Which appointment reminder method works best?

Reminder systems are sold on a reduction in no-shows and evaluated against published ranges that are almost impossible to use, because the ranges pool methods that do different jobs. A postcard, an automated voice call, a one-way text and a two-way text are not four intensities of the same intervention. They differ in one property that decides most of the outcome: whether the patient can respond through the same channel, at the moment they read it, without calling the practice.

That property matters because of what a reminder can and cannot do. It cannot make somebody attend an appointment they have decided against, and the modelling that assumes it can is where the disappointing results come from. What it can do is surface that decision early enough to be useful. A patient who receives a one-way message and is not coming does nothing, and the slot is lost silently. The same patient who can reply with a single word releases the slot, and the practice now has a recoverable gap rather than an empty room.

Which reframes the comparison entirely. The right question is not which method reduces no-shows most, but which method converts the largest share of non-attendance from silent to announced, and how much notice it buys. Notice is the variable that determines whether the slot can be refilled at all — a release the evening before is worth far more than one an hour ahead, and both are worth more than a patient who simply does not arrive.

The cost side deserves the same scepticism as the effectiveness side. Methods differ materially in cost per appointment, and the cheapest per message is not the cheapest per released slot. A practice that sends three messages per appointment through a channel with a low unit price may be spending more per recovered slot than one sending a single message through a channel that permits a reply. The denominator that matters is the slot, not the message.

Before any of this is chosen on effectiveness, a constraint applies. Automated reminders to a mobile number sit under the rules implementing the Telephone Consumer Protection Act, at title 47 of the Code of Federal Regulations, part 64, and the treatment-related allowance carries conditions on frequency, content and how an opt-out is honoured. That is not a reason to avoid automated reminders — it is the reason the method is chosen inside a boundary rather than purely on measured effect, and the reason an opt-out has to work on the first request rather than at the next billing cycle.

The evaluation to run is a straightforward one and takes a quarter. Hold the reminder method constant per clinic, count released slots separately from attended appointments and from silent non-attendance, and record how much notice each release gave. That produces a figure specific to your patients, which is the only kind that can be acted on, and it costs nothing beyond deciding to count three things instead of one.