What frightened people actually read at 2am

Every interface decision on a patient-facing clinical service should be tested against one scenario.

It is two in the morning. The person is on a phone, in bed, not sleeping. They received a diagnosis two days ago — theirs, or their father's. They understand perhaps sixty percent of the words in the report they are holding. They have read forty pages this week and are not sure any of them applied to their situation.

That is your reader. Almost nothing in standard web practice was designed for them.

These are the things we changed after building and running a cross-border second-opinion service, and after watching what actually happened when frightened people used it.

They are not reading. They are scanning for one thing.

The assumption behind most clinical page design is that a reader progresses down the page absorbing an argument.

They do not. A frightened reader arrives with one question — does this apply to me — and scans until they find it or give up. They are not evaluating your practice. They are trying to work out whether this page is about their situation at all.

What follows from this: the specificity has to be at the top. Not "advanced cancer treatment options" but the actual diagnosis, the actual stage, the actual drug or isotope name as it is spelled in their report.

Patients read their own reports. They search the terminology in them — PSMA, DOTATATE, Gleason, ejection fraction, HER2. A page that uses the clinical vocabulary from the document in their hand tells them in two seconds that they are in the right place. A page that says "personalised care pathways" tells them nothing.

Practical rule: the first sentence under every heading should be the answer, and the heading should be a question the patient would actually ask.

Marketing register reads as a warning sign

This is the finding that most surprised us, and it is consistent.

Warm, reassuring, professionally written marketing copy — the register that performs well in almost every other category — produces suspicion in this one. Not indifference. Active suspicion.

The mechanism seems to be that a frightened person is running one background question about every source: is this a serious clinical operation, or is this a business? Polished promotional language answers that question the wrong way, quickly, and the judgement is difficult to reverse.

What works instead is plainness that verges on flat. Describing what happens, in order, in the words a clinician would use. "You send the scan and the report. A nuclear medicine consultant reviews it. You receive a signed report within 72 hours." No adjectives, no reassurance, no promise.

Reassurance offered by a stranger is not reassuring. Specificity is.

Urgency mechanics are read as predatory

Countdown timers, limited slots, "book before the end of the month."

Three reasons this fails here, and one of them is legal in Australia.

The urgency is already present and it is not yours to manufacture. A person with a serious diagnosis is under genuine time pressure. Adding a synthetic layer registers as someone attempting to profit from real fear.

It answers the serious-operation-or-business question wrongly and instantly.

And under AHPRA's advertising rules, advertising that exploits health anxiety is a breach — not a grey area.

The hardest engineering problem is the upload, not the reporting

This is the thing nobody plans for and it consumed most of our build effort.

A 400MB DICOM study, arriving over an unstable mobile connection, from a person who has never used a medical portal, at two in the morning, in a different time zone, with no technical support available.

Everything about that sentence is hostile to a normal file upload.

What it required:

• Chunked, resumable uploads. A connection dropping at 380MB cannot mean starting again. It will drop, repeatedly.
• Progress that is honest. A bar that sits at 94% for four minutes is worse than one that says "uploading — this may take 20 minutes on a mobile connection."
• Tolerance for the wrong file. People send photographs of a screen showing a scan. They send the ZIP unextracted. They send one slice instead of the study. The system has to accept, inspect and explain rather than reject.
• A human fallback that actually works. A WhatsApp number where a person responds. In India and much of the world this is where these conversations happen, and treating it as an informal channel rather than core infrastructure is a design error.

None of this appears in a screenshot. It is where the effort goes and it is the difference between a service that works and one that has a beautiful interface nobody completes.

Every field on the form costs you a patient

The intake form is where the most damage is done, because each field seems individually reasonable.

Ask only what the reporting clinician needs to do the work. Not what a CRM would like. Not what a marketing team wants for segmentation. If a field does not change the clinical output, remove it.

Two fields in particular:

Free-text "tell us about your condition" is where people write the most important information and where they get most stuck, because they do not know what is relevant. Replace it with two or three specific questions. "What has the diagnosis been so far?" "What treatment has been offered or started?" These get answers; the open field gets paralysis.

Anything asking for financial or insurance detail before a clinical response signals the wrong priority order at exactly the wrong moment. It can be asked afterwards.

Time promises are clinical, not commercial

Every additional day of uncertainty has a real cost for a family, and it is not a customer-service cost.

Two things follow.

Do not promise a turnaround you cannot hold on a bad week. We committed to 72 hours to the patient and ran a 60-hour internal deadline behind it, so that the commitment did not depend on everything going right. If you cannot hold the number on your worst week, publish your worst week.

Say what happens if it slips. Almost no service does this, and it costs nothing. A sentence explaining that you will contact them before the deadline if there is a delay converts an anxious silence into a manageable one.

Independence, said plainly, does more work than any credential

The most powerful sentence on our service pages was structural rather than persuasive: the service earns from interpretation and never from treatment. No referral fee, no facilitation commission, no treatment pathway downstream that an opinion could be nudged toward.

Patients raise this more often than turnaround, price or the consultants' publication records when explaining why they chose the service.

The reason is that a person seeking a second opinion is, by definition, in a state of suspicion about advice. They have been told something they are unsure of. Every subsequent source is being assessed for motive. A structure they can read, that visibly costs the provider money, resolves that question in a way no amount of credential display can.

If your model has this property, say it plainly and early. If it does not, the patient will work that out, and the reassuring copy will not survive the discovery.

The test

Before anything patient-facing goes live, read it against the scenario at the top of this piece. Phone, 2am, two days after a diagnosis, sixty percent comprehension.

Then ask three questions:

1. Can they tell within five seconds whether this applies to them?
2. Does the language sound like a clinician or like a brochure?
3. What is the smallest true next step, and is it described rather than demanded?

Most clinical pages fail all three, and they fail them because nobody who wrote them imagined the person reading them.

The practices that get this right are almost always the ones where a clinician read the page before it went live. It shows, and the readers can tell.