Childhood Immunisation Reminders: AI-Drafted Letters and Texts

Use AI to draft the wording only: a first reminder text, a second reminder, and a final letter, each with placeholders for the child's name, date and booking link. Build the recall list and the overdue check inside your practice software or a spreadsheet, so identifiable data never reaches a chat tool.

Most templates stop at the letter. The part that protects children is the step after it: a weekly check that finds who has not booked, and a clear rule for the third reminder and for the family that says no. That step is what this tutorial adds. Everything here is administrative; clinical advice, consent conversations and decisions about whether to vaccinate stay with clinicians.

Follow me on Instagram@sagnikteaches

What AI does here and what stays in your systems

Split the job in two. One half is language, which AI handles well: tone, length, reading level, translation drafts. The other half is data, which belongs to the practice: who is due, when, by what route they prefer to be contacted, whether they have opted out. Keeping the halves apart is the whole safety design.

Connect on LinkedInSagnik Bhattacharya

If your practice software cannot produce a due-and-overdue report, that is worth fixing before any wording work. The report is the engine; the letters are the paint.

Subscribe on YouTube@sagnikteaches

The recall list: what to ask your software for

Whatever system you use, ask for a report with these fields, filtered by the schedule your practice follows: internal record number (not the name, on the working copy), age in months, the scheduled visit due, due date, last contact date, preferred contact method, and an opt-out flag. Sort by days overdue, longest first.

Agree the definitions in writing, because they change what the list says. "Due" means the visit window has opened. "Overdue" means a set number of days have passed beyond the window, for example 14. "Lost to follow-up" means three reminders have gone and no booking has been made. Three clear categories mean three clear actions, and your team will not argue each week about which pile a child belongs in.

Drafting the reminder sequence

A sequence of three touches works well in most small practices. Draft all three in one go so the tone stays consistent, and so the later ones do not contradict the first.

Draft a three-step immunisation reminder sequence for a family
practice. Placeholders: [PARENT/CARER NAME], [CHILD FIRST NAME],
[PRACTICE NAME], [BOOKING LINK], [PHONE], [CLINIC HOURS].
Step 1: SMS, under 160 characters, friendly.
Step 2: SMS, under 160 characters, a different phrasing.
Step 3: a letter of about 120 words with a clear next step.
Rules: say it is time for a scheduled immunisation visit; do not
name any vaccine, age or condition; no guilt, no pressure, no
statistics; include an opt-out line in each; plain English at a
reading age of about 10; offer a phone route for questions.

An illustrative first SMS: "Hi [PARENT/CARER NAME], [CHILD FIRST NAME] is due a scheduled immunisation visit at [PRACTICE NAME]. Book: [BOOKING LINK] or [PHONE]. Reply STOP to opt out." A first draft often comes back longer than the limit you set, so count it: this one is 156 characters with the placeholders in place, and real names and a long booking link change the total, so test the merge on your longest values. The second SMS leads with the phone number and a line about slots at different times of day. The letter adds the clinic hours, says how long the visit usually takes (a figure the practice must supply), and notes that a clinician is happy to answer questions beforehand.

Check each text against two rules. Would the text embarrass a family if read by someone else in the house? And does the text say anything that might not be true, such as "no appointment needed"? Delete anything you did not supply.

Main example: a practice with 420 children on its schedule

Take an illustrative practice with 420 children currently within the age range for scheduled visits. In a typical month, about 35 are due. The practice manager finds, by looking at her own records, that around 9 of those 35 never book without a nudge, and that reminders go out when someone remembers.

She builds the routine in four steps. On the first Monday of the month, the software produces the due list for the next 6 weeks. On the same day, step 1 goes out to those 35 through the preferred channel. Two weeks later, the software lists the children with no booking, which in this example is 17, and step 2 goes to them by the other channel. A further two weeks on, the list is 9, and step 3 goes by letter. At six weeks, anyone still without a booking, say 5 children, moves to a nurse or receptionist for a phone call, with a note of the outcome.

If each hand-written reminder used to take 3 minutes and the sequence now runs from templates, the saving on drafting is 35 first texts, 17 second texts and 9 letters, or 61 messages at 3 minutes each, which is about 3 hours a month before the system does the merging. More important than the minutes, no child waits on a missed list. The practice measures progress by how many are booked at the end of each cycle, not how many texts were sent.

The overdue tracking sheet

If your software cannot track reminders by itself, use a restricted spreadsheet on the practice's own drive, not a personal one, with access limited to staff who need it. Keep the working copy to internal record numbers rather than names where you can.

