Methodology

What we publish

We publish structured facts about where treatment is available in China and what it is reported to cost. Each fact is one row. Every row carries the source it came from, the date that source was retrieved, and an evidence level.

We do not publish treatment recommendations, and we do not rank hospitals by outcome. There is no hospital page that exists because a hospital paid for it.

A hospital is listed in this directory when we hold a confirmed external record about it - a place on the Fudan general or specialty list, a national key clinical specialty at evidence level A or B, or a current JCI or CIHA accreditation - or when it appears on the September 2025 Shanghai public-hospital pilot list for international medical tourism. That covers 137 of the 731 hospital records on this site. The other 594 are kept on file but are not listed, not searchable and not indexed.

The evidence model is applied today to:

• National key clinical specialties, stated by a health commission or a university.
• Hospital grade and the date the name list was published.
• Reported cost ranges for a treatment path, with the year and the payment basis.
• Published five-year survival figures, with their population and period.

The same model is planned, but not yet published, for:

• Insurance direct billing, which today exists only in insurer provider directories and hospital self-descriptions.
• Organ transplant programmes, where volume figures are not published institution by institution.
• Proton and heavy-ion facilities, where the equipment list comes from hospital statements rather than from a register.

Sources and evidence levels

A level describes the source, not the hospital. A lower-tier source cannot be promoted by finding a second copy of the same claim, and a level is never upgraded for presentation reasons.

Level C is shown differently, on purpose

Level C is a statement about our search, not about the hospital. Rows at level C are marked "No independent official source found for this row", they are never sorted above level A or B rows, and they are never counted as if they were verified.

How rows are ordered

Ordering on this site reflects the signals listed on the page, not clinical outcomes. A national key specialty list, a hospital grade and a cost range are administrative and financial signals. None of them predicts how your treatment will go, and we do not present them as if they did.

Where a page lists hospitals, it sorts them by province, then city, then hospital name, so the directory reads like an atlas rather than a leaderboard. Specialty records on a hospital page and the hospital list on a condition page keep the order of the official source or the same province-city-name order. An evidence level is a marker on a row, never a position. No hospital can buy a higher position.

How a condition shortlist is chosen

Selected using official public data sources

These hospitals are selected from official public data sources — the Fudan University hospital specialty reputation rankings, the national key clinical specialties list and national clinical research centres — so each line on a card can be checked against its published source. This is not a paid or promoted listing.

A condition page lists up to eight hospitals for that condition. The record table further down the same page is a different thing: it is every hospital holding a national key specialty record in the specialty the condition is filed under, in the province-city-name order described above. The shortlist above it follows the rule below, and each line on a card links to the published source it is taken from.

Admission. A hospital is listed for a condition when the specialty that condition is filed under carries at least one of: a place or a nomination in the Fudan University hospital specialty reputation ranking; a national key clinical specialty record at level A, B or C; or a national clinical research centre or a state key laboratory in that specialty. JCI accreditation is shown as a badge; it never decides admission and it never decides order.

Order. Step 1: a Fudan place between 1 and 10 for the specialty. Step 2: another Fudan place, or a nomination. Step 3: a level-A national key clinical specialty. Step 4: a level-B national key clinical specialty. Step 5: a national clinical research centre or a state key laboratory in the specialty, a centre before a laboratory. Step 6: a level-C national key clinical specialty, which is a record we publish without an independent official source and which is therefore listed after the others.

Ties. Within one step: the better Fudan place first, then the stronger key-specialty level (A before B before C), then the province, then the city, then the hospital name - the same province-city-name order the hospital directory uses. At most eight hospitals are listed.

When the list is short. Five hospitals or more: the first eight are listed. Three or four: all of them are listed, and a further fold adds the remaining records up to eight, each carrying its evidence level. One or two: the same, with that fold open and a line stating how many hospitals have a strong public record. None: the page states that no strong public record was found and links to the specialty directory. A place, an institution or a count is never invented to fill the list.

The Fudan rankings are the hospital specialty reputation ranking most widely recognised by Chinese clinicians, published annually by Fudan University's hospital management institute.

Clinical studies

A condition page also lists clinical studies that public registers record as recruiting or not yet recruiting in China. Only publicly registered study information is shared; it is neither a recommendation nor an encouragement to join, participation is entirely voluntary, and the risks of an unapproved treatment are unknown.

Aggregate figures and reported side effects

Some numbers on a condition page describe a group rather than a person: a national satisfaction score, or the rate of one side effect in a single clinical trial. We publish those as aggregate figures only. We do not publish patient testimonials, and we do not present a group result as if it were one patient's outcome.

Every aggregate row states the source it came from and the year or date it refers to, and it carries its own source line with a link and an evidence level. A figure with no source is not published: where we find no sourced number, nothing is rendered rather than an estimate.

These figures are information, not medical advice. A national average or a single-trial rate cannot tell you what will happen in your own case, and figures from different settings are not a comparison of care.

How often data is refreshed

Datasets are refreshed in batches rather than continuously, because the sources themselves publish on a cycle: hospital grade lists and specialty lists appear once a year or less often, while cost ranges follow the year they describe.

Each row keeps the date its source was retrieved, so a stale row is visible as a stale row. A fixed public refresh schedule has not been committed to yet, and this page will state it once the content operation starts.

Corrections

If a row is wrong, we correct the row and record what changed, why, how widely it applied and which source it rests on. The record is public and permanent.

Read the corrections log

How to read the survival figures

The five-year survival figures we publish are age-standardised relative survival. Relative survival compares people diagnosed with a cancer against people of the same age and sex in the general population, so a figure is not the share of patients who are still alive five years after diagnosis. Age standardisation keeps the comparison from being driven by the age mix of the patients.

The 2019–2021 values are published as an estimated value (the source calls the method a hybrid analysis). Cancer registries report with a lag, so the National Cancer Center combined the follow-up already available with a modelled estimate for the most recent years. The figures are the latest national estimates, not a final count of every case.

For context, the same national source reports 43.7% five-year relative survival for all cancers combined in 2019–2021; the longer-term comparison in the same research line gives 30.9% for 2003–2005. Those two figures come from different sets of population-based cancer registries (17 and 281), so they are two published national estimates, not one group of patients followed over time. The all-cancer figure covers every cancer type together and is not a benchmark for any single cancer.

Sources: the National Cancer Center's 2019–2021 national publication (JNCC, 2024); the 2003–2005 comparison is published in a 2025 review in the Chinese Journal of Cancer Research.

What we do not know

• There is no official, institution-level register of surgical volumes in China. Volume claims come from hospitals describing themselves, which is level B by definition.
• Cost ranges are reported ranges, not quotes. They exclude travel, accommodation and complications, and the hospital issues the actual bill.
• Survival figures are population-level. A population figure cannot tell you what will happen to one patient, and we do not present it as a prognosis.
• Survival figures are published for all stages combined. For most cancers, the stage at diagnosis is the strongest single factor behind survival differences, and comparisons between cancers are also shaped by how early or late each one tends to be diagnosed.
• Language support, insurance direct billing and appointment channels are not covered by any official statistic, so those fields are not published yet.

Who is accountable