Choosing the Right Hazard Identification Technique for Digital Health Products
Most digital health companies approach clinical safety as a compliance checkbox. They download a generic FMEA template, spend weeks filling in spreadsheets about theoretical component failures, and wonder why their safety documentation feels disconnected from the actual risks their product poses to patients.
The problem isn't that FMEA is wrong—it's that hazard identification isn't one-size-fits-all. Different products create different types of clinical risk, and the technique you choose should match the reality of how your system can actually cause harm.
Understanding the Landscape
Clinical safety in digital health isn't about preventing mechanical failures. It's about understanding the complex ways that technology, clinical workflows, human cognition, and organisational pressures intersect to create patient harm. The right hazard identification technique helps you see these interactions clearly.
Here's how to match method to system type:
FMEA: For Component-Based Systems with Clear Functional Flows
Failure Mode and Effects Analysis works brilliantly when you can map discrete functions that might fail in predictable ways.
Best suited for:
Medication management applications
Diagnostic calculators
Appointment booking systems
Anything with clear inputs, processes, and outputs
Why it works: FMEA forces you to ask: "What happens if this specific component fails?" For a medication reminder app, that's straightforward—what if the notification doesn't fire? What if the dose is displayed incorrectly? What if the user dismisses the alert?
The limitation: FMEA struggles with emergent risks that arise from system interactions rather than component failures. It's less effective when the risk isn't "this feature broke" but "this feature worked exactly as designed and still caused harm."
When to use it: Your product has well-defined features with clear success/failure states. You can map user journeys step-by-step. The clinical risk is primarily about things not working as intended.
HAZOP: For AI/ML Systems and Clinical Pathway Tools
Hazard and Operability Study excels at identifying risks that emerge from process deviations rather than component failures.
Best suited for:
AI-driven diagnosis or triage systems
Clinical decision support tools
Digital therapeutics that mediate care pathways
Any system where the algorithm's behaviour is the core clinical intervention
Why it works: HAZOP uses structured "what if" prompts to explore deviations: What if the model drifts? What if the input data quality degrades? What if the clinician over-relies on the algorithm's output? What if the patient population changes?
These aren't failures—they're deviations from intended operating conditions. The system might work perfectly, but the context has shifted in ways that create risk.
The critical insight: For AI/ML systems, the risk isn't that the code crashes. It's that a time-pressured clinician trusts the triage algorithm without questioning it, or that the model was trained on data that doesn't represent the actual patient population it encounters in a community mental health setting.
When to use it: Your product uses machine learning, mediates clinical decisions, or sits within complex care pathways where context matters as much as function. The risks you're worried about involve human-algorithm interaction, automation bias, or process disruption.
SWIFT: For Agile Teams and Early-Stage Products
Structured What-If Technique offers systematic hazard coverage without FMEA's time investment.
Best suited for:
Startups with rapid iteration cycles
Products in early validation stages
Teams without dedicated safety engineers
When you need "good enough" coverage quickly
Why it works: SWIFT combines HAZOP's deviation-based thinking with a faster, more flexible structure. You identify system elements, apply structured prompts, and capture hazards efficiently. It's less comprehensive than full FMEA but significantly faster.
The trade-off: You sacrifice some depth for speed. SWIFT won't give you the exhaustive documentation of FMEA, but it will surface the major hazards that matter most.
When to use it: You're iterating quickly and need to reassess safety regularly. Your team is small. You'd rather have decent hazard identification done this month than perfect analysis done never. You're building safety culture alongside safety documentation.
HAZID: For Engaging Clinical Stakeholders
Hazard Identification workshops excel at surfacing risks that pure technical analysis misses.
Best suited for:
Products requiring clinical stakeholder buy-in
Complex sociotechnical systems
When you need to understand how clinicians will actually use (or misuse) your product
Identifying human factors and workflow risks
Why it works: HAZID brings clinical stakeholders into structured brainstorming sessions. Clinicians who've never heard of FMEA can still answer: "What could go wrong when you use this during a crisis?" "How might a tired nurse misinterpret this alert?" "What happens when this tool meets the chaos of a weekend on-call rota?"
The value: You get insights you'd never find from technical specifications alone. The GP who points out that your beautifully designed interface will be completely unusable during a 10-minute consultation. The psychiatrist who identifies that your mental health screening tool could trigger crisis presentations without adequate safety protocols.
When to use it: You need clinical credibility for your safety work. Your product's success depends on workflow integration. You're trying to understand real-world implementation risks, not just technical failure modes. You want to build relationships with clinical advisers while doing safety work.
Get the Complete Framework
Understanding which technique to use is the starting point. The real challenge is knowing how to combine these methods strategically, facilitate workshops that surface genuine clinical risks, and avoid the common pitfalls that turn safety work into compliance theatre.
The full guide includes:
✓ The Hybrid Approach – How to combine techniques strategically for maximum coverage
✓ Practical Prompt Libraries – Ready-to-use question sets for HAZID and HAZOP workshops
✓ Decision Framework – Choose the right technique based on product type, team size, and timeline
✓ Case Study – Identifying safeguarding risks in a family therapy platform (illustrative)
✓ Red Flags Guide – Spot the difference between real safety work and compliance theatre
By downloading, you'll occasionally hear from me about digital health safety, regulatory updates, and NHS implementation insights. Unsubscribe anytime. I don't share your details with anyone.
