Risk Management Specialist

ISO 14971:2019 risk management implementation throughout the medical device lifecycle.

Table of Contents

• Risk Management Planning Workflow
• Risk Analysis Workflow
• Risk Evaluation Workflow
• Risk Control Workflow
• Post-Production Risk Management
• Risk Assessment Templates
• Decision Frameworks
• Tools and References

Risk Management Planning Workflow

Establish risk management process per ISO 14971.

Workflow: Create Risk Management Plan

1. Define scope of risk management activities:
   • Medical device identification
   • Lifecycle stages covered
   • Applicable standards and regulations
2. Establish risk acceptability criteria:
   • Define probability categories (P1-P5)
   • Define severity categories (S1-S5)
   • Create risk matrix with acceptance thresholds
3. Assign responsibilities:
   • Risk management lead
   • Subject matter experts
   • Approval authorities
4. Define verification activities:
   • Methods for control verification
   • Acceptance criteria
5. Plan production and post-production activities:
   • Information sources
   • Review triggers
   • Update procedures
6. Obtain plan approval
7. Establish risk management file
8. Validation: Plan approved; acceptability criteria defined; responsibilities assigned; file established

Risk Management Plan Content

Risk Acceptability Matrix (5x5)

Risk Level Actions

Risk Analysis Workflow

Identify hazards and estimate risks systematically.

Workflow: Conduct Risk Analysis

1. Define intended use and reasonably foreseeable misuse:
   • Medical indication
   • Patient population
   • User population
   • Use environment
2. Select analysis method(s):
   • FMEA for component/function analysis
   • FTA for system-level analysis
   • HAZOP for process deviations
   • Use Error Analysis for user interaction
3. Identify hazards by category:
   • Energy hazards (electrical, mechanical, thermal)
   • Biological hazards (bioburden, biocompatibility)
   • Chemical hazards (residues, leachables)
   • Operational hazards (software, use errors)
4. Determine hazardous situations:
   • Sequence of events
   • Foreseeable misuse scenarios
   • Single fault conditions
5. Estimate probability of harm (P1-P5)
6. Estimate severity of harm (S1-S5)
7. Document in hazard analysis worksheet
8. Validation: All hazard categories addressed; all hazards documented; probability and severity assigned

Hazard Categories Checklist

Analysis Method Selection

Probability Criteria

Severity Criteria

See: references/risk-analysis-methods.md

Risk Evaluation Workflow

Evaluate risks against acceptability criteria.

Workflow: Evaluate Identified Risks

1. Calculate initial risk level from probability × severity
2. Compare to risk acceptability criteria
3. For each risk, determine:
   • Acceptable: Document and accept
   • ALARP: Proceed to risk control
   • Unacceptable: Mandatory risk control
4. Document evaluation rationale
5. Identify risks requiring benefit-risk analysis
6. Complete benefit-risk analysis if applicable
7. Compile risk evaluation summary
8. Validation: All risks evaluated; acceptability determined; rationale documented

Risk Evaluation Decision Tree

Risk Estimated
      │
      ▼
Apply Acceptability Criteria
      │
      ├── Low Risk ──────────► Accept and document
      │
      ├── Medium Risk ───────► Consider risk reduction
      │   │                    Document ALARP if not reduced
      │   ▼
      │   Practicable to reduce?
      │   │
      │   Yes──► Implement control
      │   No───► Document ALARP rationale
      │
      ├── High Risk ─────────► Risk reduction required
      │   │                    Must demonstrate ALARP
      │   ▼
      │   Implement control
      │   Verify residual risk
      │
      └── Unacceptable ──────► Design change mandatory
                               Cannot proceed without control

ALARP Demonstration Requirements

Benefit-Risk Analysis Triggers

Risk Control Workflow

Implement and verify risk control measures.

