How to Automate Patient Follow-Ups: The Five Triggers a Philippine Clinic Actually Needs

Open your appointment book to any week from six months ago. Go down the page and ask one question about every name on it: did anything happen after that person walked out?

For most clinics the honest answer is a reminder before the visit, and then silence.

The molar came out on a Tuesday and nobody checked on Wednesday. The patient who was told she needed three visits came for the first one in March. The man due for a cleaning last November is still due, and he is not avoiding you, he just has not thought about it since.

None of that shows up as a loss, because there is no empty slot with a name against it. The appointment was never made, so nothing looks wrong.

A small clinic can close most of that gap without hiring somebody whose whole job is remembering, which is what automating follow-ups actually buys you. The trouble is that it gets sold as one thing when it is really five. Different events fire them, they land on different channels, and each one needs its own rule about when to stop.

The short version

Eight things, in the order I would do them.

1. A follow-up is defined by the event that fires it, not by the message. Write your triggers down before you write a single word of copy.
2. Patient follow-up automates well. Lead follow-up mostly does not, because Meta requires a human in the only window you have. They are different jobs and one tool rarely does both.
3. Almost every patient follow-up comes due after Messenger has closed, so plan for SMS. That means no links, a registered sender ID, and a message that works without anybody tapping anything.
4. Start with the check-in after the visit. It is the only one of the five with clinical evidence behind it, and it is a single message.
5. Set the recall interval per patient, not per clinic. Six months for everybody is convention, not evidence.
6. Write your stop rules before your messages. Automation is what turns one missed cancellation into six months of texts to somebody who rebooked in February.
7. Run the whole thing in dry-run for two weeks and read every message it would have sent. You will find things.
8. Any reply describing a symptom goes straight to a person, and your system never answers it.

Most clinics get as far as buying something that sends messages, then discover that nothing in it knows when to stop.

Five follow-ups, not one sequence

Here is the whole thing on one page. The first column is the only one that matters at the start, because everything else follows from it.

Trigger, the event that fires it	Fires when	Channel actually open	Safe to automate	Stop after
Appointment completed	24 to 72 hours later	SMS. Messenger only if they messaged you in the last day	Yes	One message
Patient did not arrive	Same day	SMS	Yes	Two
Treatment started, next visit not booked	About a week, then about a month	SMS, and a phone call for high-value work	Partly	Two, then a call
Due for a check-up	The interval the dentist set for that patient	SMS	Yes	Three over three weeks, then leave it a year
Enquiry that never became a booking	Hours, then days	Messenger, then SMS	Only the first 24 hours	Two

That last row is a different job with different rules, and it needs a person typing for most of its life. We pulled it apart separately in how to follow up with dental leads. Everything else in this article is about people who have already sat in your chair.

Why patient follow-up automates and lead follow-up doesn't

This one decides everything else, and the imported guides skip it, because American clinics run both jobs on email.

Meta gives a business 24 hours to reply to a customer, and the Human Agent tag stretches that to seven days for a human being who is genuinely typing. Both clocks start from the patient's last message. Not from their appointment. From the last thing they sent you.

A lead follow-up lives entirely inside that clock, which is why so little of it can be automated. A patient follow-up is fired by something in your calendar, and by then the clock has usually run out. Plenty of your patients booked by walking in or by ringing the front desk, so there is no Messenger thread at all and the clock never started.

So patient follow-up is an SMS job, in almost every case, whether or not that is what you wanted.

Which turns it into a delivery problem, and Philippine delivery rules are not intuitive. Links in texts are dropped by the networks with no error of any kind. Alphanumeric sender IDs have to be pre-registered and generic ones get refused. We went through the whole lot, including how to test across Globe, Smart and DITO before you trust it, in how to automate dental appointment reminders.

Design every follow-up message on the assumption that nobody can tap anything. If it only works when the patient clicks a link, you have built something that will fail silently and look fine in your own logs.

Trigger one: the check-in after the visit

Start here. It is one message, nobody minds getting it, and it is the only follow-up in this article with a proper trial behind it.

