AI, Virtual Care, and Accreditation

Strengthening Indigenous Health Systems Through Evaluation and Accreditation

Sidney Shapiro, PhD

Dhillon School of Business

University of Lethbridge

What is AI? Presentation slides

AI, Virtual Care, and Accreditation

For the Regional Accreditation Networking Group (RANG)

Welcome

This session is designed for accreditation leaders in First Nations health organizations. It focuses on how AI-enabled virtual care can support quality improvement, safety, and accreditation readiness—while respecting Indigenous data governance and self-determination.

AI is not a technology project in isolation. It is a quality and systems issue. The materials here introduce what AI in health actually means, where it can bridge gaps (documentation, triage, chronic disease management), and how adoption can align with the standards and values you already uphold.

Core question How can AI-enabled virtual care strengthen quality and accreditation readiness in your organizations?

Why this conversation now

  • Severe health human resource shortages across Canada
  • Colonial legacies, underfunding, and barriers to care have contributed to health disparities.
  • Higher prevalence of chronic conditions in many First Nations communities
  • Geographic and infrastructure barriers to care access
  • Increasing digital infrastructure in rural and remote regions
  • Accreditation focus on quality, safety, and continuous improvement

Examples of AI in healthcare

AI is already in use in clinical and operational settings. These examples illustrate where it can support—or where it is being evaluated for—quality, documentation, and safety.

Examples of AI in program evaluation

Frameworks and initiatives that use or evaluate AI in health care emphasize fairness, usefulness, reliability, and real-world impact—all relevant to accreditation and quality improvement.

What you'll find here

Presentation slides

Full slide deck (~20 minutes) covering context, AI in health, accreditation lens, Indigenous innovation, and discussion prompts.

Open presentation

Other tabs & pages

  • What is AI? — Plain-language explanation of AI, ML, and LLMs
  • First Nations AI examples — Links to programs and platforms (Mila, FNTC, language, practical applications)
  • Prompt builder — Sample prompts for healthcare and evaluation, with PDF download
  • AI safety — Examples and best practices
Open presentation slides

First Nations AI Examples

Programs, initiatives, and platforms where AI is being used with or by Indigenous communities—for leadership in AI, governance, language and culture, and practical applications. Links are to official or widely cited sources.

Indigenous leadership in AI

Language and cultural preservation

Practical applications: environment and knowledge

Health and data governance

First Nations and Inuit health authorities and regional organizations often publish their own data and technology frameworks. Check your region’s health authority for local policies on data sovereignty and AI use.

Prompt builder for healthcare and evaluation

Use the formula Context + Task + Format + Quality to get more consistent, useful outputs from AI tools. Build your prompt below, copy it into ChatGPT, Copilot, or a secure internal assistant, and always verify output before using it in records or decisions.

The core prompt formula

📝

Context

Who, where, what background

+

Task

What to do

+
📐

Format

Structure & layout

+
🎨

Quality

Tone & style

=

Effective prompt

Better, usable output

Build your prompt

Prompting patterns

Different situations call for different ways of prompting. Use these patterns to get better results.

Context-rich prompting

Give full background upfront. Best when you need an accurate, complete result on the first try (e.g. accreditation evidence checklist, report outline).

Example: “Context: I am an accreditation coordinator at a First Nations health centre. We are preparing for [standard] review. Task: Create a checklist of evidence we need for each criterion. Format: Table with criterion, type of evidence, and where we store it. Quality: Plain language for staff.”

Best for: Complex tasks, first-time outputs, when accuracy matters.

Iterative refinement

Start with a short prompt, then refine with follow-ups. E.g. “Draft a one-page QI summary” → “Add dates to each action” → “Make the recommendations more specific.”

Best for: When you’re still shaping what you want; drafts and revisions.

Template-based prompting

Specify the structure. E.g. “Create a report with these sections in order: 1) Executive summary, 2) Objectives, 3) Methods, 4) Findings, 5) Limitations, 6) Recommendations. Use headings; leave placeholders for our data.”

Best for: Consistent formatting, evaluation reports, completeness.