Workflow: Implement Risk Controls

1. Identify risk control options:
   • Inherent safety by design (Priority 1)
   • Protective measures in device (Priority 2)
   • Information for safety (Priority 3)
2. Select optimal control following hierarchy
3. Analyze control for new hazards introduced
4. Document control in design requirements
5. Implement control in design
6. Develop verification protocol
7. Execute verification and document results
8. Evaluate residual risk with control in place
9. Validation: Control implemented; verification passed; residual risk acceptable; no unaddressed new hazards

Risk Control Hierarchy

Risk Control Option Analysis Template

RISK CONTROL OPTION ANALYSIS

Hazard ID: H-[XXX]
Hazard: [Description]
Initial Risk: P[X] × S[X] = [Level]

OPTIONS CONSIDERED:
| Option | Control Type | New Hazards | Feasibility | Selected |
|--------|--------------|-------------|-------------|----------|
| 1 | [Type] | [Yes/No] | [H/M/L] | [Yes/No] |
| 2 | [Type] | [Yes/No] | [H/M/L] | [Yes/No] |

SELECTED CONTROL: Option [X]
Rationale: [Justification for selection]

IMPLEMENTATION:
- Requirement: [REQ-XXX]
- Design Document: [Reference]

VERIFICATION:
- Method: [Test/Analysis/Review]
- Protocol: [Reference]
- Acceptance Criteria: [Criteria]

Risk Control Verification Methods

Residual Risk Evaluation

Post-Production Risk Management

Monitor and update risk management throughout product lifecycle.

Workflow: Post-Production Risk Monitoring

1. Identify information sources:
   • Customer complaints
   • Service reports
   • Vigilance/adverse events
   • Literature monitoring
   • Clinical studies
2. Establish collection procedures
3. Define review triggers:
   • New hazard identified
   • Increased frequency of known hazard
   • Serious incident
   • Regulatory feedback
4. Analyze incoming information for risk relevance
5. Update risk management file as needed
6. Communicate significant findings
7. Conduct periodic risk management review
8. Validation: Information sources monitored; file current; reviews completed per schedule

Information Sources

Risk Management File Update Triggers

Periodic Review Requirements

Risk Assessment Templates

Hazard Analysis Worksheet

HAZARD ANALYSIS WORKSHEET

Product: [Device Name]
Document: HA-[Product]-[Rev]
Analyst: [Name]
Date: [Date]

| ID | Hazard | Hazardous Situation | Harm | P | S | Initial Risk | Control | Residual P | Residual S | Final Risk |
|----|--------|---------------------|------|---|---|--------------|---------|------------|------------|------------|
| H-001 | [Hazard] | [Situation] | [Harm] | [1-5] | [1-5] | [Level] | [Control ref] | [1-5] | [1-5] | [Level] |

FMEA Worksheet

FMEA WORKSHEET

Product: [Device Name]
Subsystem: [Subsystem]
Analyst: [Name]
Date: [Date]

| ID | Item | Function | Failure Mode | Effect | S | Cause | O | Control | D | RPN | Action |
|----|------|----------|--------------|--------|---|-------|---|---------|---|-----|--------|
| FM-001 | [Item] | [Function] | [Mode] | [Effect] | [1-10] | [Cause] | [1-10] | [Detection] | [1-10] | [S×O×D] | [Action] |

RPN Action Thresholds:
>200: Critical - Immediate action
100-200: High - Action plan required
50-100: Medium - Consider action
<50: Low - Monitor

Risk Management Report Summary

RISK MANAGEMENT REPORT

Product: [Device Name]
Date: [Date]
Revision: [X.X]

SUMMARY:
- Total hazards identified: [N]
- Risk controls implemented: [N]
- Residual risks: [N] Low, [N] Medium, [N] High
- Overall conclusion: [Acceptable / Not Acceptable]

RISK DISTRIBUTION:
| Risk Level | Before Control | After Control |
|------------|----------------|---------------|
| Unacceptable | [N] | 0 |
| High | [N] | [N] |
| Medium | [N] | [N] |
| Low | [N] | [N] |

CONTROLS IMPLEMENTED:
- Inherent safety: [N]
- Protective measures: [N]
- Information for safety: [N]

OVERALL RESIDUAL RISK: [Acceptable / ALARP Demonstrated]
BENEFIT-RISK CONCLUSION: [If applicable]

APPROVAL:
Risk Management Lead: _____________ Date: _______
Quality Assurance: _____________ Date: _______

