How to prepare your medical records for treatment in China

1. The five documents that matter

Institutions assessing a case from abroad are trying to answer three questions: what is going on, what has already been tried, and what state the patient is in now. Five document types carry almost all of that.

Two things people commonly under-supply. The first is history: sending only the latest discharge summary hides the trajectory, and trajectory is often what determines whether a case is workable. The second is earlier laboratory results, a single value in isolation says far less than the same value three times over eighteen months.

2. Imaging: why DICOM and not a photograph

This is the most common practical failure, and the most avoidable.

DICOM is the standard format medical imaging equipment produces. A DICOM file is not simply a picture, it holds the complete series of slices, the acquisition parameters, and the measurement scale. Opened in the right software, a radiologist can scroll through the whole study, adjust the window levels to look at different tissue types, and measure a lesion accurately.

A photograph of a screen, a phone picture of a printed film, or a flattened JPEG in a PDF report gives none of that. It shows one slice, at one setting, at unknown scale. A radiologist can look at it, but cannot re-read the study, and re-reading the study is usually the point.

Bring the written report as well. The report tells the reader what the original radiologist concluded, which is information the images alone do not carry, and where a second reader disagrees with the first, that disagreement is itself clinically useful.

3. What not to send, and to whom

Genetic test results

Do not send genetic or genomic testing results to an agency, facilitator or intermediary, including us.

Genetic data is treated as a special category of personal information in most jurisdictions, and its handling across borders is regulated separately and more strictly than ordinary medical records. Where an institution needs those results, it can request them from you directly and receive them under its own framework. There is no situation in which an intermediary needs to hold them in order to do its job.

If anyone asks you to send genetic results to them so they can "pass them on", that is a reasonable point at which to ask why.

Everything at once, by email

Sending sixty pages of scanned records as email attachments to a first contact is a poor idea for two reasons. Email is not a controlled channel, and nobody has yet established whether your case is even a sensible fit. Start with a description in your own words and a list of the documents you hold. A secure transfer method comes after there is a reason to use one.

4. Four things that get lost in translation

This is the part almost nobody checks, and where the damage is quietest. None of these errors produce an obviously wrong document. They produce a plausible one.

Dates

03/04/2026 is 3 April in the United Kingdom, Australia and most of Southeast Asia; 4 March in the United States; and 3 April again, written with dots, in Russia and Central Asia. A translator who silently picks one convention can shift your entire timeline by a month.

A month matters. "Recurrence four months after surgery" and "recurrence three months after surgery" are not the same clinical picture. Dates should be rendered unambiguously, 2026-04-03, with a note recording which convention the original used.

Units

Laboratory units differ between countries, and the differences are large enough to change interpretation entirely.

A fasting glucose of 126 mg/dL is roughly 7.0 mmol/L. Read as though it were already in mmol/L, 126 is not a plausible number, but a reader skimming quickly may not stop on it. The safe practice is to keep the original value and unit, place the conversion beside it, and keep the laboratory's own reference range. Conversions should never replace the original figure.

Staging, and which edition

Cancer staging systems are revised. The AJCC staging manual moved from its 7th to its 8th edition, and for several tumour types the criteria changed enough that the same patient can be assigned a different stage under each. A report that says "Stage IIIA" without naming the edition is incomplete, and a translation that carries that omission forward silently is passing on an ambiguity as though it were a fact.

The same applies to Gleason scoring, where grade grouping changed, and to performance status scales, ECOG and Karnofsky are different instruments and should never be treated as interchangeable.

Drug names

Brand names are territorial. The same brand can correspond to different active substances in different countries, and the same active substance appears under many brands. A medication list translated by brand name alone is not reliable.

Each medication should appear as generic name, with the original brand name retained in brackets, followed by dose, frequency and dates. Dose deserves particular attention: a figure given per kilogram of body weight or per square metre of body surface area is not the same as a fixed total dose, and confusing the two can move the number by a factor of several.

5. How to actually request your records

In most countries you have a legal right to a copy of your own medical records. The specific law differs, and so does the timeframe, but the request itself rarely needs to be justified, you are asking for something that is already yours.

1. Write to medical records, not to your doctor. Most hospitals have a medical records or health information department. Going through your treating physician usually adds a step rather than removing one.
2. Ask for a complete copy, and name the formats. A written request works better than a phone call, because it creates a record: "I am requesting a complete copy of my medical records relating to [condition], including discharge summaries, clinical notes, pathology and laboratory reports. Please provide imaging studies in DICOM format together with the radiology reports."
3. Expect a fee and a waiting period. Both vary. A modest charge for media or copying is normal; a fee large enough to discourage you is not, and is worth questioning.
4. Check what you receive before you do anything else. Missing pages and missing reference ranges are common. It is far easier to go back once, immediately, than to discover a gap three weeks later.
5. Keep the originals. Give copies to anyone who needs them. Never hand over your only set.

6. What a usable translation looks like

Whoever does your translation, us, another provider, or a bilingual relative, these are the properties that make the result usable rather than merely readable. They are worth asking about before you commission the work.

• Nothing is omitted. Unfavourable findings, a poor prognosis, prior treatment failures, a physician's reservations, all translated in full. A translation that has been quietly softened is worse than no translation, because it looks complete.
• Nothing is upgraded. Poorly differentiated stays poorly differentiated. Hedged language stays hedged. The strength of the original wording is preserved exactly.
• Ambiguities are flagged, not resolved. Where a source document is unclear, contradictory or illegible, the translation says so in a note. A file carrying three honest footnotes is more valuable than a clean one containing a silent guess.
• Numbers keep their context. Original value, original unit, original reference range, with any conversion placed alongside.
• Someone with a medical background has reviewed it. A general translator renders words correctly and cannot be expected to know that a staging edition changes a conclusion. That review is the difference between a document and a usable medical file.

7. A realistic timeline

Two to four weeks end to end is normal. The largest single variable is your own hospital's records department, which is also the part nobody can accelerate for you: which is why it is worth starting there first, before anything else.