SCUELJapan healthcare data

Solutions · SCUEL RD

Japanese hospitals publish more about themselves than you would expect

Which conditions they treat. Which tests they can run, down to the method. Which specialty clinics they hold, and on which afternoons. What equipment they own, by model. It is public, it is detailed, and almost none of it appears in any claims dataset. SCUEL RD is the service that reads it.

What a Japanese hospital page actually contains

An example from a cardiovascular department, paraphrased.

Conditions treatedHeart failure, ischaemic heart disease, valvular disease, aortic disease, peripheral arterial disease, arrhythmia, cardiomyopathy, congenital heart disease. Diagnosis using imaging and blood biomarkers to international standards, with medical and surgical teams working jointly on drug, catheter and surgical treatment.

Tests available12-lead ECG. Ankle-brachial index, pulse wave velocity, exercise ABI. Skin perfusion pressure, transcutaneous oxygen tension. Echocardiography, transoesophageal echocardiography, vascular ultrasound. CT, MRI, myocardial scintigraphy, cardiopulmonary exercise testing.

Specialty clinicsArrhythmia clinic, weekday mornings. Foot care clinic, Tuesday afternoons. Pacemaker clinic, Thursday mornings.

If your product depends on a specific test being available on site, that page has just answered the question. No claims dataset will, because running a test and billing for it under a code you can identify are different things.

How the survey runs

Three stages. Each narrows the previous one.

Find the pages Search terms you specify are used to locate the pages within Japanese medical institution websites most likely to carry the information. Those pages become the material for everything that follows.
Match the strings Test names, treatment approaches, compound names — specified as literal strings and flagged where they occur.
Read the context A string appearing on a page does not mean the institution does it. A language model reads the surrounding text against criteria agreed with you, and returns a judgement.

The judgement has three values, not two

A page that says the institution performs a procedure is one thing. A page that explains what the procedure is, in general terms, is another — and a page that states the institution does not perform it is a third.

So the result is likely treating, not treating, or cannot tell. We keep the third category rather than forcing it into one of the other two. A list that claims certainty it does not have is worse than one that admits where it is unsure.

What you receive

Two files, so the reasoning stays visible.

One row per institution Facility name and address, type, how many pages were found, which keywords and strings matched, and the judgement for each criterion.
One row per page found The same, broken out by URL. This is the part clients tell us matters: when someone internally asks why a facility is on the list, the answer is a link.
Your own facility codes attached If you hold Japanese facility records already, we can join the results to them.

Where it is used

Consistently the same shape of problem.

Rare disease, where a condition is hard to attribute to a specialty and the diagnosing physician may sit anywhere. Products with no predecessor, where there is no prescribing history to work from. Tests and screening, where the question is whether a facility can run something rather than whether it has billed for it. Vaccination, where administering a vaccine is not an insured procedure and therefore leaves no claims trace.

What these have in common is that the fact you need is one the institution has published and the billing system has not recorded. The rare disease page shows how this combines with physician-level data.

What it will not do

Worth reading before you scope a survey.

Institutions without a website are out of scope There is no way around this. The method reads what is published; where nothing is published, there is nothing to read.
Some pages cannot be reached or parsed Information held deep in a site, presented as an image, or built in ways that resist extraction may be missed even where the website exists.
Some keywords are the wrong tool Terms that map cleanly onto a displayed specialty or a board certification — rheumatology, for instance — will mostly return institutions you could have identified from the physician database instead. We will say so rather than run the survey.
The analysis is not guaranteed complete Language model judgement is a strong signal, not a certainty. Clients typically validate a sample against their own field knowledge before committing.
Websites change A page carrying a keyword at the time of the survey may not carry it by the time you look.

Tell us what you need to find and we will tell you whether this is the right instrument for it — including when it is not. Working with us from outside Japan? Read how engagements run first.

Send an enquiry

Engagements typically start from several million Japanese yen — indicatively, tens of thousands of US dollars. The US dollar figure is for scale only and is not a quoted price. Scope, coverage and contract term determine the final figure.