Decision Frameworks

Risk Control Selection

What is the risk level?
        │
        ├── Unacceptable ──► Can hazard be eliminated?
        │                    │
        │                Yes─┴─No
        │                 │     │
        │                 ▼     ▼
        │            Eliminate  Can protective
        │            hazard     measure reduce?
        │                           │
        │                       Yes─┴─No
        │                        │     │
        │                        ▼     ▼
        │                   Add       Add warning
        │                   protection + training
        │
        └── High/Medium ──► Apply hierarchy
                            starting at Level 1

New Hazard Analysis

Risk Acceptability Decision

Tools and References

Scripts

Risk Matrix Calculator Features:

• ISO 14971 5x5 risk matrix calculation
• FMEA RPN (Risk Priority Number) calculation
• Interactive mode for guided assessment
• Display risk criteria definitions
• JSON output for integration

References

Quick Reference: ISO 14971 Process

Related Skills

AI-Specific Risk Management (ISO 14971 + AI Risk Considerations)

AI/ML Medical Device Risk Categories

Traditional ISO 14971 hazard categories must be extended for AI/ML-based devices:

AI Risk Analysis Methodology

Step 1: AI System Characterization
        → Define intended use, user population, clinical context
        → Classify: locked algorithm vs. adaptive vs. continuously learning
        → Map to SaMD risk framework (IMDRF)

Step 2: AI-Specific Hazard Identification
        → Apply standard ISO 14971 hazard categories
        → ADD: data quality hazards, algorithmic hazards, integration hazards
        → Consider: training data representativeness, edge cases, failure modes

Step 3: AI Failure Mode Analysis
        → Extend FMEA with AI-specific failure modes:
           - False positive/negative beyond acceptable rates
           - Performance degradation over time
           - Out-of-distribution input handling
           - Feature importance shift
        → For each failure mode: determine harm pathway to patient

Step 4: AI-Specific Risk Controls
        → Confidence thresholds (reject uncertain predictions)
        → Human-in-the-loop for high-risk decisions
        → Input validation and out-of-distribution detection
        → Continuous performance monitoring with drift detection
        → Automated model retraining safeguards
        → Fail-safe modes when AI system is unavailable

Step 5: AI Risk Monitoring Plan
        → Define performance metrics and acceptable thresholds
        → Establish monitoring frequency (real-time, daily, weekly)
        → Define retraining triggers and validation requirements
        → Plan for model versioning and rollback procedures

AI Risk Acceptability Considerations

Cybersecurity Risk Integration (IEC 81001-5-1)

Health Software Cybersecurity Risk Management

IEC 81001-5-1:2021 establishes cybersecurity lifecycle requirements for health software. Integrate with ISO 14971:

Cybersecurity Threat Categories for Medical Devices

Cybersecurity FMEA Extension

Add these columns to standard FMEA for cybersecurity failure modes:

CYBERSECURITY FMEA EXTENSION

| ID | Component | Security Function | Threat | Attack Vector | Exploitability | Impact | S | O | D | RPN | Security Control |
|----|-----------|-------------------|--------|---------------|---------------|--------|---|---|---|-----|-----------------|
| CS-001 | Auth module | User authentication | Credential theft | Phishing | High (8) | Full access | 8 | 6 | 4 | 192 | MFA + session management |
| CS-002 | Data store | Data confidentiality | SQL injection | Network input | Medium (5) | Data breach | 9 | 4 | 3 | 108 | Parameterized queries + WAF |
| CS-003 | Update mechanism | Integrity | Supply chain | Compromised update | Low (3) | Malware install | 10 | 2 | 7 | 140 | Code signing + integrity verification |

Supply Chain Risk Management

Medical Device Supply Chain Risks

Supply Chain Risk Assessment Workflow

Step 1: Supply Chain Mapping
        → Identify all direct suppliers (Tier 1)
        → Map critical Tier 2 and Tier 3 suppliers
        → Document component criticality (safety-critical, quality-critical, standard)