A simple formula counts the days since the window opened. With the due date in column C, the booked entry in column G and the opt-out or declined note in column H, use =IF(OR(LEFT(G2,3)="Yes",H2<>""),"",TODAY()-C2), so booked, opted-out and declined rows drop out of the count. The due date must be a real date, not text, or the sum fails. Sort the sheet by that column each Monday and work from the top. Add conditional formatting to colour rows red above 14 days, the overdue threshold agreed earlier, and amber above 7. This is plain spreadsheet logic, and you do not need AI for it, though you may use a chat tool to explain a formula to yourself using a fake sheet with no real data.

Before and after: a third-reminder letter

The last automated touch is the one practices write worst. An illustrative first draft reads: "Our records show that your child is overdue for immunisation. It is very important that you book immediately to protect their health." It scolds, it makes a health claim the practice has not sourced, and it uses the word "immediately", which families read as a threat. Here is a rewrite after review.

Dear [PARENT/CARER NAME], we wrote a few weeks ago about [CHILD FIRST NAME]'s scheduled immunisation visit and have not yet heard from you. If it would help, we can find a time that suits, including early or late in the day. You can book at [BOOKING LINK] or ring [PHONE] between [CLINIC HOURS]. If you have decided not to go ahead, or you have questions, please tell us so that we can stop these reminders and talk it through whenever you like. Kind regards, [PRACTICE NAME]

The rewrite makes an offer rather than a demand, mentions the opt-out in a kind way, and invites questions, which is where a clinician can help. It also contains no clinical claim for anyone to challenge.

A Monday routine that takes twenty minutes

A routine only works if it is small. Here is a version a receptionist can run with a checklist taped to the monitor.

1. Run the due-and-overdue report and save it to the restricted folder, never to a desktop.
2. Send reminder 1 to everyone newly due, by their preferred channel.
3. Send reminder 2 to those who got reminder 1 at least 14 days ago and have not booked.
4. Send reminder 3 by letter to those who got reminder 2 at least 14 days ago.
5. Make the calls for those who got reminder 3 at least 14 days ago, and note outcomes in the sheet.
6. Update the opt-out and declined columns from replies received over the week.
7. Tell the clinical lead the count of children on the phone-call list, and any with a note of concern.

A tick-box record of this checklist each week is evidence that the routine happens, which matters if anyone asks later how your practice handles recalls.

A mistake that happens: the message that reached the wrong parent

An illustrative case: a child's records list both parents, who have separated, and the reminders go to a phone number belonging to a parent who should not receive information about the child. The text names the child and the practice. The damage is real and entirely avoidable. Keep a "do not contact" or "contact restricted" flag in the practice system, and make the recall report exclude flagged records. Ask the clinical lead and the safeguarding lead to own the flag, since the facts behind it are not for the receptionist to judge. When you merge, check the sender number and the recipient field on a test record before the full run, and always send the first batch of each cycle to two staff phones to see exactly what a family sees.

What the messages cost

Text messages through a practice-software add-on or a messaging provider are usually charged per message, and the price depends on the provider and the volume, so check your own contract rather than trusting a figure from an article. As an illustration, if a provider charged 5 cents a text, the 35 children in the main example would cost under $2 in a cycle for the first step. Even adding the second step and a letter at the cost of paper and postage, the spend is small beside the value of a child not being missed. The tutorial on what reminder texts really cost covers per-message pricing and the extras that raise the bill. Postage and print are the larger costs for the letter stage, which is another reason to reserve letters for step 3.

Using AI to explain the sheet to a new colleague

A new receptionist will take a week to feel comfortable with the routine. Ask a chat tool to turn your written procedure into a one-page training guide, with the sample data replaced by invented children whose names and numbers are obviously fake, such as "Child A" and "Record 0001". Add a five-question quiz: what makes a child overdue, what happens after the third reminder, how an opt-out is recorded, who handles a declined family, and what to do if a text is sent to the wrong number. Checking the answers with the practice manager takes ten minutes and prevents most early mistakes. Keep the guide in the shared policy folder with a review date.

Where the clinical line sits

Admin staff and AI should not answer questions such as whether a child can have a vaccine while unwell, what a reaction means, or whether a delayed visit matters. Put a line in the sequence and the phone script that routes any such question to a nurse or doctor, with a promised call-back time the practice can keep. A template that stays out of clinical territory is also easier to approve, because the clinical lead has to read the logic and not every sentence.

Channel choices, opt-outs and the rule for stopping

Different families prefer different channels. Record the preference at registration and at each visit, and use it for the first reminder; use a different channel for the second so you reach people whose number or address has changed. Letters still matter, because not every family has a working mobile number or checks email.