Chain-of-thought

Ask for step-by-step reasoning before the answer. E.g. “Analyze our program outcome data step-by-step: 1) Summarize the metrics, 2) Compare to target, 3) Identify gaps, 4) Suggest possible causes. Show your reasoning for each step.”

Best for: Analysis, interpretation, when you need to see how the AI got there.

Role-based prompting

Assign a role so the tone and level fit the audience. E.g. “Act as an experienced accreditation coordinator. Draft a one-page summary of our quality improvement activities for our board. Use professional but accessible language; no jargon.”

Best for: Board reports, stakeholder summaries, external reviewers.

Essential prompting tips

Be specific

Instead of “summarize the meeting,” try “Summarize the quality improvement meeting as bullet points: key decisions, action items with owners, and follow-up date.”

Use examples

“Format the recommendations like this: [paste one example]. Generate 3 more in the same style but for [your topic].”

Set constraints

“Keep it under 300 words.” “Use exactly 5 bullet points.” “One page only.” “No medical jargon; suitable for community members.”

Iterate and refine

Use follow-ups: “Make it more concise,” “Add a limitations section,” “Change the tone to be more formal.”

Specify tone and audience

“Professional, suitable for leadership.” “Plain language for staff and community.” “Clear and concise for external accreditors.”

Reference your data (when safe)

If your tool and policy allow: “Using the following summary of our QI activities [paste], draft a one-page report.” Never paste identifiable patient data into public AI.

Common mistakes to avoid

  • Too vague: “Write something about our program.” → Better: “Draft a one-page program description for our website: who we serve, what we offer, how to access. Plain language.”
  • Missing context: “Create a report.” → Better: “Create an internal evaluation report outline for our chronic disease program, with sections for objectives, methods, findings, and recommendations.”
  • No format: “Summarize the notes.” → Better: “Summarize the meeting notes as bullet points with dates and action items.”
  • Ignoring quality: “Write an email.” → Better: “Write a professional, respectful email to a community partner confirming the next meeting and agenda.”

Healthcare & accreditation prompts

Copy or adapt these. Always verify AI output before using it in records or for decisions.

Documentation summary

Context: Accreditation coordinator at a First Nations health centre.
Task: Summarize the last six months of quality improvement meeting notes into one page: key decisions, actions, and follow-ups.
Format: Bullet points with dates.
Quality: Professional, suitable for leadership.

Evaluation report outline

Context: Preparing an internal evaluation report for a chronic disease program.
Task: Create an outline for the report that includes: objectives, methods, findings, limitations, and recommendations. Leave placeholders for data.
Format: Report sections with headings.
Quality: Clear and concise for external reviewers.

Policy checklist

Context: Our organization is aligning policies with [accreditation standard].
Task: Generate a checklist of evidence we might need to show for each criterion (e.g. written policy, training records, audit log).
Format: Checklist with short explanations.
Quality: Plain language, accessible to staff.

Meeting agenda from goals

Context: We have a 90-minute QI meeting; our goals this quarter are [list].
Task: Draft an agenda with time allocations: review of last actions, discussion of [topic], decisions needed, next steps.
Format: Numbered agenda with times.
Quality: Realistic and actionable.

Risk register template

Context: We need a simple risk register for accreditation and board reporting.
Task: Create a table template: Risk description | Likelihood | Impact | Mitigation | Owner. Include 2–3 example rows (generic) we can replace.
Format: Table.
Quality: Plain language; suitable for non-specialists.

Plain-language consent summary

Context: We have a consent form for [program/service]. We need a one-page plain-language summary for participants.
Task: Summarize: what the program does, what we ask of participants, how we use and protect their information, that participation is voluntary. No legal jargon.
Format: Short sections with headings.
Quality: Accessible; suitable for community members; do not replace legal review.

Program evaluation tools & prompts

Prompts for logic models, theory of change, Gantt charts, M&E plans, and related evaluation tools. Copy or adapt; verify output before use.

Logic model

Context: We are planning or evaluating a community health program (e.g. diabetes prevention, mental health support).
Task: Draft a logic model with: inputs, activities, outputs, short-term outcomes, and long-term impact. Include brief definitions for each column. Leave blanks for us to fill with our program details.
Format: Table or one-page diagram with clear column headers.
Quality: Usable for funders and accreditation; plain language.

