Automating patient follow-up in a fertility clinic: what to automate, what to keep human

Every fertility clinic loses patients it never meant to lose. Not to a competitor, and rarely to a decision: they simply stop hearing from the clinic, or the clinic stops hearing from them, and nobody notices for weeks. Follow-up is the work between the moments a clinic is built around, and most clinics run it on memory.

This is a guide to automating that work sensibly: what the evidence says, which parts a machine should do, which it must not, and how consent shapes all of it.

Where follow-up breaks

Three cases keep coming up.

The enquiry that never gets a call back. Someone fills in a web form on a Sunday evening, or calls when the line is busy, and the person who would answer is off, on the phones, or assumes someone else has it. By Thursday the enquirer has spoken to another clinic.

The faxed referral that sits in a queue. A family physician faxes a referral. The patient, who was told "the clinic will call you", waits. Two weeks later she calls to ask whether it arrived, and nobody can tell her without going to look. The clinic has not lost the referral; it has lost the time, and often the patient's confidence, we cover automating referral intake itself in a separate guide.

The patient who goes quiet after a failed cycle. There is a follow-up consultation to book. It is not booked. Weeks pass. The clinic does not know why, because the last contact was the hardest conversation in the whole process and nobody wants to be the one who calls again.

These are failures of memory, hand-offs and timing, exactly the things software is good at.

What the evidence says

Patients leave for reasons that include the clinic. A 2012 systematic review in Human Reproduction Update covered 22 studies and 21,453 patients from eight countries. The reasons patients gave most often were postponement of treatment (39.18%), physical and psychological burden (19.07%) and relational and personal problems (16.67%); but organisational problems (11.68%) and clinic problems (7.71%) together account for close to a fifth, and the authors concluded that treatment burden "should be addressed by better care organization and support for patients"[1].

Much of the loss happens before treatment starts. A Dutch cohort of 1,391 couples referred to a secondary-care fertility service found that 319 dropped out, 76.8% of them on their own initiative, and about half stopped before any treatment had begun[2]. That stretch, referral to first treatment, is exactly the one follow-up is meant to cover.

Speed matters more than anyone expects. A Harvard Business Review study audited how 2,241 US companies responded to a web-generated test enquiry, and 23% never replied at all. In a separate analysis of sales-lead data, the same authors found that firms which made contact within an hour were nearly seven times as likely to reach a meaningful conversation with a decision maker as those that waited even an hour longer, and more than 60 times as likely as those that waited a day or more[3]. A fertility patient is not a sales lead, but the mechanism (the enquirer is still at the keyboard, still deciding) carries over.

Reminders work, and a text is as good as a call. A Cochrane review of eight randomised trials with 6,615 participants found that text-message reminders improved attendance at healthcare appointments compared with no reminders, moderate-quality evidence from the seven trials that could be pooled, risk ratio 1.14 (95% CI 1.03 to 1.26). Across the review, attendance was 67.8% with no reminder, 78.6% with a text and 80.3% with a phone call, and in the two trials that costed them, a text cost 55% and 65% less per attendance than a call[4]. The evidence is rated low to moderate, and none of the trials reported on harms such as loss of privacy, a fair warning where a text about a "cycle" can land on a shared kitchen table.

What to automate

The rule we use is simple: automate the parts of follow-up that are about timing and memory, and keep the parts that are about judgement and feeling.

The first response. Every enquiry should get an acknowledgement within minutes, at any hour, that says what happens next. It does not need to answer the clinical question; it needs to stop the clock the enquirer is running in their head. An AI phone agent that answers the clinic line and takes a callback number, reading it back before queuing it for staff, does this for the one channel where "we'll call you back" has always been the weakest promise.

Referral acknowledgement. The moment a referral is confirmed, the patient should hear that it arrived and what happens next. It is the cheapest item on this list.

Reminders. On the patient's preferred channel. The Cochrane figures above are the case for them.

Consent-gated follow-up by text or email. A short, clinic-authored sequence (a text, a wait, a call, a booking offer) for patients who have opted in, that stops the moment they book, opt out or say they do not want to proceed.

Logging every attempt. Every call, text and message, whether a person or the system sent it, should land on one record, in order, with the outcome. "Left a message on Tuesday" is only useful if the person picking up the file on Thursday can see it.

