AI Wound Assessment & Support System for Care Homes

Skin and wound-related quality indicators and inspection findings in Ontario long-term care homes continue to challenge clinical teams and operators, underscoring how much is at stake when assessments are inconsistent or delayed. As healthcare leaders navigate increased regulatory scrutiny and resident acuity, the question isn't whether we need better tools—it's how we implement them responsibly while maintaining clinical excellence.

The Challenge: Staging Accuracy Matters More Than Ever

Let's address the elephant in the room: wound staging errors are common, costly, and consequential.

Studies show that pressure injury staging accuracy and concordance between assessors can be highly variable, with interrater agreement in many studies falling well below 100% and frequently in the moderate range (Beeckman et al., 2018; Kottner et al., 2019). The implications extend far beyond documentation:

• Incorrect staging leads to inappropriate treatment selection

• Under-staging delays necessary interventions

• Over-staging wastes resources on unnecessary treatments

• Inconsistent staging undermines quality metrics and regulatory compliance

When a Stage 2 pressure injury is misidentified as Stage 1, we might apply a less protective dressing such as a simple film instead of a foam dressing designed for exudate management—potentially delaying optimal healing. When unstageable wounds are incorrectly classified, opportunities for appropriate debridement and offloading strategies can be missed.

The clinical stakes are too high for guesswork.

Beyond Pressure Injuries: Comprehensive Wound Assessment

During our pilot program, an experienced RN provided feedback that reshaped our entire approach: "This would be amazing if it worked for all wounds, not just pressure injuries."

She was absolutely right. In long-term care, clinicians manage far more than pressure injuries. They routinely encounter:

Diabetic foot ulcers requiring offloading and glycemic control

Venous leg ulcers needing compression therapy when arterial supply is adequate

Arterial ulcers requiring vascular assessment and often avoidance of compression

Skin tears demanding gentle, non-adhesive approaches

Moisture-associated skin damage (MASD), which is often confused with pressure injuries

Surgical wounds with unique healing timelines and infection risks

Each wound type requires different assessment criteria, classification systems, and treatment approaches. Applying a pressure injury protocol to a venous ulcer, or vice versa, leads to poor outcomes and frustrated clinical teams.

Accurate wound classification is the foundation of everything that follows.

The Maker-Checker Model: AI Precision Meets Clinical Judgment

The maker-checker model recognizes a reality that is increasingly well described in the literature: AI can process visual patterns and structured rules with consistency, but only licensed professionals possess the contextual judgment to validate and apply those findings safely (Topol, 2019; Rajkomar et al., 2019).

The "Maker" Phase (AI Analysis)

When a nurse captures a wound photograph, the AI can perform structured analysis across several domains as a clinical decision-support tool, not an autonomous decision-maker.

Wound Type Classification

The model can be designed to:

• Differentiate pressure injuries, diabetic foot ulcers, venous ulcers, and arterial ulcers based on visual and contextual features

• Identify skin tears vs. partial-thickness traumatic injuries

• Distinguish MASD from true pressure damage by distribution and exposure patterns

• Flag potentially mixed etiologies that require more complex assessment

Staging Precision (Pressure Injuries)

For pressure injuries, widely accepted criteria from NPUAP/EPUAP/PPPIA (now NPIAP/EPUAP/PPPIA) define (EPUAP/NPIAP/PPPIA, 2025 https://internationalguideline.com/the-international-guideline.):

Stage 1: Non-blanchable erythema on intact skin

Stage 2: Partial-thickness skin loss with exposed dermis, without slough

Stage 3: Full-thickness skin loss with visible fat; slough may be present but does not obscure depth

Stage 4: Full-thickness skin and tissue loss with exposed or directly palpable fascia, muscle, tendon, ligament, cartilage, or bone

Unstageable: Full-thickness skin and tissue loss in which the base is obscured by slough or eschar

Deep Tissue Injury: Persistent non-blanchable deep red, maroon, or purple discoloration, sometimes with blood-filled blister

For other wounds, classification schemes such as partial vs. full-thickness skin loss, diabetic foot ulcer grading, or CEAP classification for chronic venous disease may be relevant.