Theory of change

Context: We need a theory of change for [program name] to show how our activities lead to outcomes and impact.
Task: Create a theory-of-change outline: if we do [activities], then [short-term outcomes], so that [long-term impact]. Include assumptions and risks we should document. Use a flow or “if-then-so that” structure.
Format: Numbered or bulleted flow; optional simple diagram description.
Quality: Clear, logical; suitable for proposal or evaluation planning.

Gantt chart / timeline

Context: We are planning a [program rollout / evaluation / accreditation preparation] over [timeframe, e.g. 12 months].
Task: Suggest a Gantt-style timeline: list key phases or workstreams (e.g. planning, recruitment, delivery, data collection, reporting) with suggested duration and order. Output as a table: Phase | Start | End | Dependencies or notes. We will build the actual chart in Excel or project software.
Format: Table with columns for task phase, start, end, notes.
Quality: Realistic for teams with limited capacity; include buffer for reporting.

Outcome indicators

Context: We run a [type of program] and need to report on outcomes for our board and accreditors.
Task: Suggest 5–7 outcome indicators we could measure (process and result). For each: how to measure and one limitation to watch for.
Format: Table: Indicator | How to measure | Limitation.
Quality: Practical and feasible for many teams.

Results framework

Context: We need a simple results framework linking our program objectives to indicators and data sources.
Task: Create a table: Objective | Indicator(s) | Data source | Frequency. Leave 1–2 example rows; we will fill the rest. Include a short note on what a “results framework” is for.
Format: Table with clear column headers.
Quality: Plain language; suitable for M&E or funder reporting.

M&E plan outline

Context: We are designing monitoring and evaluation for [program name].
Task: Draft an M&E plan outline with sections: (1) purpose and key questions, (2) indicators and targets, (3) data collection methods and schedule, (4) roles and responsibilities, (5) reporting and use of findings. Leave placeholders for our specifics.
Format: Numbered sections with sub-bullets.
Quality: Usable for a community or health program; not overly technical.

Stakeholder summary (evaluation)

Context: We have completed an internal evaluation of [program name] and have raw findings.
Task: Turn our findings into a one-page stakeholder summary: what we did, what we found, what we recommend. No jargon; suitable for community and leadership.
Format: Short sections with headings; bullet points for recommendations.
Quality: Clear, respectful, actionable.

Accreditation prompts

Prompts for the accreditation cycle: evidence collection, self-assessment, reviewer preparation, and follow-up. Copy or adapt for your standard and timeline. Always verify AI output before using it in your submission or records.

Evidence checklist by standard

Context: We are preparing for [accreditation standard name] review. We need to map evidence to each criterion.
Task: For each criterion in [standard/section], list the type of evidence typically required (e.g. policy document, training record, meeting minutes, audit result). Include a column for “where we store it” as a placeholder. Leave room for our notes.
Format: Table: Criterion | Evidence type | Where we store it | Notes.
Quality: Plain language; usable as a working checklist for staff.

Self-assessment by criterion

Context: First Nations health organization conducting an internal self-assessment against [standard] before our review.
Task: Create a self-assessment template: for each criterion, include (1) a one-line description of what we need to demonstrate, (2) space for our rating (e.g. met / partial / not met / N/A), (3) space for evidence we will cite, (4) space for gaps and action plans.
Format: Table or numbered list with sub-bullets.
Quality: Clear and concise; suitable for leadership review.

Reviewer briefing / one-pager

Context: External reviewers are visiting for our [accreditation] review. We want a one-page briefing about our organization and how we have prepared.
Task: Draft a one-page briefing structure: who we are (organization, community context), our accreditation scope, key strengths we will highlight, how we have organized evidence (e.g. by standard, by site), and one or two questions we would like to discuss. Leave placeholders for us to fill in our specifics.
Format: One page; short sections with headings.
Quality: Professional, respectful; suitable for external accreditors.

