Remote patient monitoring depends on the FHIR Subscription mechanism in a way that few other clinical workloads do. A glucose reading from a connected device, a blood pressure measurement from a home cuff, or a step count from a wearable lands as an Observation, and the care team has to see the relevant ones within seconds. The five subscription engines below have been used in 2026 production remote patient monitoring stacks and handle the topic-based subscription model under real device volume. For the healthcare interoperability hub, the broader FHIR coverage walks through the surrounding architecture.
Five Subscription Engines Used in 2026 RPM Backends
- Aidbox Subscriptions. Topic-based subscription support that fits the R5 specification and handles the filtering logic remote patient monitoring needs. Used in several RPM stacks running tens of thousands of devices.
- HAPI FHIR Subscriptions. The reference open-source implementation, used by RPM teams who self-host and want a familiar Java extension point for custom delivery channels.
- Smile Digital Health CDR Subscriptions. The commercial bundle of HAPI subscriptions with operational support, picked by RPM teams that want vendor backup for the message delivery layer.
- Google Cloud Healthcare API Notifications. Integrates with Google Cloud Pub/Sub for downstream message delivery, picked by RPM teams already running on Google Cloud.
- Medplum Subscriptions. Native subscription support with WebSocket and REST hook delivery, picked by smaller RPM startups for the speed of getting a first integration live.
The five represent the realistic options for an RPM team evaluating FHIR subscription engines in 2026.
What RPM Stresses About FHIR Subscriptions
Remote patient monitoring puts a specific kind of pressure on FHIR subscriptions that hospital-internal workflows rarely match. Volume: a single RPM customer with 10,000 patients on connected devices can produce a million Observations a day, and the subscription engine has to filter that stream without backing up. Reliability: a missed glucose alert is a clinical safety issue, not just an engineering bug. Topic granularity: the care team only wants the readings that cross a threshold, not every reading, and the topic filter has to support that level of detail without expensive server-side scans.
A subscription engine that handles these three under real device load lets the RPM product focus on the care team workflow layer. A subscription engine that drops messages under load forces the team to build a parallel delivery system, which is the situation FHIR subscriptions were designed to prevent.
How to Test an RPM Subscription Setup Before Production
The most useful pre-production test for an RPM subscription engine is a synthetic load that mimics a realistic device fleet: tens of thousands of patients, each emitting Observations at the device's actual cadence, with topic filters configured to fire on threshold crossings. A subscription engine that holds latency under one second across 24 hours of that load is unlikely to surprise the RPM team at launch. A subscription engine that backs up during the first hour signals an architectural mismatch that will not improve in production.
The FHIR server complete guide covers the server choices that underlie these subscription engines, the telemedicine backend roundup is the natural next read for RPM teams that also run synchronous virtual visits, and the bulk export performance guide covers the read side of the same data path.
Sources
- topic-based subscription model for RPM, evergreen - HL7 FHIR R5 Subscriptions specification
- backporting topic-based subscriptions to R4 - HL7 FHIR Subscriptions R5 Backport IG v1.2.0-ballot
- R5 Subscriptions (community-implementer presentation) - PDF slides, Gino Canessa, DevDays 2020
