Design the patient workflow before you call it a registry

22 September 2026 · Guy Solan

When we design a medical registry, we do not start with the database. We start with the workflow a person actually walks: how they are invited, whether they qualify, what they agree to share, and whether the same questions come back later.

If you are setting one up for a device, a clinic pathway, or a study, these are the steps we go through. A form on day one does not tell you whether the knee still hurts in June, whether a patient can see three months after surgery, or whether the wound has healed since receiving the product. Leave that later check to chance and you will have a busy first month and a hole where month six should be.

Who this is for

This is for the person who has to show what happens to patients after a product or a procedure is already in use. Usually that is a medical device company, a pharma or biotech company, or a clinic that wants the same answers for its own pathway.

A product registry is normally built around one product and a defined group of patients. Here are some examples of the sort of thing that could be relevant.

The same design shows up across most of medicine: spine, dental, ear nose and throat, women's health, dermatology, diabetes devices. The details change, but the basic sequence is often similar.

Why you need one

Once a product is in wider use, effects can appear that the original trial did not capture. Regulatory bodies and the buyers of medical products know this. That is why they care about real-world data: what actually happens to patients once the product is in use, not only what the trial showed. Some of the people that data is for include:

While national registries can sometimes serve a purpose, often there is no device-specific data. That is why you might be in a position where you need to design your own registry. After setting up dozens of these, we've found that success usually comes down to getting these eight steps in the right order:

1. Invite them

We divide invitations into two routes, depending on whether you already have the patient's contact details.

2. Check they qualify

Our approach starts with a short screener. Before anyone spends time on long forms, we check the basics: their age, whether they have the condition or the symptoms you care about, and whether they have seen a clinician inside the window your protocol allows.

We design that as branch logic. The next question depends on the last answer, and a fail stops the path. They do not drift into consent by accident. You can count completed screeners and exclusions, although open QR recruitment cannot tell you how many people saw the invitation.

3. Take a name and a way to reach them

The goal here is data minimisation. We collect only the identity and contact details the follow-up actually needs - enough to reach them, but no more. It feels administrative, but it is how the later steps work at all.

Before asking for those details, explain why you need them and how they will be used. The consent in the next step then covers joining the registry and any sharing the participant agrees to.

Repeat questionnaires need a person to send them to. Reminders need a number or an inbox. If a care team is allowed to see the patient, they need a real identity, not a code you invented after the fact. Miss this and step 6 is a form with nowhere to go.

4. Consent, and decide who sees the name

We've found it's best to split e-consent into parts. One might cover the registry itself, while another covers sharing with the care team. They are separate because the answer can be different.

That choice decides who may see the identifiers. The registry still needs to hold them securely so it can send the follow-up.

5. Ask the one-off questions once

Gather the fixed facts at the start. Demographics, medical history, and BMI from height and weight are things you only need to ask once:

These describe who is in the registry. They are not the outcome.

6. Then the questionnaire you will ask again

This is where the actual tracking begins. We take a first reading (the baseline) using a validated outcome measure like EQ-5D, then set the schedule for the repeats. Validated specialty-specific scores are chosen by the protocol. There are thousands of these, across almost every specialty. The design job is to put the right one onto a schedule you will actually repeat.

Copyright usually sits with someone else. If a fee is owed, it is owed to the owner of the questionnaire. Putting the questions on a form is not the same as owning them. The practical version of that, for knees, is in which knee questionnaire to use.

7. Remind them on a schedule

A schedule only works if you stick to it. We've found that "one and done" messages don't work; we keep to the agreed reminder sequence. Quarterly, weekly, or every day for ten days and then weekly. There are a lot of shapes, and they do not have to share one timetable. A schedule is not one message. If they have not done it, we remind them again. Silence is not an answer, and missing the first one is not a reason to stop.

