R-009-001 Implementation Plan: Visiba Integration Enhancement

Executive Summary

Following our meeting on September 25, 2025, where Oscar from Visiba Care demonstrated the Red Robin triage engine and its integration with Legit.Health, we have identified significant opportunities to enhance the value delivered to Visiba Care's customers.

Key Findings

• Visiba Care has built a sophisticated Bayesian network-based triage system that effectively integrates Legit.Health's diagnostic support capabilities
• The current integration successfully uses the diagnosis probability distribution to inform Red Robin's decision matrix
• Several high-value API features remain untapped, representing significant opportunities to reduce engineering workload and increase customer value

Objective

This document provides a strategic roadmap for Visiba Care to maximize the return on their Legit.Health integration across four implementation phases.

Current Integration Assessment

What Works Well

Based on the September 2025 demonstration, the current integration demonstrates:

Capability	Implementation Status	Assessment
Diagnosis probability distribution	Fully integrated	Excellent
Bayesian network integration	Fully integrated	Sophisticated
Condition mapping (~100 diagnoses)	Manually maintained	Functional but labor-intensive
Confusion matrix adjustments	Implemented	Good error handling
Image capture workflow	Integrated	Effective

Understanding the Three Capability Types

The API provides three fundamentally different types of clinical intelligence, each serving a distinct purpose. These are often conflated, but understanding the distinction is essential for maximizing the integration's value.

Diagnosis

What does this person have?

A classification neural network identifies the most probable conditions from a dermatological image. This is what Visiba Care currently integrates: the probability distribution across conditions that feeds into Red Robin's Bayesian network. Diagnosis answers the question: What might this condition be?

Detected conditions (Top-5)

1. Generalised pustular psoriasis77.03%
2. Pustular psoriasis2.31%
3. Pemphigus0.95%
4. Zoster0.48%
5. Cutaneous tuberculosis0.48%

---

• Generalised pustular psoriasis
• Pustular psoriasis
• Pemphigus
• Zoster
• Cutaneous tuberculosis
• Other

Probability

0102030405060708090100

Example output of a report generated by Legit.Health

Clinical Indicators

What should I do about it?

Derived metrics that trigger workflows and provide actionable guidance for clinicians. Unlike diagnosis, which identifies what something is, indicators answer: What should I do about it? Examples include malignancy suspicion (flagging potentially cancerous lesions), urgent referral (cases requiring attention within 48 hours), and high priority referral (cases needing specialist review within 7-15 days). These are not diagnoses—they are call-to-action signals designed to support clinical decision-making regardless of the specific condition.

Malignancy suspicion

Suspicion of malignancy

62%

Predicted by the algorithm

Referral

• High-priority referral (15 days)Low (5.82%)That is, there is a low probability that the patient will need care in the next 15 days.
• Urgent referral (48 hours)
  High (35.07%)
  That is, there is a high probability that the patient will need care in the next 48 hours.
  Create Dermatology Referral ➤

Severity Measurement

How much of it is there?

Quantifies the degree of disease involvement rather than identifying the condition itself. A diagnosis tells you a patient has psoriasis; severity measurement tells you they have moderate-to-severe psoriasis with a PASI score of 12.5. This is achieved by measuring the intensity of individual clinical signs—erythema, induration, desquamation—and computing standardized scores using clinically validated instruments like PASI, SCORAD, IHS4, or EASI. Severity measurement answers: How bad is it?

Severity gauge

Severity Score

Automatic Psoriasis Area and Severity Index

Mild

Evolution

Severity
(PASI)

Time

Identified Opportunities

The following API capabilities are currently not utilized but offer substantial value. They span all three capability types described above:

Feature	Current Status	Potential Impact
Malignancy suspicion score	Not used	High: Replace manual benign/malignant mapping
Entropy (confidence indicator)	Not used	High: Enable safer triage decisions
Urgent referral indicator	Not used	High: Automate 48-hour urgency flagging
High priority referral indicator	Not used	High: Automate 7-15 day referral needs
Severity scoring (PASI, SCORAD, etc.)	Not used	High: Enable clinical treatment decisions
Image quality score	Not used	Medium: Improve AI performance feedback
Visual signs (erythema, desquamation, etc.)	Not used	Medium: Complement chat-based questions

