An information failure wearing a hardware costume.
The perimeter deserves its due, so take it off the table first. The exits are alarmed and the alarms fire. The badge readers work. Nothing about the hardware is broken, and this page is not going to sell you a better door — if it did, it would be doing the exact thing that has not worked for you so far. Set the door aside completely, because the argument only lands once the door is no longer the suspect.
Here is where it actually breaks. The identical event — a person leaving through a door — is a discharge when the patient has capacity and an elopement when they do not, and that difference is a clinical determination made upstream, sometimes hours before, sometimes on a different floor, by someone who never touches the security system. The documented root causes bear this out: elopement traces to breakdowns in patient assessment and team communication far more than to hardware. In one case in AHRQ’s own analysis, a missing inpatient was found near the emergency department, registered as a new ED patient, medicated, and discharged — while still listed as admitted upstairs. The hospital physically encountered the person it had lost and still lost him. No door would have helped, because the failure was that the risk flag never reached the place the person actually was.
You make this argument yourself, all the time. You tell a client that a sensor reporting to nobody is not a system — it is a component, and a component that alerts an empty hallway is decoration. Then look at your own elopement perimeter: when a flagged patient reaches a door, where does the alert go, and to whom, on what device — and if nobody acknowledges it, what happens next? If the honest answer is “it sounds, and someone nearby usually notices,” you have a component, not a system.
The pattern worth admitting is simple and repeated: we have sold perimeter into an information problem for years, because perimeter is what we make. A door contact is easy to quote, easy to install, easy to point at. The routing, the escalation, the record of who was told and whether they acknowledged — the parts that actually address the documented root cause — are harder to sell and less satisfying to install, so the industry keeps answering an information failure with more hardware. That is our pattern, and it is worth naming.
So the instrument is the only one of ours that is half information, not just a walk. You look at the doors, quickly — and then you trace a single elopement-risk flag through the building and ask whether the exit system knows about it, whether the alert reaches a named responder, and whether anything escalates when it goes unacknowledged. That trace is the shape of the whole problem, done in twenty minutes with a clipboard.
“The question that lands is always the same one. Pick a patient your board flags as an elopement risk today, and go find where that flag physically lives — the chart, a wristband, a whiteboard, a verbal handoff at shift change. Then ask whether the door knows. Nine directors out of the room go quiet on that one.”
Sources: AHRQ PSNet published case analysis of an inpatient elopement (patient found near the ED, re-registered, medicated, discharged while still admitted) — carried as one documented case, never as a frequency or rate. Root-cause characterization (assessment and team-communication breakdowns as primary contributors) from patient-safety literature. Location/time routing and escalation-logic material is RTLS-vendor-sourced and carried as design reasoning only, with no numeric claim. Standard references verified against the current accreditation manual at publish.
What would change this: New published elopement root-cause analyses; changes to elopement/wandering guidance; any RTLS field data that would let directional reasoning become a sourced claim.
Reviewed July 2026 by the Eagle Systems healthcare practice. The systems referenced are how a facility meets the accountability the regulation creates, not a mandate; the accreditation and compliance determinations belong to the organization and its regulators. Current as of Accreditation 360 (eff. Jan 1 2026).