Step 2: Supplier Risk Scoring
        → Quality risk: past performance, certification status, audit results
        → Financial risk: stability, dependency on your business
        → Geographic risk: natural disaster, political stability
        → Cyber risk: supplier's information security posture
        → Concentration risk: single-source, regional concentration

Step 3: Risk Treatment
        → Critical suppliers: quality agreements, annual audits, dual-sourcing
        → High-risk suppliers: enhanced monitoring, contingency plans
        → Medium-risk suppliers: periodic review, performance metrics
        → Low-risk suppliers: standard purchasing controls

Step 4: Ongoing Monitoring
        → Supplier scorecard tracking (quality, delivery, responsiveness)
        → Annual supplier risk reassessment
        → Trigger-based reassessment (quality event, financial change, M&A)

Post-Market Risk Monitoring Automation

Automated Signal Detection

Risk Management File Update Automation

Automated Trigger → Risk Review Decision Tree

New complaint received
    → Classify by hazard category (auto or manual)
    → Check: Known hazard?
        YES → Update frequency data → Recalculate risk level
                → Risk level changed? → Flag for risk management review
        NO  → New hazard identified → Initiate risk analysis
              → Estimate initial risk → Determine controls needed
              → Update risk management file

Trend threshold exceeded
    → Generate trend report with statistical analysis
    → Convene risk management review within 30 days
    → Update risk management file with new probability estimates
    → Evaluate if additional risk controls needed
    → If safety issue: initiate FSCA/field action assessment

Cross-Reference: NIST Cybersecurity Framework Risk Assessment

Map ISO 14971 risk management to NIST CSF 2.0 for comprehensive risk coverage:

See also: ../information-security-manager-iso27001/SKILL.md for ISO 27001 security controls that serve as risk mitigations.

Cross-Reference: DORA ICT Risk Management

For medical device companies operating as or supplying to financial entities in the EU, the Digital Operational Resilience Act (DORA, Regulation 2022/2554) adds ICT risk requirements:

Enhanced FMEA with Cybersecurity Failure Modes

Combined Safety-Security FMEA Template

COMBINED SAFETY-SECURITY FMEA

Product: [Device Name]
Subsystem: [Subsystem]
Date: [Date]

TRADITIONAL SAFETY FAILURE MODES:
| ID | Item | Function | Failure Mode | Effect | S | Cause | O | Detection | D | RPN | Control |
|----|------|----------|--------------|--------|---|-------|---|-----------|---|-----|---------|
| FM-001 | Sensor | Measure vital sign | Incorrect reading | Wrong diagnosis | 8 | Calibration drift | 4 | Self-test | 3 | 96 | Auto-calibration |

CYBERSECURITY FAILURE MODES:
| ID | Asset | Security Objective | Threat | Attack Vector | Exploitability (O) | Impact (S) | Detection (D) | RPN | Security Control |
|----|-------|-------------------|--------|---------------|-------------------|-----------|---------------|-----|-----------------|
| CS-001 | Sensor data | Integrity | Data manipulation | MITM attack | 3 | 8 | 5 | 120 | TLS + data signing |
| CS-002 | Firmware | Integrity | Malicious update | Supply chain | 2 | 10 | 6 | 120 | Secure boot + code signing |
| CS-003 | User interface | Availability | DoS attack | Network flooding | 5 | 6 | 4 | 120 | Rate limiting + redundancy |

AI/ML FAILURE MODES (if applicable):
| ID | Component | ML Function | Failure Mode | Clinical Effect | S | Cause | O | Detection | D | RPN | ML Control |
|----|-----------|-------------|--------------|----------------|---|-------|---|-----------|---|-----|-----------|
| AI-001 | Classifier | Diagnose condition | False negative | Missed diagnosis | 9 | Distribution shift | 4 | Performance monitoring | 5 | 180 | Drift detection + human review |
| AI-002 | Classifier | Diagnose condition | Biased output | Health disparity | 8 | Unrepresentative training data | 3 | Subgroup analysis | 6 | 144 | Fairness constraints + diverse data |

COMBINED RPN THRESHOLDS:
>200: Critical — Immediate action required (all categories)
100-200: High — Action plan within 30 days
50-100: Medium — Monitor and consider action
<50: Low — Accept and monitor

