Case Study — MedTech / Medical Lifecare
Client: Leading UK-Based Medical Device Company (Remote Patient Monitoring) | Category: MedTech Hardware & Software Testing
Our client is a pioneering UK-based medical device company that has built a Patient Status Engine (PSE) — a continuous, wireless vital signs monitoring platform designed to deliver near-ICU grade patient surveillance both inside hospitals and in remote home-care settings. The PSE captures, transforms, and analyses real-time vital signs data from wearable biosensors, generating predictive clinical insights and Early Warning Score (EWS) alerts that enable care teams to intervene proactively before patient deterioration becomes critical.
The platform is designed for deployment across hospital wards, high-dependency units, and community/home care facilities — extending clinical-grade monitoring beyond hospital walls. Given the life-critical nature of the application, the client required rigorous QA testing that went far beyond conventional software testing, demanding hardware-level signal validation and full IEC 62304 / FDA-aligned verification evidence.
Testing a Patient Status Engine is fundamentally different from testing standard enterprise software. The system sits at the intersection of medical hardware, embedded firmware, wireless protocols, cloud infrastructure, and clinical decision logic — each layer introducing unique failure modes that can directly impact patient safety. The client faced five critical, interconnected testing challenges.
The development team was initially testing vital sign capture using software-generated mock data. This approach passed all unit tests but masked critical hardware-level issues: ADC (analog-to-digital conversion) inaccuracies, signal noise rejection failures, and timing drift between sensor channels. A 2% SpO2 offset that went undetected in software testing would translate to a clinically significant misread — potentially delaying intervention for a deteriorating patient.
The PSE implements a NEWS2-based Early Warning Score engine that aggregates six vital sign parameters into a composite risk score, triggering clinical escalations at defined thresholds. During exploratory testing, Thoughtcoders discovered that the scoring algorithm produced incorrect composite scores at boundary conditions — specifically when two parameters simultaneously sat at adjacent-threshold values, causing the total score to be under-reported by 1–2 points. In a system where a score of 5+ triggers urgent clinical review, this under-reporting could directly delay life-saving intervention.
The PSE's remote monitoring capability relies on BLE sensors transmitting to a bedside gateway, which forwards data via cellular/Wi-Fi to the cloud. In real-world deployments — hospital corridors, homes with inconsistent Wi-Fi, community facilities — connectivity is intermittent. Testing revealed that when connectivity dropped for more than 90 seconds, the local buffer on the gateway would begin overwriting older readings rather than queuing them for transmission on reconnection. Vital signs data from the gap period was permanently lost — invisible to the clinician dashboard, with no gap indicator displayed.
The PSE wearable sensor includes ECG arrhythmia detection — one of its highest clinical value features. Testing with the Fluke ProSim 8 revealed that when motion artifact signals (generated via the simulator's artifact injection mode) were applied simultaneously with specific arrhythmia waveforms (AF, PVCs), the device's artifact rejection algorithm was incorrectly classifying the combined signal as pure motion artifact and suppressing the arrhythmia alert entirely. A patient with paroxysmal atrial fibrillation who moved during an episode would generate no alert.
The client's regulatory submission timeline for CE/MDR approval was 4 months away. The existing test documentation consisted of informal sprint test notes and JIRA tickets — none of which were formatted as traceable verification evidence linking software requirements to test cases, test results, and defect disposition. The regulatory body requires a complete Software Verification and Validation (V&V) package with full traceability and risk-based coverage justification per ISO 14971. Without this, the submission could not proceed.
Thoughtcoders deployed a dedicated MedTech QA team with hardware simulator access, embedded systems testing expertise, and IEC 62304 documentation capability. The engagement ran as an embedded sprint team alongside the client's development and regulatory teams over 16 weeks.
