Epic’s ADT — Admit, Discharge, Transfer — event stream is one of the most comprehensive real-time data feeds in regional hospital operations. Every patient movement generates an HL7 message: registration, bed assignment, transfer between units, discharge. In a 300-bed regional hospital running Epic, this stream produces hundreds of discrete events per day. Most of them never reach the people who could act on them.
The gap is not a data problem. It is a surfacing problem. Epic was built to support clinical documentation and revenue cycle workflows. The ADT stream is technically accessible, but the workflows built around it were designed for billing and clinical record completeness, not for the operational decision a house supervisor makes every 20 minutes.
An ADT feed from Epic carries more information than most hospital operations teams use. Each event includes the patient’s current location and unit, admission type and acuity flags, attending physician, expected discharge indicators flagged by the care-management module, isolation and infection-control requirements, and the timestamp of the last status change. Combined with the OR schedule from Epic Surgical Services, these signals provide enough information to predict unit-level census with reasonable accuracy 4 to 8 hours ahead.
The challenge is that this information is distributed across multiple Epic modules — ADT, Hyperspace, the care-management workflow, the surgical services schedule — none of which share a unified view designed for the bed-assignment decision. A house supervisor who wants to know which admitted ED patients can be placed in which beds over the next two hours has to navigate four to six separate Hyperspace screens and synthesize the information manually.
Epic’s FHIR R4 API enables real-time streaming of ADT events to external applications without disrupting existing clinical workflows. The interface publishes patient encounters, location updates, discharge flags, and care-management status changes as they happen, typically with a latency of under 30 seconds from event to stream availability.
For patient flow applications, this means a 15-minute rolling prediction loop is technically feasible with standard Epic integration. The prediction model ingests the live ADT stream, cross-references it against historical arrival and discharge patterns calibrated per unit and per day-of-week, and produces updated census forecasts without requiring any manual data entry from clinical staff. The data already exists inside Epic — it just needs to be connected to a decision-support layer built for operational use.
This is where the integration architecture matters. Health IT teams at regional hospitals have limited engineering capacity. An ADT integration that requires custom FHIR endpoint development, ongoing maintenance of proprietary message formats, and coordination with Epic’s professional services team is unlikely to stay operational after the initial implementation team moves on. The most durable integrations use Epic’s published FHIR R4 endpoints and avoid custom HL7 v2 message parsing where possible.
Tools that surface ADT-based predictions inside Epic Hyperspace as sidebar gadgets see substantially higher adoption than standalone web applications. The practical reason is context switching. A house supervisor who has to leave Hyperspace to check a separate bed-management dashboard will stop checking the dashboard within two to three weeks under normal operating conditions. The same information presented as a gadget in the existing Hyperspace session stays in the workflow.
Epic’s App Orchard framework enables third-party applications to embed as Hyperspace components. Regional hospitals evaluating patient flow tools should prioritize solutions that use this integration pathway over solutions that require separate login credentials, separate browser windows, or dedicated hardware at the nursing station. The operational adoption data consistently favors embedded solutions over standalone ones.
A few prerequisites affect how quickly an ADT integration goes live. Most regional hospitals running Epic need to verify that their FHIR R4 endpoints are enabled and current — hospitals on older Epic versions may need an upgrade before the streaming API is available. Epic’s implementation team can confirm the version requirement during the initial technical discovery call.
Health IT also needs to confirm that the FHIR API scopes required for ADT events, care-management flags, and surgical services data are included in the integration’s OAuth configuration. These scopes are not enabled by default and require an Epic administrator to configure. The process is straightforward but takes time to schedule and complete in most regional hospital IT queues.
Finally, the clinical informatics team should be involved early in defining which ADT event types are most relevant to the bed-assignment workflow. Ingesting the full ADT stream is technically possible but creates noise. The most useful configurations filter for admitted patients awaiting bed assignment, flagged discharge-readiness indicators, and OR case completions expected within the next four hours. That filtered view is what gives supervisors actionable information without overwhelming them with irrelevant events.