Cybersecurity-Safety Interaction Analysis

Enhanced Risk Management — AI, Cybersecurity & Cross-Framework Integration

AI-Specific Risk Management

When managing risk for AI/ML medical devices, extend ISO 14971 with:

• AI Model Risk: Training data bias, model drift, adversarial attacks, explainability gaps
• Performance Degradation: Monitor for distribution shift, concept drift, and data quality issues
• Algorithmic Bias: Demographic parity, equalized odds, calibration across subgroups
• Human-AI Interaction Risks: Over-reliance, automation bias, alert fatigue, trust calibration
• Cross-reference: See eu-ai-act-specialist for EU AI Act risk classification

Cybersecurity Risk Integration (IEC 81001-5-1)

• Health Software Cybersecurity: IEC 81001-5-1 extends ISO 14971 for cybersecurity
• Threat Modeling: STRIDE methodology applied to medical device architecture
• Cybersecurity FMEA: Failure modes include unauthorized access, data breach, ransomware, supply chain attack
• Vulnerability Management: CVSS scoring integrated with ISO 14971 severity/probability matrix
• Cross-reference: See infrastructure-compliance-auditor for technical security checks

Supply Chain Risk Management

• Component Risk: Third-party software vulnerabilities (SBOM-based assessment)
• Supplier Risk: Single-source dependencies, geopolitical risks, quality history
• Cloud Risk: Data residency, service availability, vendor lock-in
• Cross-reference: See nis2-directive-specialist for NIS2 supply chain requirements

Cross-Framework Risk Mapping

Troubleshooting

Success Criteria

• Risk Management Plan approved with defined scope, risk acceptability matrix, RACI chart, and post-production monitoring plan before design input phase
• 100% of ISO 14971 hazard categories analyzed (electrical, mechanical, thermal, radiation, biological, chemical, software, use error, environmental) with documented rationale for each
• All identified risks evaluated against the 5x5 acceptability matrix with no uncontrolled "Unacceptable" residual risks remaining
• Risk controls implemented following the priority hierarchy (inherent safety first, then protective measures, then information for safety) with verification records for every control
• Overall residual risk evaluated as acceptable or ALARP demonstrated, with benefit-risk analysis completed for any residual risks remaining in High territory
• Post-production risk monitoring operational with defined information sources, review triggers, and a documented process for updating the Risk Management File
• For AI/ML devices: AI-specific risk categories (bias, drift, adversarial inputs) assessed per BS/AAMI 34971:2023 or equivalent, with continuous performance monitoring thresholds defined

Scope & Limitations

In Scope:

• ISO 14971:2019 risk management process implementation (planning, analysis, evaluation, control, residual risk, production/post-production)
• 5x5 risk matrix calculation and FMEA RPN scoring
• Hazard analysis methodology guidance (FMEA, FTA, HAZOP, Use Error Analysis, PHA)
• Risk control hierarchy application and verification planning
• Benefit-risk analysis framework
• Post-production risk monitoring and risk file update triggers
• AI/ML-specific risk management extensions (model bias, drift, adversarial inputs)
• Cybersecurity risk integration per IEC 81001-5-1
• Supply chain risk assessment methodology
• Cross-framework risk mapping (ISO 14971, NIST CSF, DORA, NIS2)

Out of Scope:

• Clinical investigation design or execution (risk management informs clinical strategy but does not execute studies)
• Software hazard analysis per IEC 62304 (the skill references software risk but detailed software lifecycle management requires IEC 62304 expertise)
• Biocompatibility testing or ISO 10993 evaluation (the skill identifies biological hazards but does not execute biocompatibility testing)
• Cybersecurity penetration testing or vulnerability scanning (use infrastructure-compliance-auditor for technical security testing)
• CAPA root cause analysis execution (use capa-officer for 5-Why, Fishbone, FTA, FMEA-based root cause investigation)
• Regulatory submission of risk management files (use regulatory-affairs-head for submission strategy and packaging)

Integration Points

Tool Reference

risk_matrix_calculator.py

Calculates ISO 14971 risk levels and FMEA Risk Priority Numbers.