Turnout time is the one interval in a fire response that a department controls almost completely. The call volume is not yours, the traffic is not yours, and the distance to the house on the far county line is not yours, but the seconds between the alert hitting the station and the apparatus clearing the bay door belong entirely to how you built the station, the alerting system and the habits around both. This piece covers where the tones came from, what the cardiac research actually establishes, how to measure turnout honestly, and what to do when the primary alerting path quits.

The interval nobody else owns

Response time analysis splits the clock into named intervals, and the names matter because people argue past each other when they do not use them. Alarm handling time runs from the moment the call is answered at the communications center to the moment the notification of the responding units begins. Turnout time runs from the start of that notification to the point at which the apparatus begins travel. Travel time runs from there to arrival on scene. Total response time is the sum, and the public and the press almost always mean the sum when they say “response time,” while the department usually means travel time when it defends itself.

NFPA 1710, the deployment standard written for career departments, sets turnout time objectives of 80 seconds for fire and special operations responses and 60 seconds for emergency medical responses, measured at the 90th percentile rather than as an average. NFPA 1720, the companion written for departments staffed wholly or substantially by volunteers, frames the problem differently, using staffing and total response time by population density demand zone instead of breaking out turnout as its own measured interval. NFPA has spent the last several years consolidating its emergency response and communications documents, and the numbering has moved, so pull the current edition from NFPA before you write a number into a policy or a grant application rather than quoting an edition you remember from a class.

The reason turnout deserves attention out of proportion to its length is that it is cheap to improve and it is measured in the same seconds as everything else. A department that shaves twenty seconds off its 90th percentile turnout has bought the same amount of time as a new station would buy on some fraction of its calls, at a fraction of the cost. The other reason is that turnout time is where the alerting system lives, and the alerting system is the piece of the chain that most agencies inherited, never specified, and cannot describe when you ask them how it works.

Two tones and a voice: where the vocabulary came from

The two-tone sequential page is the sound the American fire service grew up on. A tone encoder at the dispatch center transmits one audio tone for a nominal one second, then a second audio tone for a nominal three seconds, and a receiver programmed for that specific pair of frequencies in that specific order opens its audio and passes the dispatcher’s voice announcement. A long second tone of several seconds is used in many systems as a group call that opens every pager in a set. Motorola’s Quick Call II implementation became the de facto standard in the United States, and Plectron receivers covered much of the same ground earlier, which is why you still hear the word “Plectron” used generically by people who have not owned one in forty years.

What made the scheme durable is that it is entirely analog and entirely stateless. The pager does not register with anything, there is no channel grant, there is no database at the other end that has to be correct, and the receiver either hears the tones or it does not. That simplicity is also the reason so many counties that moved their dispatch to a trunked P25 system kept a separate VHF paging transmitter alive. Digital trunking, and encryption on top of it, breaks the legacy belt pager completely, and the cheapest answer for a county with four hundred volunteers carrying pagers has usually been to keep a conventional paging channel and simulcast the dispatch announcement on it.

Digital paging came out of a different lineage. POCSAG, the code standard developed by the British Post Office Code Standardisation Advisory Group and later adopted as CCIR Radiopaging Code No. 1, carries addresses and alphanumeric text at 512, 1200 or 2400 bits per second, and Motorola’s FLEX protocol runs faster and handles more subscribers per channel. Digital paging gives you text on the belt and, depending on the system, a delivery path that is more spectrally efficient than tone and voice. It does not by itself tell you whether the page was received, because one-way paging has no return path, and that limitation drives a good deal of the redundancy discussion later in this piece.

From CAD to the bay door: what modern station alerting does

A current station alerting system is a networked appliance in the firehouse that takes a message from computer aided dispatch over IP and acts on it locally. The CAD interface is the heart of it, and it is usually one of two shapes: a proprietary interface written by the CAD vendor to the alerting vendor’s specification, or a message on a standard transport that the alerting controller parses. The message carries the incident type, the address, the units assigned, and enough information for the station controller to decide which parts of the building to alert and what to say.