What to keep human

Clinical questions. "Should I stop the progesterone?" is not a follow-up task and must not be answered by a sequence, an auto-reply or an agent. It should be routed to a nurse, quickly, with the original wording intact.

Distress. A patient who writes "I don't think I can do this again" has not answered a menu. Anything that reads as distress should stop the automation on that person and put a coordinator in front of it. The safe failure mode is silence followed by a human, never a cheerful automated nudge.

Anything after a negative result. The days after a failed cycle are the point where a patient is most likely to leave and least able to tolerate being processed. A gentle reminder a week or two later may be welcome, but the first contact after bad news should come from a person the patient recognises, and its timing should be a clinical decision, not a timer.

Consent is the gate, not a checkbox

Follow-up by text or email that carries anything about a patient's care is only acceptable where the patient has opted in to that channel. A first message that carries nothing private (that a referral has arrived, and a question about how the patient would like to be reached) is how that consent gets asked for; everything private waits for the answer. In our own platform, before the assistant shares any personal health information by phone or text, two separate things must be true: the patient's identity has been verified, typically by date of birth, and consent is on file for that channel. A patient who has proved who they are but has not consented is invited to opt in rather than shown anything private.

Consent is tracked per channel with a full audit history and can be captured three ways: the patient switches it on in the portal, confirms by text (a double opt-in), or a staff member records it from the patient's record. Replying STOP to any text both stops further texts and withdraws text consent; clinic email carries an unsubscribe link. An email the clinic starts is refused outright without consent on file. Two distinctions matter: booking a nurse intake call does not grant consent to receive health information by text or phone, and unsubscribing from email is not the same as withdrawing consent or withdrawing from the referral, three statements, recorded separately.

How Fertiligent does it

Follow-up runs on one prospect pipeline. Every intake channel (web chat, phone, a confirmed referral, text, email, an import) lands as a prospect in the same list, and each prospect records where they came from. The record carries a timeline of every touch across every channel, with system-sent texts logged automatically, and shows when the patient read a secure message, so an unanswered message can be told apart from an unopened one. That single record is the same principle behind our approach to data management for fertility clinics.

A prospect nobody has picked up is flagged overdue once the referral has sat longer than the clinic's turnaround target. An Awaiting reply toggle shows the people the clinic has contacted who have not answered since. Logging a contact attempt records the outcome (no answer, left a message, reached) on the timeline without moving the funnel stage.

Automated outreach is a sequence the clinic authors itself (text, automated call, wait, consent question, preference question, booking offer) started by a confirmed referral and run on business days. The preference question confirms receipt of the referral and asks how the patient would like to be reached. The sequence stops on its own when the patient books, opts out or says they no longer wish to proceed, and prospects it could not reach surface on a worklist so staff can follow up directly rather than let it go quiet. Appointment reminders go out on the patient's preferred channel.

The human boundaries are built in. A withdrawal typed in the patient's own words ("not interested", "stop contacting me") records nothing by itself: outreach on that person is put on hold and a coordinator is asked to confirm, because withdrawing is a care decision the referring physician has to hear about. Inbound messages are assessed as safe, needs review or escalate, anything sensitive goes to a person, and every AI action is recorded. Every call and text the assistant handles is logged in one worklist and, once matched to a person, on their record: the same way our AI medical scribe for fertility clinics captures calls and messages.

None of this makes a clinic warmer. It makes sure the warmth the clinic already has is not wasted on a patient who has stopped hearing from it.

See it yourself: take a referral through to a booking in the live demo, a fictional clinic with synthetic data.

References

• [1] Why do patients discontinue fertility treatment? A systematic review of reasons and predictors of discontinuation in fertility treatment, Human Reproduction Update (2012)
• [2] When and why do subfertile couples discontinue their fertility care? A longitudinal cohort study in a secondary care subfertility population, Human Reproduction (2009)
• [3] The Short Life of Online Sales Leads, Harvard Business Review (2011)
• [4] Mobile phone messaging reminders for attendance at healthcare appointments, Cochrane Database of Systematic Reviews (2013)

Related:

• Patient-Support Infrastructure: How Fertility Clinics Scale Without Adding Headcount
• AI in Fertility Clinics: Benefits, Implementation, and ROI

Sergei Gorlovetsky, CEO, Fertiligent