Key Challenges Identified

1. Difficulty evidencing impact: The optional nature of photos makes it challenging to quantify AI value
2. Manual mapping workload: Maintaining condition mappings requires ongoing engineering effort
3. Limited HCP visibility: Healthcare professionals see minimal AI output, reducing perceived value
4. Customer value narrative: Sales teams need clearer materials to communicate the AI benefit

Strategic Recommendations

Phase 1: Quick Wins (Weeks 1-4)

Objective

Implement changes that deliver immediate value with minimal development effort.

Malignancy Suspicion Indicator

Challenge

Currently, Visiba Care manually maps each condition to benign or malignant categories.

Solution

Replace manual mapping with the malignancy indicator from the API.

API Response: Clinical Indicator

{
  "clinicalIndicator": {
    "malignancy": 0.15 // Value between 0 and 1
  }
}

Recommended Thresholds

Score Range	Risk Level	UI Display	Suggested Action
0-15%	Low	Green	Standard triage flow
16-30%	Moderate	Orange	Flag for clinical review
31-100%	High	Red	Prioritize specialist referral

Business Value

• Eliminates manual mapping maintenance for benign/malignant classification
• Provides consistent, validated risk assessment
• Stronger regulatory compliance story: "AI uses a validated malignancy suspicion model"

Example UI

Malignancy suspicion

Suspicion of malignancy

62%

Predicted by the algorithm

Entropy (Confidence Indicator)

Challenge

Triage decisions currently don't account for AI prediction confidence.

Solution

Implement the entropy indicator to flag low-confidence cases.

API Response: Performance Indicator

{
  "performanceIndicator": {
    "entropy": 0.25 // Normalized 0-1, lower = more confident
  }
}

Recommended Thresholds

Entropy (0-100)	Certainty Level	UI Display	Suggested Action
0-20	Very High	Green	High confidence in AI output
21-40	High	Green	Reliable for triage
41-60	Moderate	Yellow	Consider with caution
61-80	Low	Orange	Flag for manual review
81-100	Very Low	Red	Treat as inconclusive

Business Value

• Enables nuanced, safer triage decisions
• Strong regulatory narrative: "System behavior adapts based on model confidence"
• Reduces risk of over-reliance on uncertain predictions

Example UI

Entropy

🚩

The normalized entropy value is high (69%), meaning the algorithm has low certainty about its analysis. Please keep this in mind when interpreting the results.

Improved Imaging Instructions

Challenge

Current instructions are generic; image quality varies significantly.

Solution

Implement differentiated instructions based on lesion type.

For Localized Lesions (moles, pigmentation)

Step	Instruction	Rationale
1	Capture the photo with your mobile phone	Standard capture
2	Center the photo on the lesion	Ensures lesion is focal point
3	Take up to three images of the same lesion	Multiple angles improve accuracy
4	Crop the image	Removes background noise

For Other Conditions (rashes, infections, inflammatory)

Step	Instruction	Rationale
1	Capture the whole affected area	Broader context needed
2	Only send one dermatological problem at a time	Prevents confusion in analysis
3	Avoid covering the area with clothing/hair	Ensures visibility

Implementation Note

Since Red Robin already asks whether symptoms are localized or widespread, this information can trigger the appropriate instruction set.

Business Value

• Improved image quality leads to better AI performance
• Clear user guidance improves experience
• Active guidance shows commitment to quality outcomes

Example UI: Problem Type Selection

Select problem type

Type of issue

Select the type of issue

Localized Lesion

For example, a mole or a pigmentation.

Something else

Any problem that is not on a fixed spot.

Instructions for Localized Lesions

Step 1: Capture the photo with your mobile phone

Step 2: Center the photo on the lesion

Step 3: Take up to three images of the same lesion

Step 4: Crop the image to remove background noise

Phase 2: Enhanced Triage (Weeks 5-8)

Objective

Leverage additional API indicators to strengthen triage accuracy.

Referral Urgency Indicators

Solution

Integrate urgentReferral and highPriorityReferral indicators.