Thoughtcoders deployed a Fluke ProSim 8 and Rigel Uni-Sim to generate real analog physiological signals across all six vital sign parameters. Test scripts were written using the ProSim 8's Ansur USB interface to automate waveform injection sequences — replacing ad-hoc manual simulator operation with repeatable, scripted signal test runs. Every parameter was tested across its full clinical range (e.g. SpO2 70%–100% in 1% increments, ECG rates 30–300 bpm, NIBP 40/20 to 250/150 mmHg), and the ADC offset bug was identified and documented with exact signal values, providing the engineering team a precise reproduction recipe.
A 450-row combinatorial test matrix was designed covering all NEWS2 parameter combinations across low/normal/high ranges, with special focus on boundary conditions (parameters sitting at scoring thresholds). Automated Python pytest scripts executed the matrix against the PSE API endpoint — identifying the 23 specific parameter combinations where the composite score was under-reported. Root cause was traced to an integer truncation error in the score aggregation function rather than the lookup table, which would not have been apparent from manual spot-checking.
A dedicated test environment simulated network disconnection at controlled intervals using Wireshark packet capture and network throttling tools. Thoughtcoders documented the exact buffer overflow threshold (91 seconds) and the overwrite behavior, providing engineering with the failure reproduction steps. Post-fix, a gap-detection and queuing mechanism was implemented and re-validated: vital signs captured during connectivity gaps were now correctly queued locally, transmitted on reconnection, and displayed on the clinician dashboard with a gap-indicator annotation — maintaining full clinical record integrity.
Using the Fluke ProSim 8's simultaneous waveform and artifact injection capability, Thoughtcoders systematically tested 12 arrhythmia types combined with 4 motion artifact intensity levels. The test suite confirmed the suppression bug was specific to AF + moderate/high artifact at HR >90 bpm. Post-fix, re-testing with the same simulator configuration confirmed arrhythmia alerts were correctly generated even under high motion artifact conditions — a test that is not reproducible without hardware signal simulation.
Thoughtcoders authored a complete Software Verification and Validation package aligned to IEC 62304 Clause 5.7 and ISO 14971 risk management. This included: a Software Verification Plan (SVP), a full Requirements-to-Test-Case traceability matrix covering all 214 software requirements, test execution records with pass/fail evidence for every test case, defect disposition records for all open and closed defects, and a risk-based coverage justification document identifying safety-critical functions and their enhanced test coverage rationale. The package was submitted to the regulatory body on schedule, with no additional information requests received.
Thoughtcoders integrated the ProSim 8 Ansur-controlled signal injection scripts into the client's CI/CD pipeline. On every nightly build, 80 core signal accuracy tests were automatically executed against the latest firmware — providing continuous hardware-level regression coverage that would catch any signal processing regression before it reached clinical testing. This was the first time the client had hardware-in-the-loop testing in their automated build process.
The Patient Status Engine engagement demonstrated that software-only QA is fundamentally inadequate for life-critical medical monitoring devices. Every one of the five critical defects found during this engagement — SpO2 offset, EWS boundary under-reporting, data loss during connectivity gaps, arrhythmia suppression under motion artifact, and missing regulatory documentation — would have either reached clinical trials undetected, or blocked the CE/MDR submission entirely.
The arrhythmia + motion artifact defect is particularly illustrative: it is physically impossible to reproduce this failure mode without a hardware simulator generating simultaneous real ECG waveforms and motion artifact signals. No software mock, unit test, or manual testing approach could have uncovered it. This is the core case for hardware-first testing in medical device software QA.
By embedding hardware simulators, combinatorial algorithm testing, connectivity stress testing, and IEC 62304-aligned documentation into a single integrated QA engagement, Thoughtcoders enabled the client to:
"A Patient Status Engine that clinicians trust enough to act on — and patients trust enough to sleep soundly under — has to be tested at the hardware signal level, not just the software logic level. That is the standard we bring to every MedTech engagement."
— Thoughtcoders MedTech QA TeamTalk to our MedTech QA specialists about hardware simulator testing, EWS validation, and IEC 62304 documentation for your device.