Designing a Triage Form That Sorts Real Dental Emergencies

Most practice triage forms are built backwards. Someone sits down meaning to capture everything the dentist might want to know before the patient walks in, and out comes nineteen fields: a tooth chart, a 1 to 10 pain slider, a duration dropdown, hot or cold sensitivity, full medical history, what they’ve already taken for it. The form submits. It just doesn’t sort.

Sorting is the only thing a triage form has to do. Every submission has to land in one of three buckets before it reaches a human: send to hospital now, see today, book into the next routine urgent slot. Three outcomes. You do not need nineteen questions to pick between three outcomes, and the practices getting real value from digital triage have worked out that six will do it.

What actually changes the bucket

Run the exercise honestly. Take your current form, print it, and go through field by field asking one question of each: if the answer flipped from best case to worst case, would the patient move buckets?

For most fields, no. Pain score is the clearest example. Ask a hundred people in pain to rate it out of ten and the answers pile up at seven, eight and nine. Somebody with a cracked cusp and somebody two days into a spreading buccal infection will both write 8. The distribution is compressed at the top, so the field carries almost no sorting information while costing you a completion step. Duration is similar, and slightly worse, because it is actively misleading: three weeks of grumbling pulpitis is not more urgent than a submandibular swelling that started yesterday afternoon, but a form that sorts on “how long” will put them in the wrong order.

The fields that do move the bucket are the ones that map onto genuine escalation criteria. There are about six of them, and they come straight out of guidance your clinicians already work to: SDCEP’s Management of Acute Dental Problems, the IADT trauma guidelines, and NICE’s head injury criteria for deciding who belongs in A&E rather than in your chair.

Here is the whole set of dental emergency triage questions that does the sorting work, written the way a patient in pain at 11pm can answer them:

Six screens, seven lines, and one of them is a follow-up that only appears when there’s swelling. That is the form.

The arithmetic on question ten

The trade is not free, and this is where practice managers usually push back: surely one more field costs nothing? It costs completions, and completions are the only reason the form exists.

Model it with your own numbers. A 6,000-patient mixed practice taking around 14 unscheduled contacts a day, roughly 300 a month. Suppose your six-question form gets completed by 80% of the people who open it. Every additional required field pulls completion down by a few points, which is the consistent direction of every form conversion study anyone has published, Formstack’s included. Say three points per field, conservatively. Go from six fields to fourteen and you are at 56%. On 300 monthly contacts that is 72 people who started your triage form, abandoned it, and phoned reception instead. Or worse, didn’t.

Those 72 are not evenly distributed either. The person who abandons a long form at 11pm is disproportionately the one in most distress, with the least patience for a tooth chart, holding a phone one-handed with a cold compress in the other. Long forms are a filter, and they filter out exactly the wrong people.

The counter-argument is that the extra fields save clinical time later. They don’t, much. A clinician who has the patient in front of them will establish tooth, character, duration and trigger in about twenty seconds of questioning and get better answers than the form produced. You are trading a reliable twenty seconds of chairside history for an unreliable eight fields that lose you a quarter of your submissions.

Worked output

This is what a well-built form hands to the front desk. Real practices wire this into whatever they already run, so the output lands in the right place: Dentally’s portal, Software of Excellence EXACT, or Carestream R4, usually via a form tool like Dengro or Zesty sitting on the practice site.

23:41  Submission #4471
DOB 03/11/1978   Tel 07### ######   Reg: YES (last seen 14/02/25)

Q1  Swelling closing eye / under tongue / hard to swallow?   NO
Q2  Bleeding after 20 min pressure?                          NO
Q3  Tooth knocked out or moved?                              NO
Q4  Blackout / vomiting / memory gaps since injury?          N/A
Q5  Chemo, bisphosphonate, denosumab, anticoagulant?         NO
Q6  Kept awake / analgesia not working?                      YES

ROUTE:  URGENT-TODAY
SLOT:   16:20 emergency (England: Band 1, £27.40, 1.2 UDA)
PHOTO:  1 attached
Completion time: 47s

Now the one that matters:

23:52  Submission #4472
DOB 22/06/1961   Tel 07### ######   Reg: NO

Q1  Swelling closing eye / under tongue / hard to swallow?   YES (swallowing)
Q2  Two fingers between teeth?                               NO
Q5  Chemo, bisphosphonate, denosumab, anticoagulant?         YES (alendronic acid)

ROUTE:  ESCALATE - DO NOT BOOK
ACTION: Call patient now. Direct to A&E / on-call maxfax.
        Form stopped after Q2. Remaining questions suppressed.
Completion time: 19s

Note what the second form did: it stopped. It did not ask which tooth, it did not ask for a pain score, and it did not offer a booking link. A triage form that keeps collecting data after it has identified an airway risk is a form that has misunderstood its job.

Photos, and where remote assessment stops

Ask for one photo, not five. A single intra-oral or extra-oral shot resolves a genuine ambiguity in a decent share of swelling and trauma cases, which is why tools like SmileMate from Dental Monitoring and the Toothfairy app are built around patient-captured images. Ask for five and you get one blurry photo of a cheek, because the person is in pain and gave up after the second attempt.

Be clear about what the photo is for. It is not a diagnosis. Photographic triage tells you whether a swelling is diffuse or localised and whether an incisor is displaced; it cannot tell you about the periapex, which is why radiograph AI (Pearl’s Second Opinion, Overjet, Videa) is a separate product category solving a separate problem. Conflating the two is the most common mistake in procurement conversations. If you want the fuller picture of what remote assessment can and cannot carry, our triage and remote assessment guide sets out the boundaries in more detail.

Testing it before you trust it

Don’t launch on faith. Pull the last 100 unscheduled contacts from your call log or NHS 111 referrals, have someone who wasn’t involved in building the form run each one through the six questions on paper, and compare the bucket the form produces against what actually happened to that patient.

You are looking for two error types, and they are not equal. A patient the form sent to an urgent slot who turned out to need a routine appointment costs you one wasted 1.2-UDA slot. A patient the form routed to “book next week” who ended up in A&E on IV antibiotics is a clinical governance incident. Set your thresholds accordingly: it is entirely reasonable for the form to over-escalate swelling, and you should expect a false positive rate on Q1 somewhere around one in four.

Watch for the specific failure that catches everyone. Patients answer “no” to the anticoagulant question because they think of apixaban as “a tablet for my heart”, not a blood thinner. Naming the drugs in the field helper text fixes most of it. Warfarin, apixaban, rivaroxaban, edoxaban, clopidogrel, on the screen, in plain type.

Then go and look at the questions you removed. Every one of them is now somebody’s twenty seconds at the chair, and that is a much better place for them than a field that 72 people a month never reached.