That only works if we know what they have actually done. Finished this form, and the reminders for it stop. Not finished, and they continue. Nagging someone who completed it yesterday is how you teach them to ignore you. Stopping because they missed a single message is how the registry goes quiet.

You also need an exit, written down before anyone is invited. The schedule has ended. They withdraw. They no longer meet the criteria. A clinician closes the pathway. Without that, reminders either run forever or stop because someone got tired of sending them.

Longer forms can sit on the quieter schedule. A short check can be dense at the start and then ease off. The knee registry follow-up model is one worked example: heavier in year one, light after that. The hard part of any of these is not picking the interval. It is getting the later form finished.

8. Pay them for the time

Respect the participant's time. If the registry is built strictly for patient care - helping a care team monitor progress - the benefit to the patient is often enough. But when a registry is for research or commercial evidence, questionnaires are a chore, and a registry asks for them more than once. In those cases, we think you should include a voucher, cash, or another incentive as part of the core design.

It can sit on a completed form, so the reward follows the work. The best way we've found to run this is to automate the distribution: the voucher or the payment trigger happens the moment the form is finished. This stops patients having to chase their reward, and it stops your team from having to manage a manual list of who to pay. The point is simple: if you aren't part of their direct clinical care, treating participants' time as free can make the later follow-up harder to complete.

Don't wait for an export to see how you're doing

We keep a live view running from the very first invite, so you can see exactly where people are in the flow: screened out, consented, baseline done, follow-up due, reward sent.

The dashboard is not a static count. We design it to be explorable, so you can filter the outcomes by the groups that matter to your evidence. For example, you might want to look at pain scores over 12 months specifically for a certain age group, or compare adverse events across different SKUs in your product portfolio. Having that view live means you are watching the evidence build, rather than waiting for a statistician to return a report months later. Export it to a spreadsheet when you need the raw file.

If consent includes the care team, those clinicians get their own view of the patients they are allowed to see. That second dashboard is common in registry work, because the clinic is often why the patient agreed to be followed at all. The study file and the clinician view are not the same screen, and they should not be.

Where most registries fail

The steps are simple to write down and easy to break once a person has to actually do them. These are the ones we keep seeing.

An app is the same mistake in a bigger box. Download it, make an account, then find the questionnaire inside a home screen built for something else. A registry patient does this once, or once a quarter, and many will not install software for it. A new tab is the smaller version: consent in one, the form in another, the voucher in a third. On a phone those tabs are how people get lost.

The design problem is friction. Keep them on the page they already have, show one thing, then the next. If they have to work to stay in the flow, you will not get month six.

There are other common traps too:

• Baseline collected, follow-up left as a later project. You get a contact list.
• Every form on the same intensive schedule. The long questionnaire should not ride the daily reminder.
• Consent that never decides who sees the name. A sponsor gets identifiers they should not have, or the care team cannot see their own patients.
• Screening done afterwards in a spreadsheet. People who should have stopped are already through.

How we usually run it

We have the infrastructure so those steps can happen quickly, in one user flow. No app to install, and the next step does not open a new tab. The patient sees one thing at a time, on the page they already opened, and is walked straight through. Reminders, rewards, the live view, and a separate dashboard for the care team sit behind that. The export is a spreadsheet.

The questionnaire is the easy part. The registry is whether the person can walk the whole workflow without being sent somewhere else.

If you want a tech platform just for this, get in touch.

Guy Solan

Founder, Patient Watch

I built Patient Watch at university for my dad, an orthopaedic surgeon. Clinicians pay for it because it sends questionnaires on a schedule, with reminders. A few manufacturers have paid us to collect follow-up for their own evidence. That is the experience this sequence comes from: clinic pathways first, then smaller product studies.

I have not built a national registry. I am not a CRO, and I am not the person who signs off your ethics or your questionnaire licence. The reply rate we see, about 85% against a figure around 12% often cited as a paper baseline, is directional. It is what happens when the reminder is done properly. It is not a promise.