Pippi and colleagues randomised 638 extraction patients in a 2018 study published in Medicina Oral, Patologia Oral y Cirugia Bucal. Half got a phone call at 24 hours and again at 72 hours, asking about pain, bleeding, difficulty swallowing, jaw stiffness and fever. Everybody was examined at suture removal on day seven. In the group that got the calls, 15.7% had a complication by then. In the group that did not, 30.7% did.

Read that carefully, because it is easy to oversell. A phone call does not prevent inflammation. It finds the problem on day one, while it is still small, instead of on day seven when the patient walks in carrying it. A separate randomised study on third molar surgery found that patients who got a follow-up call stuck to the post-operative instructions better than patients given the same instructions verbally and in writing.

Two honest caveats. That was a phone call from a clinician, and a text message is not the same instrument. It was also a surgical population, so nobody has shown the same effect for a routine cleaning.

What a text can do is get a symptom into your inbox on day one rather than day six. That is a smaller claim, and it is enough.

What the check-in has to contain

Four things, and it fits in one message.

1. Who you are, in the first line. A caring question from an unknown number reads as spam.
2. What they had done, named. "Your extraction" beats "your recent visit".
3. One specific question they can answer in a few words.
4. A number they can ring, and permission to ring it.

What it must not do is ask "how was your experience?" That is a survey, and a survey is a different job you are welcome to do later. The check-in is there to catch a problem early. Sounding warm is a side effect, not the point of it.

The rule that matters more than the wording

Some of the replies will describe symptoms. That is the point of sending it.

So decide now, before you switch anything on, what happens to a reply that says masakit pa rin po at nine in the evening. Two rules, and neither is negotiable.

The system never answers a clinical question. Not with reassurance, not with "that's normal after an extraction", not with a painkiller suggestion. That is diagnosis, and it is not the front desk's job whether the front desk is a person or a piece of software.

Any reply mentioning pain, bleeding, swelling, fever or a bad taste goes to a named human immediately, and that human knows they own it. If the clinic is shut, the reply says so and gives the number to call.

Get that wrong and an automated check-in is worse than no check-in at all, because you asked the question and then did not listen to the answer.

Trigger two: the patient who did not arrive

Same day, one message, and its only job is to make rebooking easier than explaining.

There is no scolding in it, no "we missed you", and no fee talk unless your policy already said so in writing before the appointment. We covered the whole no-show problem, including why an easy cancel route beats a fourth reminder, in how to reduce dental appointment no-shows.

The automation part is small. The trigger is a booking whose time has passed with nothing marked against it, which means somebody has to mark attendance the same day. Clinics that skip that step end up texting patients who were sitting in the chair an hour earlier, which is the fastest way to make the whole system look stupid.

Trigger three: the treatment nobody finished

This is where the money is, and it is the least automated thing in most clinics.

A patient is told she needs two fillings. She has one done, says she will book the second after sahod, and then life happens. Nothing in your appointment book records that a second visit was ever supposed to exist, so it is neither a cancellation nor a no-show. Nobody comes to tell you about an appointment that was never made.

I could not find a Philippine figure for how often this happens, and the case-acceptance percentages that circulate online come from American practice consultancies with no study attached. So do what you did with your no-shows and count your own.

For one month, whenever a dentist recommends work that needs another visit, write down four things: the date, the patient, what was recommended, and whether the next visit was booked before they left the building. At the end of the month you have a number that is yours, and the gap between "recommended" and "booked before leaving" is the thing you are actually automating.

That last column is the interesting one. The best fix for unfinished treatment is not a follow-up at all. Book the next visit at the desk, while the anaesthetic is still wearing off and the recommendation is fresh. Everything else in this section is what you do for the ones who slip past that.

For the ones who slip past, two touches:

• About a week later, an SMS naming the specific work and offering two real times.
• About a month later, one more. For braces, implants or anything over a few thousand pesos, make this a phone call from a named person, because the patient is usually deciding with a spouse and a text cannot answer what the spouse asks.