Wound Characteristics Assessment

AI models can be trained to extract:

• Wound bed composition estimates (granulation, slough, eschar)

• Apparent exudate level and character, which are critical for dressing selection

• Periwound skin condition (maceration, erythema, induration)

• Presence of undermining or tunneling

• Visual cues suggestive of local infection or heavy biofilm burden (for clinician confirmation)

The "Checker" Phase (Clinical Validation)

Professional nursing judgment remains essential. The RN or NP reviews AI output and:

✓ Validates staging at the bedside, including blanching vs. non-blanching erythema and palpation findings

✓ Applies clinical context such as recent surgery, immobility, or moisture and shear patterns

✓ Considers resident-specific factors like frail skin, sensory impairment, and pain

✓ Integrates systems assessment: nutrition, perfusion, infection risk, comorbidities

✓ Corrects AI errors with minimal friction and documents the final professional assessment

The result is AI-supported consistency combined with clinical wisdom, which can support better and more defensible outcomes when embedded responsibly.

Protocol Selection: Matching Evidence to Individual Wounds

Accurate staging and classification are only the first step; protocol selection is where many facilities struggle with standardization. Consider a Stage 3 sacral pressure injury—facilities may have protocols from multiple manufacturers and internal guidelines that differ in exudate management, infection control strategies, and dressing sequence.

The AI-Assisted Approach

A well-governed system can maintain the facility's active protocol library with:

• Manufacturer specifications and indications for use

• Mapping to wound types and stages

• Version control and review dates for governance

• Crosswalks to CNO standards and FLTCA 2021 expectations where applicable

When generating recommendations, the AI can:

1. Filter protocols applicable to the identified wound type and stage

2. Match wound characteristics (exudate level, slough, infection risk) to protocol criteria

3. Consider local formulary availability and inventory

4. Present options aligned with evidence-based guidelines and facility policy for clinician review

For a Stage 3 sacral pressure injury with moderate exudate and partial slough, such a system can highlight:

• All protocols covering Stage 3 pressure injuries

• Dressing families appropriate for moderate exudate and slough management (e.g., foam with or without alginate, moisture-retentive dressings to support autolytic debridement)

• Suggested change frequencies adjusted to exudate level and periwound condition

• Triggers for nutritional support and pressure redistribution interventions

The nurse reviews these recommendations through the lens of resident comfort, prior dressing tolerance, care team capability, cost-effectiveness, and skin fragility.

Product Selection: The Right Dressing at the Right Time

Evidence and guidelines emphasize that incorrect dressing choice can delay healing, increase pain, and contribute to complications. Feedback from pilot users that "the dressing suggestions weren't correct for the wound characteristics" is exactly the type of signal needed to refine algorithms and align them more closely with evidence-based dressing logic.

Evidence-Based Dressing Selection Logic

The decision-support logic described below is broadly consistent with current guidance, though specific product choices should always be interpreted in light of manufacturer indications and local policies (Wounds Canada, 2020; WOCN, 2016).

For DRY wounds (minimal exudate):

• ❌ Avoid highly absorbent dressings that may further desiccate tissue

• ✅ Consider hydrogels to donate moisture when appropriate

• ✅ Consider hydrocolloids or other moisture-retentive dressings when there is no infection and low exudate

• ✅ For budget-conscious care plans, saline-moistened gauze under an occlusive layer can be used where consistent with guidelines and infection risk

For HEAVY exudate:

• ❌ Avoid low-absorbency films or some hydrocolloids alone, which may leak and macerate surrounding skin

• ✅ Use super-absorbent dressings, alginates, or hydrofiber dressings with appropriate secondary layers

For SLOUGH presence:

• ❌ Avoid long-term use of occlusive dressings without a clear debridement plan

• ✅ Use autolytic debridement strategies such as hydrogels or some hydrocolloids where appropriate and infection is controlled

• ✅ Consider enzymatic debriding agents or sharp/surgical debridement per clinical judgment and local policy

For suspected INFECTION or high biofilm risk:

• ❌ Avoid relying solely on standard foam or hydrocolloid dressings without antimicrobial strategy when there are signs of local infection

• ✅ Use antimicrobial dressings (silver, iodine, PHMB, or medical honey) consistent with guidelines and manufacturer instructions, alongside systemic antibiotics when indicated

For FRAGILE periwound skin:

• ❌ Avoid aggressive adhesives that can cause medical adhesive-related skin injury

• ✅ Prefer silicone-bordered dressings, non-adherent contact layers, and barrier products to protect periwound skin

For TUNNELING or UNDERMINING:

• ❌ Avoid flat dressings that fail to fill dead space and may contribute to abscess formation

• ✅ Use appropriate cavity fillers such as alginate rope, foam strips, or hydrofiber ribbon, with correct packing technique

The system can match your facility's product inventory to these criteria so that every recommendation remains both clinically grounded and operationally realistic.

