, or upload your own documents. What the assessor checks is that the standard is met and that you can show how. If you use your own document, make sure it covers the same content as the reference and that it is current, signed and dated, so that the assessor can match it to the criterion without difficulty.
What does essential or recommended training mean for my staff?
In the training tables, red means essential: the course is required for that role in that setting. Green means recommended: the course is advised for that role. Yellow means optional. Each course shows its level for humanitarian, transition and fragile, and standard settings, because the requirement can differ between them. Staff evidence is a verified micro-credential in the GMJ Academy micro-credential portfolio, or a training certificate kept in the staff file.
Do we have to use ISCO-08 codes?
The staff tables give each occupation its ISCO-08 code, so that roles can be compared between facilities and countries. You are not required to change your job titles. What matters is that each role named in the standard can be matched to the person who does that job and to the training they have had. If your titles differ, record the match in your staff file so that the assessor can follow it.
Which training do staff need?
The e-learning table under each criterion shows who must take each course and who is recommended. Two criteria, 5.2 and 5.3, name no required course; an optional academy course is shown there instead. The table is read together with the staff table, which shows the level for each setting. Each course name opens the course page in the GMJ Academy. Staff can show evidence through a verified micro-credential in the GMJ Academy micro-credential portfolio, or through a training certificate kept in the staff file.
What will the surveyor see on the day?
The surveyor works through the criteria, following the “Sample” and the checks given for each one. For each digital care system, the surveyor reviews the assessment, the testing records and the support log. For each AI system, the surveyor reviews the assessment, the policy, the monitoring register, the staff consultation record and the patient information. The surveyor also talks to people: staff are asked how they get technical support and whether they were consulted, clinicians are asked to describe their accountability, and patients are asked whether they were told that AI supported their care.
How does the surveyor choose which records to check?
Each criterion states its own sample, such as “Review the assessment for each digital care system in use” or “Review the monitoring register for each AI system in use over the last 12 months”. The surveyor uses the sample given on the criterion rather than choosing records to suit the facility. Preparing for the survey means keeping a list of every digital care system and every AI system in use, with its records organised so that any item can be produced quickly. A facility that cannot produce the sampled records loses the criterion, whatever else it has in place.
What if we do not have the written protocol yet?
A criterion cannot be met on the basis of an unwritten or informal arrangement, because the assessor needs something to check against. If a protocol does not yet exist, write one before the survey, using the reference draft or your own version, and make sure it is signed, dated and in use. Staff must be able to explain it in practice.
Where can I get the full text?
The full text of each criterion, with its rating, why it matters, the assessor checks, the immediate-fail rule and the documents, is on this page. A PDF with the same criteria, for printing and for assessors, can be downloaded from the box at the top of this page. The questions and answers on this page summarise the criteria, and if they ever differ, the criterion text is the authority.
Who can I ask about this standard?
For questions about this standard, contact Accréditation Sans Frontières through the Contact page of this website. Please give the standard number and the criterion (for example 5.4) so that your question can be answered against the right text.
E-learning by standard
These are the GMJ Academy courses named in the criteria of this standard. Each course name opens the course page in the academy. Where a course is published in English and in Georgian, both versions are linked. Evidence of completion is a verified micro-credential in the GMJ Academy micro-credential portfolio, or a training certificate in the staff file.
| E-learning course | Who must take | Who is recommended | Criteria |
|---|---|---|---|
| Technology Platform Reliability & Security (TM-01) GMJ Academy: English · Georgian |
Manager (ISCO 1342) | Secretary (ISCO 3344) | 5.1 |
| Artificial Intelligence in Healthcare Safety (AI-001) GMJ Academy: English · Georgian |
Chief executive (ISCO 1120), Manager (ISCO 1342), Nurse (ISCO 2221) | None | 5.4, 5.5, 5.6, 5.7, 5.8 |
| Artificial Intelligence Systems for Care (TELE-10, Georgian version only) GMJ Academy: Georgian |
Manager (ISCO 1342) | Nurse (ISCO 2221) | 5.4 |
Related courses in the GMJ Academy
These courses cover the same subjects in more depth or for a specific type of facility. They are optional. They are not required by the criteria.
| Course | Subject | Criteria |
|---|---|---|
| Digital Care Assessment and Management Process (BAS-06) GMJ Academy: English |
Assessing and managing a digital care system | 5.1 |
| Data Privacy & Patient Confidentiality (TM-06) GMJ Academy: English · Georgian |
Data privacy in digital care | 5.1 |
| Digital Inclusion for Patients Who Cannot Use Digital Tools (BAS-07) GMJ Academy: English · Georgian |
Patients who cannot use digital tools | 5.2 |
| Technical Support for Digital Care Systems GMJ Academy: English |
Technical support for digital systems | 5.3 |
| Digital Care and Artificial Intelligence in the Lab (LAB-12) GMJ Academy: English · Georgian |
Digital care and AI for your type of service | 5.4 to 5.8 |
| Digital Care and Artificial Intelligence (MT-10) GMJ Academy: English · Georgian |
Digital care and AI for your type of service | 5.4 to 5.8 |
| Digital Care and AI Systems for Care (FIT-12) GMJ Academy: English · Georgian |
Digital care and AI for your type of service | 5.4 to 5.8 |
Back to ASF Health Standards V11 · Standards Map · E-learning by standard · References
References and guidelines for this standard
The guidelines and references cited in the criteria above are listed here. Each link opens the full text.
- World Health Organization (2021). Recommendations on digital interventions for health system strengthening. Used for: Digital systems assessment. https://news.gmj.ge/guideline/recommendations-on-digital-interventions-for-health-system-strengthening/
- World Health Organization (2024). Implementation toolkit for accessible telehealth services. Used for: Accessible digital care and alternatives. https://news.gmj.ge/guideline/implementation-toolkit-for-accessible-telehealth-services/
- World Health Organization (2022). Consolidated telemedicine implementation guide. Used for: Technical support and testing. https://news.gmj.ge/guideline/consolidated-telemedicine-implementation-guide/
- World Health Organization (2021). Ethics and governance of artificial intelligence for health. Used for: WHO guidance on AI governance. https://news.gmj.ge/guideline/ethics-and-governance-of-artificial-intelligence-for-health/