Opt-outs need a rule that everyone knows. A reply of STOP removes the number from text reminders at once, and the receptionist notes the date. It does not stop the letter, which may be a separate care communication under your policies, so your adviser should confirm which of your messages count as reminders for care and which count as marketing. Never use a vaccination reminder to promote other services, and keep promotional wording out of the template entirely.

The stopping rule is as important as the starting rule. After the third touch, the practice moves from automation to a person. If a family has declined, the clinician records it, the automated sequence stops for that vaccine, and the conversation continues in person at a routine visit. An automated message to a family that has made an informed decision does harm.

Tone: how to ask without pressure

Parents respond to calm, practical messages. Ask the tool for three tone variants of the same text: neutral, warm and brief. Read each aloud. Remove any sentence that scolds ("you have not yet booked"), implies fear, or uses statistics you have not verified. Do not let the draft include health claims; clinical information is for clinicians, and sources such as your health authority's materials can be linked, but check the link works and that it is a page you have read. Keep the message to three jobs: it is time, here is how to book, here is how to ask a question.

Cases that need a different response

A family who has moved or changed number

Bounced texts and returned letters are data. Add a column for "contact failed", and route those children to a phone call or a check of the registration details. Do not keep resending to a dead number.

A child due several visits at once

Combined recalls confuse families. Ask the tool for a version that says "a visit is due" once, and lets the receptionist explain at booking. Do not list several items in the text.

A parent who replies with a question

Route all replies to a human. If your system auto-responds, the auto-response should say that the practice will reply within a set time and give a phone number for urgent matters, with no clinical content at all.

A translated reminder

Draft translations with AI if you need a first version, then have a fluent colleague or professional translator check them. A wrong date or phone number in a translated text is worse than none. For wider paediatric work, the tutorial on vaccine recalls and school forms goes further.

Checking that the routine works

Each month, look at four numbers from your own records: how many children were due, how many booked within six weeks, how many reached the phone-call stage, and how many opted out or declined. Compare month to month. If bookings rise after the first reminder but flatline after the second, the second message probably adds little and the phone call should come sooner. If opt-outs spike, read your wording again. If the same families appear on the phone-call list every cycle, ask whether the channel, opening hours or booking route is the barrier rather than the message.

Also audit the AI side. Once a quarter, read every template for anything that has crept in: an offer, a vaccine name, an unsupported claim. Store the approved wording in one shared document with a version date, and keep names and numbers out of it. The related reading on recall, reminder and review flows for doctors shows how the same structure works across a whole practice, and the tutorial on reducing no-shows with reminders covers the appointment itself once a family has booked.

If you want help building the sequence and the weekly check around your own software, that is what my AI implementation consultation covers; start at the contact page.

Questions practices ask about immunisation reminders

Should the reminder name the vaccine?

Usually no. Say it is time for your child's scheduled immunisation visit and give the booking route. Names of vaccines, ages and conditions can be read by anyone who opens the letter or sees the phone, so keep specifics for the appointment, unless your own policy and consent say otherwise.

Can I paste the overdue list into ChatGPT?

No. A list of children, dates of birth and due dates is health data about minors. Use AI only for the wording with placeholders, and run the overdue logic in your practice software or a spreadsheet that stays inside your own systems.

How many reminders are too many?

Three touches over about six weeks is a sensible ceiling for a small practice: a first message, a second by a different channel, and a final letter or call. Past that, move to a person making a phone call. Always honour opt-outs and record them straight away.

What if a family has chosen not to immunise?

Record the decision, stop automated reminders for that vaccine, and let a clinician hold the conversation. A repeated automated message to someone who has made an informed choice damages trust. Review the decision at routine visits, in person.

Further reads

• Can AI Handle Vaccination and Check-Up Reminders for a Vet? — The same reminder logic applied to animal practices.
• How a Clinic Can Use AI to Cut Missed Appointments — Cut missed appointments with reminders and rebooking.
• AI Chatbot for Healthcare: A Small Clinic's Data and Consent Checklist — A data and consent checklist for patient messaging.
• How Dental Practices Use AI for Recalls and Appointment Reminders — How a dental practice runs recalls, for comparison.
• How a Small Clinic Can Use AI at the Front Desk Safely — Using AI at a clinic front desk without losing trust.
• AI in General Practice UK: Telling Patients and Updating Privacy Notices — Draft patient notice, privacy-notice paragraph and poster wording for a GP practice using AI scribes or phone tools, with a summary of the NHS guidance.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: OpenAI help centre article on data controls and OpenAI's enterprise privacy page (checked 4 November 2026); the practice's own immunisation schedule and consent rules, which must be confirmed locally.