Real-World Impact: When Staging and Selection Are Right

The scenarios below are directionally consistent with what is documented in wound care literature: earlier correct treatment is associated with faster healing and fewer complications, though exact time frames can vary by patient and study.

Scenario 1: Correctly Identified Stage 2 vs. Misclassified Stage 1

Intervention: Immediate use of an appropriate moisture-balancing dressing such as foam for Stage 2 injuries with exudate

Outcome: Can support faster healing compared with delayed escalation. Some studies and guidelines suggest that properly managed Stage 2 pressure injuries may heal over a period of weeks, while delayed or inadequate treatment risks progression to deeper stages and complications.

Scenario 2: Properly Differentiated MASD from Pressure Injury

Intervention: MASD responds best to moisture management, barrier protection, and addressing incontinence or perspiration, rather than pressure redistribution alone (Gray et al., 2011; Beeckman et al., 2015)

Outcome: Correct classification avoids misreporting pressure injury incidence and reduces unnecessary use of pressure-redistributing equipment

Scenario 3: Venous Ulcer Correctly Classified vs. Arterial Assumption

Intervention: Venous leg ulcers are typically managed with compression when arterial perfusion is adequate, while significant arterial insufficiency is a contraindication to high-level compression (O'Donnell et al., 2014; WUWHS, 2016)

Outcome: Early correct classification can reduce risk of ischemic complications and optimize healing trajectory

The key "time dividend" here relates to time to correct treatment and time saved by preventing complications rather than documentation speed alone.

Quality Metrics: From Subjective to Objective Assessment

For Quality Improvement leads and Directors of Care, AI-assisted assessment can support more objective and auditable measurement frameworks when implemented with robust governance. Potential benefits include:

Reduced inter-rater variability in staging and wound description, with structured fields and photo-based review

Clear audit trails of staging rationale and changes over time

Linkage of dressing and product selection to documented wound characteristics and approved protocols

More consistent outcome tracking through serial standardized photos and measurements, enabling earlier identification of non-healing wounds

These data can inform protocol refinement, formulary optimization, and regulatory survey defense.

Clinical Governance: Protecting Professional Accountability

Regulatory and ethical guidance on AI in healthcare emphasizes that licensed professionals must retain accountability for clinical decisions (WHO, 2021; FDA, 2021). In this model, the AI:

Does NOT:

• Make treatment decisions independently

• Override nursing judgment

• Replace hands-on assessment

• Determine plans of care

• Communicate with physicians autonomously

Does:

• Analyze visual data consistently

• Reference protocols systematically

• Match products to wound characteristics logically

• Flag inconsistencies for clinical review

Every AI suggestion should appear explicitly as a recommendation requiring professional validation, with easy mechanisms for clinicians to override or correct it. This framing makes it clear that the nurse or nurse practitioner always has the final say and that the tool exists to augment—not automate—nursing practice.

Privacy and Compliance: PHIPA-First Design

In Ontario, PHIPA sets rules for the collection, use, and disclosure of personal health information, including images used for clinical care (Government of Ontario, 2004). A PHIPA-conscious design for AI-assisted wound photography and assessment should include:

✓ Capability to perform assessments without unnecessary identifiers when clinically appropriate (data minimization)

✓ Encrypted storage and secure transmission of wound photos and associated data

✓ Robust access controls and audit trails for all access and modifications

✓ Clear policies about retention, destruction, and integration of images into the health record

✓ Alignment with CNO expectations and FLTCA 2021 requirements regarding documentation, consent, and quality of care.

These elements help ensure that technical sophistication does not come at the expense of resident privacy and trust.

Implementation Considerations for Clinical Leaders

Implementation guidance from studies on AI decision support in nursing and broader healthcare stresses change management, training, and continuous evaluation (Kelly et al., 2019; Sutton et al., 2020).