Post-review action plan

Context: We have received our accreditation report with findings and recommendations. We need to turn them into an internal action plan.
Task: Create an action plan template: for each finding or recommendation, include (1) summary, (2) responsible person/team, (3) due date, (4) status, (5) evidence of completion. Include 2–3 example rows we can replace with our actual items.
Format: Table with clear column headers.
Quality: Actionable; suitable for board or leadership reporting.

Gap analysis (readiness)

Context: We are [X months] from our next [accreditation] review and want to assess readiness.
Task: Create a gap analysis framework: list key areas (e.g. policies, training, documentation, quality improvement, governance). For each area, include: what the standard expects, how we would demonstrate compliance, current status (ready / in progress / gap), and priority actions. Leave blanks for us to complete.
Format: Table or sections with sub-bullets.
Quality: Practical; usable for planning and assigning tasks.

Continuous improvement log for accreditation

Context: We need to document ongoing quality improvement activities for our next accreditation cycle.
Task: Create a simple log template: Date | Activity / initiative | Link to standard or quality goal | Outcome or next step | Owner. Include brief instructions on how to use it (e.g. update monthly; attach to evidence file).
Format: Table with one example row; short instructions at the top.
Quality: Plain language; suitable for staff who are not accreditation specialists.

Artificial Intelligence in Program Evaluation and Accreditation

Applying AI Responsibly in Indigenous Health Systems

Artificial intelligence (AI) and machine learning (ML) are increasingly relevant to accreditation and quality improvement processes. While AI is often discussed in clinical or technical terms, its practical value for accreditation coordinators lies in its ability to enhance data analysis, strengthen reporting, and support evidence-based decision-making.

In Artificial Intelligence in Program Evaluation: Insights and Applications (Shapiro & Lam, 2024), six practical approaches are outlined for integrating AI into evaluation workflows. Although written for program evaluators broadly, the framework translates directly to accreditation contexts, particularly in Indigenous health systems where accountability, documentation quality, and continuous improvement are central.

Citation
Shapiro, S., & Lam, V. (2024). Artificial intelligence in program evaluation: Insights and applications. Canadian Journal of Program Evaluation, 39(2), 382–391. https://doi.org/10.3138/cjpe-2024-0027

The Six AI-Enabled Approaches — Adapted for Accreditation

1. Identifying Patterns in Health and Service Data

AI systems can analyze large datasets to uncover trends, correlations, and outliers that may not be visible through manual review. For accreditation coordinators, this might mean tracking changes in chronic disease indicators over time, identifying service gaps across communities, or detecting anomalies in reporting or compliance data. In remote or resource-constrained settings, this capability can surface quality risks earlier and support proactive intervention.

2. Predicting Future Outcomes

Predictive models can estimate likely future outcomes based on historical trends. In accreditation settings, that could include forecasting program completion rates, estimating future service demand, or predicting risk of performance shortfalls prior to review cycles. Predictive analytics does not replace judgment; it supports strategic planning and resource allocation decisions.

3. Discovering Areas for Improvement

AI can pinpoint where performance is strongest and where improvement is needed—for example, identifying which training modules correlate with higher staff retention, detecting service areas with higher client dropout, or highlighting units where documentation compliance is inconsistent. This strengthens quality improvement planning and aligns directly with accreditation standards that require evidence-informed adjustments.

4. Generating Visualizations and Dashboards

AI-assisted visualization tools can transform complex data into clear, interpretable graphics. For accreditation leaders, dashboards can display real-time quality indicators, monitor service delivery metrics, and support transparent reporting to leadership and community. Clear visual reporting strengthens stakeholder engagement and improves communication during accreditation reviews.

5. Automating Routine Data Tasks

AI can automate repetitive tasks such as data cleaning, sorting and categorizing responses, basic statistical summaries, and preliminary report drafting. Automation reduces administrative burden and allows staff to focus on interpretation, engagement, and strategy.

6. Building Intuitive Monitoring Systems

AI tools increasingly support interactive dashboards that allow non-technical users to filter results dynamically, explore trends by community or service type, and track indicators in real time. For Indigenous health organizations, this can improve transparency and internal accountability while maintaining community oversight.