Then stop, and let them fall into your recall list.

One line of self-discipline here. Following up unfinished treatment is clinical housekeeping, and it stops being that the moment the message starts selling. Name the work, give the normal price, offer the times, and attach nothing else. The Code of Ethics for Philippine dentists prohibits promotional rates outright and holds everything you send electronically to the same standard, so a discount to get somebody back for their second filling is out on two counts at once.

Trigger four: recall, and the interval nobody sets

Almost every clinic in the country runs recall on six months for everybody. Six months is a convention, not a finding, and the evidence has been pointing elsewhere for twenty years.

NICE guideline CG19 says the interval between oral health reviews should be set by the individual patient's disease risk, choosing from 3, 6, 9, 12, 15, 18, 21 or 24 months for an adult, and 3, 6, 9 or 12 for anyone under 18. The 2020 Cochrane review by Fee and colleagues, covering two studies and 1,736 people, found little to no difference in tooth decay, gum disease or quality of life after four years between risk-based recall and six-monthly recall in adults, and rated that evidence high certainty. For children and adolescents the same review found the evidence too weak to say anything.

Now the caveat, because that is UK research from a population that attends regularly. Filipino attendance does not look like that. Figures reported from the Department of Health's 2018 National Survey on Oral Health put dental care seeking at 57.6% of Filipinos, with close to half of those visits prompted by toothache. I could not find the DOH report itself online, only reporting of it, so treat the exact numbers with some caution. The shape of it will surprise nobody who runs a clinic here.

Stretching a recall interval in a population that already comes only when something hurts is a clinical judgement, and it belongs to the dentist rather than to whoever configures the software. What the evidence does settle is the other direction. A blanket six months is not a safety floor, so a longer interval for a low-risk adult is not the reckless option it sounds like.

The automation part of this is one field. Your system needs a recall interval stored per patient, set by the dentist at the end of the visit, and your follow-up fires off that. If the only interval your tool understands is one global number, you have bought a mailing list rather than a recall system.

What to send once it fires, how to build the list, and the three-touch sequence that works here are all in how to get old dental patients back. No point repeating it.

What a follow-up message has to do here

Three things keep Filipino patients out of the chair, and the same DOH survey reporting puts them in this order: cost at 54.8%, fear at 22.8%, and a busy schedule at 14.3%.

Look at what that does to a standard follow-up template. "You're due for your check-up, please call us to schedule" answers none of them. It hands the patient the work of asking what it costs, hides how long it takes, and gives them nothing to say yes to.

So say the price, say how long it takes, and offer two specific times. That is one message, and it deals with all three obstacles in the order patients actually feel them.

Four more things that are specific to sending here:

• No links. They are dropped silently, and your own sending report will still say delivered.
• Register your sender ID, and put the clinic name in the first line anyway.
• Mirror the language they use with you. A patient who messages in Bisaya should not get a follow-up in formal English.
• Mind the peso sign. Writing the amount as a peso symbol rather than PHP pushes the message into a different encoding with less than half the room, and a text that billed as one starts billing as three. We counted eighteen templates that way in dental appointment reminder SMS templates.

Write the stop rules before you write the messages

This is the section missing from every follow-up guide I have read, and it is the one that decides whether your clinic looks organised or unhinged.

A person sending follow-ups by hand suppresses things without noticing. They see that Mrs Dela Cruz already rebooked, so they skip her. They know the family lost their father last month, so they leave that number alone. Automation knows none of that unless you tell it, and it never gets tired of being wrong.

Write these down first. Each one is a rule about when a scheduled follow-up must not send.