For Directors of Care and Associate Directors:

• Establish policies that define AI-assisted assessment as decision support under professional accountability standards

• Monitor staging accuracy and consistency through audits and case reviews

• Track clinical outcomes (healing rates, progression, infection) and use data to refine protocols and formularies

For Nurse Managers:

• Train staff on what the AI can and cannot do, emphasizing "review and approve" rather than "accept by default"

• Use the tool as a coaching aid for staging, product selection, and documentation consistency

• Highlight cases where clinical judgment appropriately overrode AI recommendations to reinforce professional agency

For Quality Improvement Leads:

• Establish baseline metrics before implementation (e.g., staging concordance, time-to-appropriate-treatment, product utilization patterns)

• Monitor for improvements and unintended consequences, including equity and bias concerns

• Use data to support evidence-based updates to protocols and education programs

The Bigger Picture: Clinical Excellence Through Better Tools

The shift here is from documentation-centric thinking to decision-quality-centric thinking, which aligns with wider trends in AI for clinical decision support.

Better staging → More appropriate treatment plans

Better protocol selection → Evidence-based consistency

Better product matching → Faster healing, less waste

Better consistency → Improved quality metrics

The most meaningful time savings come from getting treatment right sooner and preventing avoidable complications, not simply from charting faster.

A Vision for Wound Care Excellence

The future of wound care is unlikely to be about replacing clinical expertise with algorithms; instead, it is about equipping clinical teams with tools that:

✓ Improve staging accuracy and consistency

✓ Support evidence-based protocol selection

✓ Match products to wound characteristics systematically while respecting formulary constraints

✓ Keep professional judgment and accountability at the center

✓ Protect resident privacy under PHIPA and related regulations

✓ Enable better outcomes through more informed decisions

Because in practice, success is not measured by how many assessments are documented, but by how many wounds progress toward healing—and how confident clinicians feel about the decisions that guided that journey.

References

Beeckman, D., et al. (2015). Incontinence-associated dermatitis: moving prevention forward. Proceedings of the Global IAD Expert Panel. Wounds International.

Beeckman, D., et al. (2018). Pressure ulcer staging: inter-rater reliability and the influence of knowledge. International Wound Journal, 15(1), 13-20.

CNO documentation updated https://www.cno.org/Assets/CNO/Documents/Standard-and-Learning/Practice-Standards/documentation-in-effect-feb-2026-en.pdf

3.Staging Precision EPUAP/NPIAP/PPIA you might want to update to 2025 https://internationalguideline.com/the-international-guideline

FDA. (2021). Artificial Intelligence and Machine Learning in Software as a Medical Device.

Government of Ontario. (2004). Personal Health Information Protection Act, 2004.

Government of Ontario. (2021). Fixing Long-Term Care Act, 2021.

Gray, M., et al. (2011). Incontinence-associated dermatitis: a consensus. Journal of WOCN, 38(4), 359-370.

Kelly, C.J., et al. (2019). Key challenges for delivering clinical impact with artificial intelligence. BMC Medicine, 17, 195.

Kottner, J., et al. (2019). Pressure ulcer/injury classification today: an international perspective. J Tissue Viability, 28(4), 197-203.

O'Donnell, T.F., et al. (2014). Management of venous leg ulcers: clinical practice guidelines. Journal of Vascular Surgery, 60(2S), 3S-59S.

Rajkomar, A., et al. (2019). Machine learning in medicine. NEJM, 380, 1347-1358.

Sutton, R.T., et al. (2020). An overview of clinical decision support systems. CMAJ, 192(8), E209-E216.

Topol, E.J. (2019). High-performance medicine: the convergence of human and artificial intelligence. Nature Medicine, 25, 44-56.

WHO. (2021). Ethics and governance of artificial intelligence for health.

WOCN. (2016). Guideline for Prevention and Management of Pressure Ulcers (Injuries).

Wounds Canada https://www.woundscanada.ca/health-care-professional/publications/bpr-new

WUWHS. (2016). Compression in venous leg ulcers: consensus document.

How is your organization ensuring consistency in wound staging and treatment selection? What tools or processes have improved your clinical decision-making? I'd love to hear your perspectives in the comments.

#WoundCare #ClinicalExcellence #LongTermCare #NurseLeadership #QualityImprovement #EvidenceBasedPractice #PatientSafety #PressureInjury #NursingLeadership #HealthcareInnovation #PHIPA #OntarioLTC #CNO #FLTCA2021 #WoundsCanada #Solventum #assety