What the controller does with that message is where the capability lives. It can alert only the bunk rooms and bays associated with the units actually assigned, so the medic crew sleeps through an engine call. It can render the dispatch text to speech and play it into the station, then repeat it, then put it on a screen in the day room and the bay with the address and a map. It can open bay doors, turn the station lights from red night lighting up to full white, shut off the stove, disconnect shore power and trigger traffic signal preemption on the way out. It can time itself, recording the moment the alert fired at the station rather than relying on a timestamp generated somewhere in the CAD.

The acknowledgment function matters more than it looks. A button in the bunk room or the bay that the crew presses gives the dispatcher a positive indication that a human being at that station heard the alert, and it gives you a timestamp for the front half of turnout that is independent of the crew’s en route button on the mobile data terminal. In volunteer and combination systems, the same idea shows up as a response confirmation in a smartphone application, which tells the officer how many people are coming before the first one arrives. The value of all of it depends on someone looking at the indication, which is a staffing and procedure question rather than a hardware question.

Waking up at 150: what the cardiac research establishes

Sudden cardiac events have been the largest single category of on-duty firefighter deaths in the United States for decades. NFPA’s annual firefighter fatality reports have historically attributed somewhere in the range of 40 to 50 percent of on-duty deaths to sudden cardiac events, with year-to-year variation and with lower proportions in several recent years as the total number of fatalities has fallen. If you are going to cite a percentage in a policy document or a grant narrative, get it from the specific NFPA report year you are citing, because the number moves and a figure from a 2005 report is not a statement about today.

The study most often cited on where in the duty cycle the risk sits is Kales and colleagues, “Emergency Duties and Deaths from Heart Disease among Firefighters in the United States,” published in the New England Journal of Medicine in 2007. The authors drew on on-duty coronary heart disease deaths over roughly a decade and compared the share of deaths occurring during each category of duty against the share of time firefighters spend in that category. Fire suppression accounted for a large fraction of the deaths while occupying a very small fraction of on-duty time, and responding to an alarm and returning from an alarm each accounted for a share of deaths well out of proportion to their share of time. The authors reported the results as odds ratios relative to non-emergency duties, and the intervals around those estimates are wide, so read the paper rather than a slide that reduced it to one number.

What that study establishes is that the response phase carries elevated risk. What it does not establish, and what nobody should claim it establishes, is that the alerting tone itself is the cause of a measurable number of deaths. The mechanism people reason from is real enough on its own terms, because the acoustic startle reflex is a well documented response whose magnitude depends on the intensity and the rise time of the sound, and because abrupt waking from deep sleep produces a catecholamine surge and a rapid rise in heart rate and blood pressure. Going from that mechanism to a claim that a particular alerting product prevents heart attacks is a leap, and I have watched vendors make it in a sales presentation without a citation on the slide. Ask for the citation, read what it measured, and note whether the outcome was heart rate over the first minute or something that anyone would call a clinical endpoint.

Ramped audio, zoning and lighting: design intent and evidence

Ramped alerting means the alert tone starts at a low sound level and rises over several seconds to full volume, instead of arriving at full volume on the first cycle. The design intent is to bring a sleeping firefighter up through the arousal process over a few seconds rather than in one step, on the theory that the peak sympathetic response is smaller and the person is oriented sooner. Zoned alerting supports the same intent from a different direction by not waking anyone who is not going, which over a year of night calls is a meaningful reduction in total sleep disruption for a crew that runs mixed fire and EMS.

The lighting side follows the same logic. Red or low-level night lighting in bunk rooms and corridors that steps up to full white illumination as the alert progresses lets people see where they are going without the visual shock of an overhead fluorescent bank at two in the morning, and it removes one of the reasons people lie in bed for three seconds getting their eyes back. None of this is exotic, since it amounts to lighting control tied to the same trigger as the audio.