API Response: Clinical Indicators

{
  "clinicalIndicator": {
    "urgentReferral": 0.05, // 0-48 hours
    "highPriorityReferral": 0.22 // 7-15 days
  }
}

Recommended Thresholds

Indicator	Score Range	Action	UI Element
Urgent Referral	31-100%	Immediate escalation	Red alert, "Urgent: 48-hour action required"
Urgent Referral	16-30%	Clinical review needed	Orange warning
Urgent Referral	0-15%	Standard flow	No special flag
High Priority Referral	31-100%	Schedule within 7-15 days	Orange priority flag
High Priority Referral	16-30%	Consider priority scheduling	Yellow note
High Priority Referral	0-15%	Routine scheduling	No special flag

Integration Approach

These indicators can either:

• Replace internal urgency calculations (simpler approach)
• Complement existing Bayesian network output (recommended for validation period)

Business Value

• Reduces risk of missing urgent cases
• Provides evidence-based urgency assessment
• Enables automated routing to appropriate care pathways

Example UI

Referral

• High-priority referral (15 days)Low (5.82%)That is, there is a low probability that the patient will need care in the next 15 days.
• Urgent referral (48 hours)
  High (35.07%)
  That is, there is a high probability that the patient will need care in the next 48 hours.
  Create Dermatology Referral ➤

Top-5 Predictions Display for HCPs

Challenge

HCPs currently see limited AI output, reducing perceived value.

Solution

Display the top diagnostic predictions to healthcare professionals.

API Response: Conclusions Array

{
  "conclusions": [
    {
      "code": { "coding": [{ "code": "EA90", "display": "Psoriasis" }] },
      "probability": 0.72
    },
    {
      "code": { "coding": [{ "code": "EA80", "display": "Atopic dermatitis" }] },
      "probability": 0.15
    }
    // ... top 5
  ]
}

Display Options

Option	Description	Best For
Probability bars	Horizontal bars showing relative likelihood	Quick visual assessment
Percentage list	Text list with percentages	Detailed review
Confidence threshold	Only show predictions above X%	Reduced noise

Recommended Approach

• Show top 5 predictions with probability bars
• Apply a minimum threshold (e.g., 5%) to avoid cluttering with unlikely conditions
• Include the entropy/confidence indicator alongside

Business Value

• HCPs see concrete AI value in daily practice
• Enables structured feedback collection
• Differentiates Red Robin from competitors

Image Quality Validation

Solution

Display image quality score to guide users toward better captures.

API Response: Image Quality

{
  "imagingAnalysis": {
    "mediaValidity": {
      "quality": {
        "score": 75 // 0-100
      }
    }
  }
}

Recommended Thresholds

Score	Quality	UI Feedback	Action
81-100	Excellent	Green checkmark	Proceed with confidence
61-80	Good	Green checkmark	Proceed
41-60	Fair	Yellow warning	Consider retaking
21-40	Poor	Orange warning	Recommend retaking
1-20	Bad	Red warning	Request new image

Business Value

• Improves overall image quality in the system
• Sets clear expectations with users
• Provides basis for performance analytics

Example UI

Metadata

Report information

Information about the report.

• TimestampFeb 01, 2024, 10:06 PM
• Analysis performed in0.65 seconds
• Image modalityDermatoscopic
• Visual image quality82%

Phase 3: Severity and Monitoring (Weeks 9-12)

Objective

Introduce severity scoring capabilities that transform the integration from diagnostic support into clinical treatment decision support.

Severity Scoring Integration

Context

While diagnosis tells clinicians what condition a patient has, severity scoring tells them how bad it is—and this distinction drives treatment decisions. A patient with mild psoriasis may need only topical therapy, while severe psoriasis requires systemic treatment or biologics. Without severity data, clinicians must assess this manually, which is time-consuming and subjective.

Severity scoring systems like PASI (Psoriasis Area and Severity Index) and SCORAD (Scoring Atopic Dermatitis) are standardized, clinically validated instruments used worldwide to quantify disease severity. They are the foundation of treatment guidelines, clinical trial endpoints, and reimbursement criteria. By integrating these scores, Red Robin can provide actionable clinical intelligence that directly informs the treatment pathway—not just the referral decision.

