Hospital Evacuation Plan Requirements: Vertical vs. Horizontal Evacuation

The CMS emergency preparedness Condition of Participation requires every Medicare-certified hospital to have a policy and procedure for safe evacuation, and a separate policy for tracking the location of staff and sheltered patients. Both requirements are one sentence long in the regulation. Neither tells a hospital which patients to move first, how to decide between relocating patients down the hall versus down the stairwell, or what a tracking method actually has to record once patients start changing physical location. Those decisions are what a surveyor, or an accreditor’s tracer, actually tests — and they are the parts a generic, all-hazards emergency operations plan (EOP) typically leaves unresolved because the EOP is written to cover every hazard type, not the specific mechanics of moving patients out of a building.

What the CoP actually requires, and where it stops

Under 42 CFR 482.15(b), the hospital’s emergency preparedness policies and procedures must address, among other subjects: a system to track the location of on-duty staff and sheltered patients in the hospital’s care during an emergency, including the specific name and location of any receiving facility if patients are relocated; and safe evacuation from the hospital, covering the care and treatment needs of evacuees, staff responsibilities, transportation, evacuation locations, and primary and alternate means of communicating with external sources of assistance. These two elements sit alongside subsistence needs, shelter-in-place, medical documentation, staffing/volunteer management, and the hospital’s role under an 1135 waiver as the eight subjects the CoP requires the plan to cover.

That is the floor, not the plan. The regulation tells a hospital that it must be able to evacuate safely and track patients — it does not tell the hospital how to decide which type of evacuation a given incident calls for, or what a tracking record needs to contain once a patient is no longer where the census system says they are. A hospital’s hazard vulnerability analysis identifies which scenarios could force an evacuation in the first place; the evacuation plan itself has to operationalize the response once one of those scenarios happens, and that operational detail is what this page covers.

Horizontal, vertical, and full evacuation: how the decision actually gets made

Hospitals built to NFPA 101’s health care occupancy chapter are designed around a defend-in-place philosophy: the building itself, divided into fire-rated smoke compartments connected by rated doors, is the first line of protection, which is why relocating patients rather than leaving the building is usually the first move, not the last. A hospital evacuation plan needs to specify, in order, which of three evacuation types applies and what triggers a move to the next one — not just assert that “evacuation procedures exist.”

Horizontal evacuation — the default first response

Patients are moved through a rated door into an adjacent smoke compartment on the same floor. This is the default response to a fire, smoke condition, or localized hazard (e.g., a gas leak, a flood confined to one wing) because it keeps patients on the same level, close to their care team and equipment, and inside a compartment built to resist fire and smoke spread for a defined period. Plans should name the paired smoke compartments for each unit in advance — deciding this during the event, rather than referencing a pre-built compartment map, costs time a fire scenario doesn’t allow.

Vertical evacuation — moving between floors

Patients are moved to a different floor, typically down a stairwell, because elevators cannot be relied on during a fire (power loss, smoke infiltration of the shaft, code restrictions on elevator use during a fire alarm) and are used for vertical movement only under specific, pre-planned conditions using firefighter or fire-service elevator control. Vertical evacuation is triggered when horizontal relocation is not sufficient — the threat has compromised the compartment itself, is spreading toward it, or the incident (structural damage, a flood rising through lower floors) makes the current floor untenable. It is the most staff- and equipment-intensive tier below full evacuation: moving a non-ambulatory or bed-bound patient down multiple flights requires an evacuation chair or sled, typically two to four staff per patient, and materially more time than horizontal relocation of the same patient.

Full (complete) facility evacuation

The entire building is emptied to an outside marshalling point and then to receiving facilities. This is the last-resort tier — triggered by conditions defend-in-place cannot survive at all (total loss of utilities beyond generator/fuel endurance, structural compromise, a hurricane storm-surge or flood scenario making the building itself unsafe to occupy) — and the one that activates the CoP’s requirement for arrangements with other hospitals and providers to receive transferred patients, and potentially the facility’s role under an 1135 waiver if CMS conditions of participation themselves need to flex during the response.

The decision among the three sits with the hospital’s incident command structure, not with unit-level staff improvising in the moment. See Hospital Incident Command System (HICS) for how that command structure is organized and who actually issues an evacuation order once the hazard picture changes.

The patient-tracking method your plan has to specify

“Have a tracking system” is not an operational answer, and a general census or ADT (admit-discharge-transfer) feed is not, by itself, an evacuation-tracking method — a census system assumes patients stay where they’re documented until a discrete transfer event is entered; an evacuation moves multiple patients across locations in a compressed window, often faster than routine ADT documentation can keep pace with. A plan element that will actually hold up under a tracer needs to specify what gets recorded at the point of movement, not just that “tracking occurs”:

• Patient identity — name/MRN or, where verbal identification isn’t reliable (sedated, non-verbal, or pediatric patients), a triage tag or wristband method established in advance.
• Origin location — the unit, room, and bed the patient moved from.
• Acuity/mobility class — which of the sequencing tiers below the patient falls into, since this drives both which staff and which equipment go with them.
• Destination — the receiving unit, floor, or, for a full evacuation, the specific name and location of the receiving facility, matching the CoP’s explicit requirement that the receiving facility be documented, not just that relocation happened.
• Time of movement and the staff member responsible for the transfer (escort or unit charge nurse), so a family member’s or surveyor’s question about a specific patient’s location has a single point of accountability, not a search across multiple units’ paper logs.
• Equipment that moved with the patient — a ventilator, IV pumps, a portable monitor — since equipment left behind during a chaotic evacuation is a documented, recurring failure mode and losing track of it downstream complicates both patient care and asset recovery.