Governance, Ethics, and Indigenous Context

The original article emphasizes that AI adoption must be accompanied by careful attention to ethics and bias. In Indigenous health contexts, this extends further to data sovereignty, community governance of information systems, transparency in algorithmic decision-making, and cultural safety in system design.

AI tools must operate within established Indigenous data governance frameworks and align with principles such as OCAP®. Implementation should follow governance, not precede it.

Moving Forward

The integration of AI into program evaluation and accreditation should be incremental, transparent, community-informed, and aligned with quality standards. When implemented responsibly, AI can support more robust, evidence-based decision-making and contribute to improved service delivery outcomes across Indigenous health systems.

Read the full article (CJPE) Presentation slides

AI safety

AI can support quality and efficiency, but adoption in health care must be guided by safety, equity, and governance. This tab summarizes key risks, established frameworks, and practical steps—with attention to equity and First Nations data sovereignty—so accreditation leaders can assess and document responsible use.

Risks and harms

Recognized risks of health AI include:

  • Clinical errors and “hallucination”: AI can produce wrong, incomplete, or plausible-sounding false information. Used without verification, summaries or decision support can contribute to misdiagnosis or inappropriate care.
  • Bias and inequity: Models trained on non-representative data can perform poorly or reinforce bias (e.g. by race, ethnicity, geography, or language). For First Nations and other historically underserved populations, this can worsen existing health inequities.
  • Privacy and data sovereignty: Sending patient or community data to third-party AI services can violate privacy law and Indigenous data governance (e.g. OCAP®). Data may be retained, reused, or used to train models without community consent.
  • Automation bias and over-reliance: Staff may trust AI output without checking, especially under time pressure. Over-reliance increases the chance that errors or bias go unnoticed.
  • Documentation drift: AI-generated notes that sound fluent but don’t match what happened can undermine record quality, accountability, and accreditation evidence.
  • Access and equity: If only some organizations or populations can access or benefit from AI, gaps in quality and outcomes can widen.

Frameworks and guidelines

These frameworks help organizations turn principles into practice. Use them to shape policy, procurement, and monitoring.

Equity and First Nations context

Safety and equity are linked. For First Nations health organizations, responsible AI adoption should include:

  • Indigenous data sovereignty: Align with OCAP® and regional governance. Data should be held and controlled by First Nations; avoid sending identifiable or aggregate community data to third-party AI without clear agreements and consent.
  • Representation and bias: Ask whether training data and evaluation include First Nations populations and contexts. If not, assume higher risk of poor performance or bias in your setting.
  • Indigenous-led design and oversight: Where possible, involve community and leadership in deciding what AI is used for, how it is evaluated, and who is accountable. General frameworks often lack detail for cultural safety—local input is essential.
  • Literacy and consent: Staff and communities should understand what AI is doing, what data it uses, and how it supports (or does not support) care. Consent and transparency build trust and support accreditation expectations.

Best practices (operational)

  • Human in the loop: Use AI to assist, not replace, professional judgment. A qualified person should review and approve AI-generated content before it becomes part of the record or a clinical decision.
  • Verify before use: Treat every AI output as a draft. Check facts, numbers, and recommendations against your own data and policies before sharing or acting.
  • Align with data governance: Use only tools and vendors that comply with your privacy and Indigenous data governance requirements. Prefer solutions that keep data on your systems or in approved jurisdictions.
  • Start with low-risk uses: Pilot AI for lower-stakes tasks (e.g. drafting summaries, outlines, checklists) before high-stakes clinical or accreditation documentation.
  • Train staff: Ensure staff know when they are using AI, how to prompt and verify, and that they are responsible for final outputs.
  • Monitor and audit: Track where AI is used, sample outputs for quality and bias, and correct issues. Document this for internal improvement and accreditation.
  • Procurement and contracts: Require transparency (how data is used, where it is stored), accountability, and the right to audit. Align contracts with Indigenous data sovereignty where applicable.
For accreditation AI should strengthen your ability to demonstrate quality, safety, and continuous improvement—not create new risks. Document how you govern, verify, and limit AI use so reviewers can see that adoption is responsible and aligned with standards.

Further reading