Supported Scoring Systems

Condition	Scoring System	API Code	Output Range
Psoriasis	PASI	apasi	0-72
Atopic Dermatitis	SCORAD, EASI	scorad	0-103
Alopecia	SALT	salt	0-100
Hidradenitis Suppurativa	IHS4, HiSCR	ihs4	Variable

And many more. See the /questionnaires?pathology={ICD-11} endpoint for full details.

API Response Structure

API Response: Severity Score

{
  "patientEvolution": {
    "apasi": {
      "score": {
        "value": 12.5,
        "interpretation": {
          "category": "Moderate",
          "intensity": 2 // 0=None, 1=Mild, 2=Moderate, 3=Severe
        }
      }
    }
  }
}

Implementation Approach

1. When top-1 diagnosis probability exceeds threshold (e.g., 40%), check if severity scoring is available for that condition
2. Use the /questionnaires?pathology={ICD-11} endpoint to determine if additional questions are needed
3. Display severity result to HCP with interpretation

Business Value

• Treatment pathway guidance: Severity scores directly map to treatment guidelines, enabling appropriate therapy recommendations
• Monitoring over time: Track whether a patient is improving, stable, or worsening across visits—essential for chronic disease management
• Clinical trial alignment: Pharmaceutical customers require standardized severity measures; this positions Red Robin as a clinical trial-ready platform
• Reimbursement support: Many payers require documented severity scores to approve biologic or systemic treatments

Example UI

Severity gauge

Severity Score

Automatic Psoriasis Area and Severity Index

Mild

Visual Signs Display

Solution

Surface individual clinical sign assessments to complement chat-based questions.

API Response: Clinical Signs

{
  "item": {
    "erythema": { "value": 2, "text": "Erythema" },
    "desquamation": { "value": 3, "text": "Desquamation" },
    "induration": { "value": 1, "text": "Induration" }
  }
}

Display Format

Clinical Sign	Value (0-4)	Interpretation
Erythema	2	Moderate redness
Desquamation	3	Significant scaling
Induration	1	Mild thickening

Integration Opportunity

Cross-reference AI-detected visual signs with patient-reported symptoms from the chat to validate consistency.

Example UI

Intensity of clinical signs

Desquamation

Moderate (2)

---

Erythema

Moderate (2)

---

Induration

Mild (1)

---

Affected area

30% (2)

Evolution Tracking

Solution

For patients with multiple visits, track severity changes over time.

Business Value

• Demonstrates treatment effectiveness
• Supports clinical decision-making for treatment adjustments
• Valuable for pharmaceutical and clinical trial customers

Example UI

Evolution

Severity
(PASI)

Time

Phase 4: Value Communication (Weeks 13-16)

Objective

Strengthen the commercial value proposition and customer engagement.

HCP Dashboard Enhancements

Proposed Elements

Element	Description	Value
AI Confidence Badge	Visual indicator of prediction certainty	Builds appropriate trust
Top Diagnoses Panel	Collapsible section showing top 5 predictions	Quick reference
Severity Gauge	Visual representation of condition severity	Clinical decision support
Image Quality Indicator	Shows quality score for submitted images	Quality feedback loop

Sales Enablement Materials

1. Value Slides (2-3 slides): For Visiba Care's sales decks

   • What AI provides in the Red Robin context
   • How AI complements rules-based triage
   • Example scenarios from primary care

2. Customer-Facing Brief (1 page):

   • How AI reduces uncertainty in primary care skin assessments
   • How it identifies high-risk cases
   • How it improves the patient journey

3. FAQ Document:

   • Non-technical answers for customer meetings
   • Handling objections about AI reliability
   • Evidence and validation references

Implementation Timeline

Phase Summary

Phase	Duration	Key Deliverables	Business Outcome
Quick Wins	Weeks 1-4	Malignancy indicator, entropy display, improved instructions	Reduced maintenance, safer triage
Enhanced Triage	Weeks 5-8	Referral indicators, top-5 display, image quality	Better accuracy, HCP visibility
Severity	Weeks 9-12	Severity scoring, visual signs, evolution tracking	Chronic condition support
Communication	Weeks 13-16	Dashboard, sales materials, impact evaluation	Stronger value proposition

