Behavioral Health Conversion: Locking, Egress, and a Blocked Horizontal Exit

I would separate the two issues, because the architect is conflating them.

Horizontal exit. A horizontal exit is a legitimate means of egress only if it discharges into a refuge area that can hold the occupant load of the discharging side at the code-specified density, the barrier is properly rated and the doors self-close and latch, and the receiving space is not itself compromised. For 18 sleeping occupants this is usually achievable, but I would require a written capacity calculation for the receiving medical unit and confirmation that its own corridor is not counted twice. If those hold, I can defend the egress concept.

Delayed egress. A 15-second delay is not clinical security. A patient who is determined enough simply waits out the delay every time, and the arrangement depends on AHJ approval plus proper integration with the fire alarm. In a behavioral occupancy the right mechanism is the code-recognized special locking arrangement for such units: staff continually available, keys immediately at hand, release on alarm, and documented AHJ approval with a written staffing and training plan.

So: defend the horizontal exit with the capacity math, but replace the delayed-egress proposal with an approved special locking arrangement. Hardware must still be single-operation in the direction of egress.

The horizontal exit is a life safety feature, not a convenience path, so the default answer is that it stays usable. If the design genuinely cannot proceed without removing that segment, I require, in writing and before demolition:

• An approved alternative egress analysis showing travel distance and clear width on the remaining path, signed by the design professional and accepted by the fire marshal.
• Interim Life Safety Measures scoped to the actual risk: the fire alarm and sprinkler systems in the work area stay in service or a documented fire watch is implemented; combustibles removed and hot work permitted separately; temporary partitions and directional signage installed; staff briefed on the altered route and on assisting patients who cannot self-evacuate.
• An ICRA for the demolition dust, with negative pressure and barriers, because the adjoining unit is occupied by potentially immunocompromised patients.
• A daily walkthrough by facilities, logged, with authority to stop work if the corridor clear width or exit signage degrades.
• Sequencing so demolition happens in the shortest window and the exit is never blocked overnight, and a costed contingency to re-route the nurse station if the AHJ refuses.

I would not sign a permit on the promise that 'it's only 10 weeks.'

I would reject both proposals, but not dismiss the underlying concern.

Pull stations. Locking a manual fire alarm box defeats its purpose and will not survive an AHJ review. The defensible options are institutional covers that remain operable without a key by anyone, or relocating the station to a supervised location outside the patient-accessible corridor with the AHJ's agreement. If neither works, the design changes, not the device.

Sprinkler heads. Painting a head voids its listing and is a straightforward code violation; covering one is worse because it delays operation. The correct answer is a listed institutional or recessed sprinkler head designed for the environment, or a concealed space arrangement with protective cage where the listing permits it.

My ruling: the ligature risk is real and must be engineered out, but the tools are listed institutional devices, design layout changes, and staff observation of high-risk points - never disabling, obstructing, or covering life safety equipment. I would require the design team to return with a device-by-device schedule showing the listed product for each location, and I would document the AHJ's concurrence before the fixtures are ordered.