My honest read on the evidence is that the physiological rationale is plausible and the arousal and startle literature supports the general mechanism, while the specific claim that ramped station alerting reduces cardiac events is not something I can point you to a definitive trial for. I would still put the feature in a station, because the sleep quality argument stands on its own and crews who have it generally do not want to go back, and because zoned alerting reduces the number of disruptions regardless of what the ramp does, but the justification you write should describe the feature accurately, since a grant narrative that overstates the science is a narrative somebody can pick apart. Audibility is the constraint on the other side, and the convention used throughout fire alarm work, which comes from NFPA 72 for public mode signaling, is that the signal needs to exceed the average ambient sound level by 15 dB or exceed a maximum sound level of 60 seconds duration by 5 dB. A ramp that never reaches an audible level in a noisy apparatus bay has traded one problem for a worse one.

Measuring turnout honestly: timestamps and percentiles

Turnout time is only as good as the two timestamps that bracket it, and in most CAD systems at least one of those timestamps is lying to you. The front end should be the moment the alert actually fired in the station. What frequently gets recorded instead is the moment the dispatcher pressed the dispatch key, or the moment CAD wrote the record, or the moment the interface process got around to sending the message, and the gap between those and the sound in the bunk room can be several seconds of queuing and middleware. If your alerting system generates its own timestamp when the tone starts and writes it back to CAD, use that one and find out from your CAD administrator exactly which field it lands in.

The back end is worse, because the en route timestamp is generated by a human pressing a button, and the habit varies by crew, by station and by unit type. I have seen officers press en route while walking to the rig, which makes turnout look excellent and travel look terrible, and I have seen crews press it at the first traffic light, which does the reverse. Automatic vehicle location can give you a wheels-moving event that is independent of anyone’s thumb, and if you have AVL with a reasonable reporting interval, comparing the AVL departure to the en route press for a month will tell you how much of your turnout number is measurement artifact.

Report the 90th percentile, not the mean, and report it broken out by call type and by hour of day. An average hides the tail, and the tail is the part that hurts people, because the call that took four minutes to get out of the station is the one that ends up in the newspaper and the average never shows it. Fire responses will run longer than EMS responses because of the protective clothing, night calls will run longer than day calls, and a station where the bunk room is upstairs and the gear is in the bay will run longer than one where the gear is at the bunk room door. Those differences are diagnostic, and a single department-wide average erases every one of them.

How the primary path fails

The failures I see most often are not dramatic. The single most common is the CAD interface itself, because the alerting system depends on a process running somewhere that takes messages from CAD and hands them to the station controllers, and when that process hangs, CAD keeps working, the dispatcher keeps dispatching, and the stations hear nothing. Nothing in the dispatcher’s workflow necessarily reveals that, which is why the heartbeat monitoring in the earlier callout matters so much. A close second is the network path, where a station is fed by a single fiber or a single microwave hop or, in more places than people admit, a business-grade broadband circuit with no service level commitment and no alternate route.

Power is the third. The alerting controller is usually on a small UPS, and the network switch it depends on may be on a different UPS with different runtime, and the station’s radio receiver for the backup page may be on neither. A generator that starts reliably solves most of this, and I would still walk the station and physically trace what is plugged into what, because a piece of equipment added three years after the generator was sized is frequently on a wall outlet. Write down the runtime you actually have on battery, not the number on the UPS label, which assumes a load you are not running.

On the RF side, the classic paging failure is coverage that was never verified. Tone and voice paging on VHF gets deployed with the assumption that anywhere the mobile radio works, the pager works, which is not true, because a belt pager with a short helical antenna at hip level inside a house is a much worse receiving installation than a mobile with a quarter wave on a roof. Simulcast paging adds a second mode of failure in the overlap zones between transmitters, where two signals of similar strength and imperfect time or frequency alignment produce distorted audio that a pager’s decoder will not reliably accept. Smartphone alerting applications introduce their own set, because delivery depends on the carrier network, on the push notification service, on the phone’s operating system power management and on the user’s do not disturb configuration, and none of those are under your control or instrumented for you.

Building the second path and proving it works

The design question is not how many alerting methods you own but how many independent ones you have. A station alerting appliance fed by IP, a smartphone application and a web dashboard are three products sharing one network path and one CAD interface, so a middleware failure takes all three at once. Genuine independence means a second path that does not share the failure, and the most common honest answer in the fire service remains a conventional radio page transmitted from the dispatch center over a transmitter the agency controls, received by a base receiver in the station and by pagers on belts, because it shares neither the data network nor the CAD interface with the primary path.