Stop rule	Why it exists	What it looks like in practice
Patient already rebooked	The commonest and most embarrassing failure	A future booking for that patient cancels every pending follow-up on the old one
Patient replied STOP, or asked in person	Required, and it is the one people honour in the thread and forget in the next campaign	The suppression lives on the patient record, not in one staff member's memory
Patient replied at all	A live conversation must never be interrupted by a scheduled message	Any inbound reply pauses the sequence until a person closes it
A complaint or dispute is open	Nothing reads worse than a cheerful recall in the middle of a complaint	A flag on the record that only a person can clear
Patient has died, or the number was reassigned	It happens, and somebody at the clinic usually knows	Manual flag, permanent, checked before any list send
Two follow-ups due in the same week	The patient does not experience your triggers separately	One message wins, the more urgent one, and the other waits
Sending window	A 6am recall is a bad first impression	Nothing sends before 8am or after 8pm, ever

The collision rule is the one that catches people out. A patient can easily be due a post-visit check-in, an unfinished-treatment nudge and a recall inside the same fortnight. To your system these are three unrelated jobs. To her, it is a clinic that has texted her three times in ten days.

Consent, and why automating it raises the stakes

Your patient list is health information, and under the Data Privacy Act that is sensitive personal information with its own narrower set of lawful grounds. There is no legitimate-interest route for it the way an ordinary shop has for a mailing list. We worked through what that permits and what it rules out in the recall guide, and the short version is that a message about the patient's own care sits on very different ground from a message advertising this month's whitening price.

Automation does not change the law. It changes how many people a mistake reaches. Somebody at the front desk who forgets a request to stop sends one more message. A system that never learned about it sends forty over the next year, punctually, and each one reminds the patient that they asked and you did not listen.

So put a way out in the message, honour it the day it arrives, and honour it on the record rather than in the thread. Ours is set out in our privacy policy if you want the shape of one to copy.

Two weeks in dry run

Before a single message leaves the building, run the whole thing with sending switched off and have it write down what it would have sent, to whom, and why.

Two weeks. Then sit with the list and read it.

Four things to look for:

• Messages to people who already rebooked, which means your attendance marking is not keeping up.
• The same patient appearing three times, which means your collision rule is missing.
• Recalls firing on a global interval instead of the one the dentist set.
• Anything addressed to a person the front desk knows should not be receiving it.

Finding any of that in a spreadsheet costs you nothing. Finding it after it sent costs you an apology to somebody who was already unhappy.

What to actually send

Adapt the wording, keep the shape. Fill in the placeholders and count the message before you trust it.

The day after a visit:

Sundial Dental: Hi Ms Reyes, si Jen po ito. Kumusta po ang extraction kahapon? Pa-reply lang po kung masakit pa, may dugo o pamamaga. Tawag po sa 0917 123 4567 kung urgent.

Same day, after a no-show:

Sundial Dental: Hi Ms Reyes, na-miss po namin kayo ngayong 2:30. Meron pa po kaming bakante Thursday 10:00 AM or Saturday 3:00 PM. Reply lang po kung alin.

A week after treatment was left unfinished:

Sundial Dental: Hi Ms Reyes, hindi pa po naka-book ang pangalawang filling ninyo. 45 minutes lang po, PHP 1,200. Bakante po kami Tue 11:00 AM or Fri 4:00 PM.

A recall, at whatever interval the dentist set:

Sundial Dental: Hi Mr Lim, due na po kayo sa cleaning. 45 minutes, PHP 1,500. Bakante po kami Wed 9:00 AM or Sat 2:00 PM. Reply STOP kung ayaw n'yo na po ng ganitong text.

Notice what none of them do. None of them apologises for messaging or says "just following up", and none carries an offer, a discount or a link. Each one names the clinic, names the treatment, and ends with something easier to answer than to ignore.

Common mistakes

Buying the tool before writing the triggers. You end up configuring your clinic around whatever the software calls a campaign, which is somebody else's model of your job.

Automating the check-in without staffing the replies. Asking a patient how they are feeling and then not reading the answer for two days is worse than staying quiet.

One recall interval for the whole practice. Convenient for the software, wrong for the patient, and unsupported by the evidence in either direction.

Letting the system speak first about anything clinical. Administrative questions, scheduling, hours and prices are fine. Symptoms are not, and that line has to be a hard rule rather than good behaviour.

No suppression list. The fastest way to convert a good clinic into a nuisance.

