Independent learning for medical-device professionals
SearchCommentaryConsulting
Worked Examples & Case StudiesMTL-403 · WORKED PRODUCT CASE STUDY

Medical-device Mobile App and Cloud Service

An end-to-end example showing how a mobile application and continuously operated cloud service form one controlled medical-device system.

How to use this case study

This example treats the app, cloud and operational service as one product. Follow a safety-significant data item across components and organisations, then ask whether responsibility, evidence and change control remain intact.

01

Case at a glance

A patient app receives measurements from a compatible home-use device, presents trends and sends encrypted records to a hosted service. A clinician portal displays prioritised review queues. The software supports monitoring but does not diagnose automatically or replace urgent-care instructions.

Patient value

Reliable access to measurements, trends and actionable status information.

Clinical value

A controlled longitudinal view that helps professionals prioritise review.

Technical boundary

Mobile apps, APIs, hosted services, databases, clinician portal and operational tooling.

Principal uncertainty

Whether data remain timely, attributable and correctly interpreted across intermittent connectivity.

02

Separate monitoring support from diagnosis

Draft intended purpose

The system is intended to collect measurements from specified compatible devices, display longitudinal information to adult patients and authorised healthcare professionals, and support prioritisation against predefined review criteria. It does not provide an autonomous diagnosis or emergency response.

The boundary constrains notifications, claims, timing expectations and clinical evaluation. Product wording must not imply that absence of an alert means absence of risk. Use MTL-102 — Intended Purpose, Users and Use Environments.

03

Make distributed responsibility explicit

Measurement deviceSource value, unit, device identity, status and time
Mobile applicationPairing, local record, presentation, consent and synchronisation
Cloud APIAuthentication, validation, deduplication and routing
Clinical serviceStorage, prioritisation, audit and professional presentation
Platform dependenciesPhone OS, cloud provider, identity, messaging and monitoring services
OperationsDeployment, support, incident response, backup and recovery
04

Define the meaning of data at every boundary

  • Every measurement shall retain source-device identity, patient association, unit, original timestamp and synchronisation state.
  • The app shall distinguish measured, manually entered, calculated and corrected values.
  • Retry and recovery shall not create duplicate records or hide missing intervals.
  • Clock uncertainty and offline acquisition shall be visible where clinically relevant.
  • Priority rules shall use controlled versions and preserve the inputs and output of each evaluation.
  • Notifications shall state their meaning, delivery limitations and required user action.
  • Deletion and retention shall respect safety, regulatory, privacy and contractual obligations.

Apply MTL-111 — Data, Connectivity and Interoperability to the semantic as well as technical interface.

05

Combine safety, security and privacy decisions

Missing measurement

Connectivity failure creates false reassurance. Show data freshness, missing intervals and degraded operation.

Wrong patient

A valid value reaches the wrong record. Control enrolment, pairing, identity and reconciliation.

Altered priority

Rule or configuration error changes clinical review order. Version, test, authorise and monitor every rule.

Account compromise

An attacker views or alters sensitive records. Use proportionate authentication, authorisation, detection and recovery.

Safety and cybersecurity share system assets and consequences but require explicit linkage. See MTL-109 — Medical-device Cybersecurity and MTL-112 — Privacy and Data Protection by Design.

06

Design for degraded service

The mobile app retains recent records and clearly identifies unsynchronised data. The API is idempotent, rejects invalid schemas and preserves provenance. Clinical prioritisation runs as a versioned service whose output can be reconstructed. No single notification channel is treated as guaranteed delivery, and immediate safety instructions remain available without cloud access.

Trust boundaries, supported versions and recovery behaviour belong in the architecture, following MTL-105 — Systems Engineering, Architecture and Interfaces.

07

Test the service as users experience it

APP

Mobile behaviour

Pairing, permissions, offline use, accessibility, OS combinations, updates and recovery.

API

Interface integrity

Schema, identity, ordering, duplication, latency, errors, rate limits and compatibility.

CLOUD

Service assurance

Infrastructure configuration, resilience, backup, restore, observability, access and deployment controls.

CLINICAL

Claim validation

Representative data, priority performance, user interpretation, workflow fit and known limitations.

08

Operational controls are product controls

Release approval covers source, build artefacts, infrastructure configuration, database migration, feature flags, rollback, monitoring and support readiness. Alerts distinguish service health from product-performance signals. Incidents enter both technical response and medical-device assessment. Apply MTL-227 — Cloud, Platform and DevOps Engineers.

09

Follow one measurement thread

User needMy care team needs a dependable view of recent measurements
RiskMissing or stale data are interpreted as current
RequirementDisplay acquisition, synchronisation and freshness state
ArchitectureProvenance and state travel with the value through every service
VerificationOffline, retry, duplicate, clock, migration and failure testing
ValidationPatients and professionals interpret incomplete data correctly
10

Use an operating-system change to test lifecycle control

A mobile OS release restricts background communication, delaying synchronisation for some patients. The team must determine affected versions and workflows, test foreground and background states, monitor real-world latency, communicate limitations, update support rules and assess whether clinical claims or regulatory obligations are affected.

The event is a product change crossing software, usability, risk, supplier dependence, clinical performance and post-market surveillance—not simply an app defect.

DISCUSSION

Questions to challenge the case

  1. Which functions remain useful and safe without cloud connectivity?
  2. How does a reviewer know that data are complete and current?
  3. Who approves infrastructure and priority-rule changes?
  4. Which third-party service failure has the greatest product consequence?
  5. Can every displayed value be reconstructed from its source and configuration?
  6. What operational signal should enter post-market surveillance?