None of this requires a specific vendor platform — a paper tag-and-log method that captures all six fields is a compliant tracking method; a verbal headcount is not. What matters for the CoP requirement, and for a surveyor tracer that follows one evacuated patient’s paper trail end to end, is that every one of those fields exists somewhere and can be reconstructed after the fact.

Sequencing by acuity: the piece most plans get backwards

The intuitive assumption — move the sickest patients first — is usually the wrong operational answer, and most functioning hospital evacuation plans sequence in close to the opposite order, for a specific reason: the goal during the movement window is to clear egress routes, stairwells, and elevators as fast as possible while committing the largest staff and equipment resource to the smallest, most complex group of patients last, once routes are clear and dedicated teams are free to focus on them.

• Tier 1 — ambulatory patients. Can walk independently or with minimal assistance. Moved first, often in groups with a single escort, because they require the least staff time per patient and their movement doesn’t compete for the equipment (chairs, sleds, portable monitors) the later tiers need.
• Tier 2 — non-ambulatory, stable patients. Wheelchair-dependent or requiring an evacuation chair/sled but not on life-sustaining equipment. Moved next, typically one or two staff per patient.
• Tier 3 — critical-care, bed-bound, and ventilator-dependent patients. ICU, OR, L&D, and other patients on life-sustaining equipment. Moved last, with the largest per-patient staff and equipment commitment (typically three or more staff, portable ventilators or bag-valve devices, and a designated receiving bed confirmed before the move starts) — not because they matter less, but because they are the most complex movement to execute correctly, and are often more stable left in place until the last safe moment (while power, medical gas, and monitoring remain functional) than exposed to prolonged transport risk before the route and receiving bed are ready.

This sequencing logic is why the CMS rule’s history matters for planning, not just for citation. The 2016 rule followed a series of well-documented hospital evacuation failures during Hurricane Katrina and Superstorm Sandy, in which hospitals without a pre-built sequencing and tracking plan had to organize the movement of critical, equipment-dependent patients under improvised conditions. A plan that names the three tiers, assigns a staffing ratio and equipment list to each, and states explicitly that Tier 3 moves last (with the rationale documented) is the difference between a policy that reads as adequate on paper and one a unit charge nurse can actually execute at 2 a.m.

Where this plan differs from your general Emergency Operations Plan

A hospital’s EOP is deliberately broad — an all-hazards framework built from the hazard vulnerability analysis, covering command structure, communications, and response across every hazard type the HVA identifies, from active-shooter to a regional power outage. Evacuation is one annex within that structure, not a restatement of it, and it needs detail the EOP’s general language doesn’t carry:

• The EOP states that a tracking system exists; the evacuation annex specifies the six data fields above and who owns capturing them at the point of movement.
• The EOP names incident command’s authority to order an evacuation; the evacuation annex specifies the horizontal/vertical/full decision criteria and the pre-mapped smoke-compartment pairings incident command uses to make that call quickly.
• The EOP addresses staffing and volunteers generally; the evacuation annex assigns a staff-to-patient ratio and equipment list to each of the three acuity tiers.
• The EOP references arrangements with other providers; the evacuation annex is where those arrangements become specific receiving-facility names and bed-confirmation steps for a full evacuation.

A surveyor or accreditor tracer that asks a charge nurse to walk through “how would you evacuate this unit right now” is testing the annex, not the EOP’s cover language — which is exactly why the annex-level detail above, not just the policy’s existence, is what a plan review needs to confirm before the next survey cycle.

Frequently asked questions

Does CMS require a written horizontal-vs-vertical-vs-full evacuation policy specifically?

42 CFR 482.15(b) requires a policy and procedure addressing safe evacuation generally — care and treatment needs of evacuees, staff responsibilities, transportation, and evacuation locations — without naming “horizontal,” “vertical,” or “full” evacuation as required categories. Building the plan around those three tiers is standard hospital emergency-management practice for meeting that requirement in an operationally usable way, not a distinct regulatory citation.

Can elevators be used for vertical evacuation?

Only under specific, pre-planned conditions using firefighter or fire-service elevator recall/control, and generally not for routine vertical movement during an active fire — code restrictions and the risk of smoke infiltration or power loss to the shaft are why stairwells, evacuation chairs, and sleds remain the default for moving non-ambulatory patients between floors.

Who decides which evacuation type to use during an actual event?

The hospital’s incident command structure, informed by real-time information about the hazard’s location and behavior relative to the affected smoke compartment or floor — see Hospital Incident Command System (HICS) for how that command decision authority is organized.

Does the patient-tracking requirement apply to shelter-in-place too, not just evacuation?

Yes — 482.15(b) requires tracking the location of staff and sheltered patients generally, whether the hospital evacuates or shelters in place. The difference in practice is that a sheltering patient’s location doesn’t change, so the tracking method only needs to confirm presence; an evacuating patient’s location changes multiple times in a short window, which is why the evacuation-specific method needs the additional origin/destination/time/escort fields described above.

Related reading

• Hazard Vulnerability Analysis and the CMS Emergency Preparedness Rule
• Hospital Incident Command System (HICS): Structure, Forms, and How It Is Run
• NFPA 101 Life Safety Code in the Hospital Survey
• CMS Conditions of Participation for Hospitals
• Choosing Emergency Notification Software for a Campus or Health System