Following up in a channel that closed weeks ago. Messenger will not deliver a six-month recall, and if you found a way to force it through, it would still be the wrong channel.

Measuring messages sent. Nobody is paid in messages.

How to tell whether it worked

Three numbers, counted monthly, and none of them is an open rate.

Appointments where the next visit was booked before the patient left. Move this one first, because every point you gain here is a follow-up you never have to send.

Recalls that become a booked appointment inside 30 days of firing. Track it per trigger rather than as one blended figure, or a healthy post-visit check-in will hide a recall list doing nothing.

Complaints, opt-outs and "please stop" replies. If this is climbing, your cadence is wrong, and no conversion number makes up for it.

Give it three months before you judge any of them. A recall system set to a nine or twelve month interval has barely cleared its throat by week six.

Where OtterFlow fits

Everything above is doable by hand. The reason clinics do not do it by hand is that the triggers fire at exactly the moments the front desk is busiest, and the recall ones fire months after anybody remembers agreeing to them.

OtterFlow sits in front of whatever clinic system you already run and takes the conversation side of it. Bookings get an SMS confirmation and reminders on the schedule you set, a check-in goes out after the visit, and a no-show gets a rebooking link. Dormant-patient reactivation, unfinished-treatment follow-up and recall runs sit in the same place, so the triggers in this article are configured once instead of remembered every morning.

The parts that map onto the harder sections above are the ones worth asking any vendor about, ours included. Patients reply to follow-ups, and the reply arrives in whatever language they wrote in, so the assistant mirrors English, Taglish or Bisaya rather than answering everybody in one register. It answers only from information the clinic approved, and when it does not have an answer it says so and hands the conversation over instead of inventing one. Anything clinical hits a hard guardrail: it does not diagnose, does not advise, and escalates. Your staff can take any conversation over completely, and the assistant stays quiet in that thread until you hand it back.

Facebook and Instagram Messenger connect through our messaging API, which we wire up for you during setup, so an enquiry arriving on your Page lands in the same flow as the ones from your website.

We are a small team in Cebu, and we set it up around your clinic's own services, prices, hours and booking rules rather than handing you an empty dashboard. Pricing is public: ₱5,000 once when we launch you, then ₱2,500 a month. If you want to see how that sits against everything else on the market first, we wrote up what dental clinic software actually costs in the Philippines, our own numbers included.

Questions clinics ask

How soon after the visit should the check-in go out?

The day after, for anything surgical. The trial that found early complications called at 24 hours and again at 72. For a cleaning, one message the next day is plenty and a second is fussing.

Can we automate the post-visit check-in on Messenger instead of SMS?

Only if the patient messaged you inside the last 24 hours, which is unusual for somebody who has just been in the chair. Assume SMS and treat a live Messenger thread as a bonus.

What if the patient replies with a clinical question?

A person answers it, and the automation stays out of it. Build the handover before you build the message.

Do we need clinic software before any of this is worth doing?

No, but you need attendance marked the same day and a recall interval written somewhere per patient. A spreadsheet does both. A follow-up system built on a book nobody updates until Friday will text people who came in on Tuesday.

Is a phone call still better than a text?

For anything expensive or anything clinical, yes, and it is not close. For a routine recall or a cleaning check-in it is heavy-handed and harder to answer than a text. Use calls where the conversation genuinely has two sides.

How many follow-ups is too many?

Count from the patient's side rather than from your triggers. If the same person would get more than two messages from you in a fortnight without having replied, that is too many, whatever the reason.

Start with a sheet of paper

None of this starts with software. It starts with somebody deciding what fires a follow-up and what ends one, and in most clinics that has never been written down at all.

So take an afternoon and write the five rows out: the event, the timing, the channel, and the rule that stops it. That will tell you more about what your clinic is losing than any demo will, including ours.

If you do that and find the triggers firing faster than anyone at the desk can keep up with, particularly the ones that land in the evening, that is the point where automating them starts to pay for itself, and we are happy to show you what it looks like running on your own clinic's services and availability.