Technical Reference

API Keys Quick Reference

Feature	Endpoint	Response Path	Type
Diagnosis probabilities	/diagnosis-support	conclusions[].probability	Float 0-1
ICD-11 codes	/diagnosis-support	conclusions[].code.coding[].code	String
Malignancy	/diagnosis-support	clinicalIndicator.malignancy	Float 0-1
Urgent referral	/diagnosis-support	clinicalIndicator.urgentReferral	Float 0-1
High priority referral	/diagnosis-support	clinicalIndicator.highPriorityReferral	Float 0-1
Entropy	/diagnosis-support	performanceIndicator.entropy	Float 0-1
Image quality	/diagnosis-support	imagingAnalysis.mediaValidity.quality.score	Integer 0-100
Severity score	/severity	patientEvolution.[system].score.value	Varies
Severity interpretation	/severity	patientEvolution.[system].score.interpretation	Object
Clinical signs	/severity	item.[signName].value	Integer 0-4

Helper Endpoints

Endpoint	Purpose	Use Case
/body-sites	Standardized body location codes	Consistent anatomical reference
/clinical-signs	Visual sign translations	Multilingual support
/questionnaires?pathology={ICD-11}	Scoring system availability	Determine if questionnaire needed

Expected Benefits

For Visiba Care Engineering

Benefit	Impact	Phase
Reduced mapping maintenance	Eliminate manual benign/malignant classification	1
Simplified urgency logic	Leverage pre-calculated referral indicators	2
ICD-11 code alignment	Reduce mapping updates when device version changes	1-2

For Visiba Care Commercial

Benefit	Impact	Phase
Stronger value narrative	"AI provides malignancy assessment, confidence scores, and severity monitoring"	All
Differentiation	Features competitors may not offer	2-3
Customer retention	Enhanced capabilities keep customers engaged	3-4

For End Customers (Healthcare Providers)

Benefit	Impact	Phase
Safer triage	Confidence and urgency indicators flag edge cases	1-2
Better informed decisions	Top-5 predictions visible to HCPs	2
Chronic condition support	Severity scoring and evolution tracking	3

Commercial Considerations

Pricing Model Alignment

The current per-capita pricing model aligns well with Visiba Care's customer pricing structure. Enhanced features can be positioned as:

• Standard tier: Current integration + Phase 1 features (malignancy, entropy)
• Professional tier: + Phase 2 features (referral indicators, HCP dashboard)
• Enterprise tier: + Phase 3 features (severity scoring, evolution tracking)

ROI Framework

Category	Metric	Estimated Impact
Engineering efficiency	Hours saved on mapping maintenance	40-60 hrs/year
Development velocity	New feature integration time	30-50% faster
Customer retention	Reduced churn from enhanced value	To be measured
New customer acquisition	Competitive differentiation	Qualitative

Joint Value Demonstration

We propose a joint case study or validation study that:

• Documents improvement in triage accuracy
• Quantifies time savings for HCPs
• Provides evidence for customer sales conversations

Next Steps

Immediate Actions (Week 0)

Action	Owner	Target Date
Prioritization meeting	Visiba Care + Legit.Health	TBD
Technical deep-dive on Phase 1	Engineering teams	TBD
Access to additional API documentation	Legit.Health	TBD

Phase 1 Kick-off

Action	Owner	Target Date
Malignancy indicator implementation	Visiba Care Engineering	Week 2
Entropy display design review	Visiba Care UX + Legit.Health	Week 1
Imaging instructions content finalization	Joint	Week 3

Ongoing Support

Legit.Health commits to providing:

• Technical support during implementation
• Access to updated API documentation
• Sales enablement material drafts
• Impact evaluation consultation

Signature meaning

The signatures for the approval process of this document can be found in the verified commits at the repository for the QMS. As a reference, the team members who are expected to participate in this document and their roles in the approval process, as defined in Annex I Responsibility Matrix of the GP-001, are:

• Author: Team members involved
• Reviewer: JD-003, JD-004
• Approver: JD-001