The other paths worth having are simple. A ringdown telephone circuit or a published direct line to each station gives the dispatcher a way to reach a human being by voice when everything automated has quit, and it works as long as the telephone service works. The station’s own radio, monitored on the dispatch channel, means a crew can be alerted by voice on the same channel they are already required to monitor. For volunteer departments, keeping the belt pager alive alongside the application is not nostalgia, because the pager depends on one transmitter and one receiver while the application depends on a chain of commercial services with no committed delivery time.

Whatever you build has to be written down and drilled, because the failure mode I see in after-action review is a dispatcher who knew the primary was down and spent four minutes deciding what to do about it. The SOP needs to name the backup path, name who declares that the agency is operating on it, state how the dispatcher confirms receipt when the automated acknowledgment is gone, and state that stations go to manual acknowledgment by radio for the duration. Test the backup path on a schedule, log the test, and make the test realistic by sending an actual page to an actual station rather than by confirming that the equipment has a green light on it.

What to do at your agency

• Have your communications officer or data analyst pull the last full quarter of CAD records and compute 90th percentile turnout time separately for fire and EMS responses, broken out by station and by overnight versus daytime, and bring that one page to the officers’ meeting that already occurs this month.
• Ask your CAD administrator in writing which timestamp field is used as the start of turnout, whether it is generated by CAD or written back by the station alerting system, and keep the answer in the communications file so nobody has to guess the next time the number is challenged.
• Have a captain at each station send a live test page over the backup radio paging path this month, confirm it was heard on the station receiver and on at least one belt pager, and log the date and the result on the same form you use for apparatus checks.
• Walk each station with the maintenance supervisor and trace what the alerting controller, its network switch and the backup radio receiver are actually plugged into, then write the measured battery runtime and the generator circuit on a label at the equipment rack.
• Confirm with your dispatch center manager that every station alerting device reports a heartbeat to a screen somebody watches, and that a station going offline produces an alarm rather than a log entry nobody reads.
• Add one paragraph to your existing communications SOP naming the backup alerting path, the person who declares that the agency is operating on it, and the requirement that stations acknowledge by radio while it is in use.
• Have the training officer time turnout at one drill per station using a stopwatch from tone to wheels moving, compare it to what CAD reported for the same event, and use the difference to decide whether your reported numbers need correcting.

Takeaways

• Turnout time is the interval between the start of the alert in the station and the beginning of travel, and it is the part of total response time a department controls most directly and can improve most cheaply.
• NFPA 1710 sets turnout objectives of 80 seconds for fire and special operations and 60 seconds for emergency medical responses at the 90th percentile, and because NFPA has been consolidating and renumbering its documents, the current edition should be pulled directly before any number goes into a policy.
• Two-tone sequential paging survived because it is analog, stateless and independent of any database, which is also why counties that moved dispatch to trunked P25 systems kept a conventional VHF paging channel running for belt pagers.
• Kales and colleagues, writing in the New England Journal of Medicine in 2007, found that on-duty coronary heart disease deaths among firefighters were concentrated in fire suppression and in alarm response far out of proportion to the time spent in those activities, which establishes that the response phase carries elevated risk without establishing that the tone itself causes deaths.
• Ramped audio and zoned alerting rest on a plausible physiological rationale and a real sleep quality benefit, and a vendor claim that either prevents cardiac events should be met with a request for the citation and a look at what the study actually measured.
• The two timestamps that bracket turnout are frequently unreliable, since the start may be a CAD write rather than the moment the tone sounded and the end is usually a human pressing a button, so an alerting system that reports its own tone time and an AVL departure event give you a more honest measurement.
• Report turnout at the 90th percentile and split it by call type, station and time of day, because an average conceals the long calls and erases the physical differences between stations that are the fixable part.
• Redundancy is counted in independent failure domains rather than in products, so a station appliance, a phone app and a dashboard that all depend on one CAD interface and one network are a single path.
• The backup alerting path only counts if it is named in the SOP, owned by a named role, tested on a schedule with the result logged, and exercised by sending a real page to a real station rather than by checking for a green light.