Understanding
The patient-facing path: clear language, practical explanations, and information that helps someone understand their healthcare.
About the project
Trans healthcare should not require becoming your own researcher.
HeldToStandard exists because trans healthcare should not require people to become researchers just to understand their own options.
What the project does
Every treatment profile is traceable and open to inspection.
HeldToStandard cross-references what the medical evidence says about trans healthcare against what official regulators require. Every treatment profile is traceable: it is tied to an archived, dated primary source that anyone can check.
The goal is to give trans people a clear, plain record of the evidence behind their care options, traceable back to where each piece of it comes from. A person should be able to look up a treatment option and see what the research supports, what official standards of care require or recommend, how certain the evidence is, and where those things align or differ.
The patient should not have to become a researcher. But if they want to check the work, they should be able to.
Verification
The evidence path: the sources, the methodology, the graded evidence, the limitations, and the reasoning behind conclusions.
What HeldToStandard does not do
Not another authority to accept blindly.
Does not replace care
The project does not tell anyone what to do with their own care. It does not replace a clinician and does not claim to see the whole of anyone's situation.
Does not advise
A treatment profile describes the state of the evidence and the regulatory position. It does not issue a recommendation in its own voice.
How trust works
Every source is archived,
Every source is archived,
dated, and open to inspection.
Linking to a URL and trusting it will stay there is not enough. HeldToStandard fetches, snapshots, and indexes everything it references.
Source retrieval
The crawler respects robots.txt on every redirect hop,
uses per-domain rate limiting, and identifies itself honestly.
Archiving
Each fetched source is stored under its final URL with a content-addressed identifier. Nothing silently overwrites anything.
Validation
A profile missing a citation key, source list, or recognised regulator is a hard error, not a warning.
Each treatment profile is a structured record pairing a standard of care
with graded evidence. The standard-of-care side records what an official body requires or
recommends. The evidence side records what the medical literature says, with certainty graded
and sources attached. Treatment profiles move through status tiers:
draft / reviewed / published.
A reviewer can always trace any treatment profile back to what was published, by whom, on what date, as it actually appeared.
The guiding principle is simple: silent failure is impossible. When the system is uncertain, its failure mode is exclusion, not inclusion. A source that cannot be verified is kept out, not admitted with a caveat. A valid source kept out can be re-fetched and admitted later. An invalid source let in corrupts the record it enters.
The academic ingestion pipeline uses Europe PMC primary and PubMed E-utilities fallback, with provenance fields on every entry so a reviewer can re-fetch and confirm. AI assists with tooling and text, but it does not author treatment profiles. The validation layer refuses any profile that cannot show an archived source.
The person behind the project
“I spent nearly ten years trying to understand my own options, while repeatedly finding myself having to become my own researcher just to know what care was possible, what the evidence said, and what the system required.”
Nikita Ilyana Kerdman, founder
My name is Nikita Ilyana Kerdman. I go by Ilyana.
I built the first version of HeldToStandard, including the crawler, evidence ingestion systems, validation tooling, archival processes, and the underlying codebase.
I am a trans woman, and my experience navigating the Swedish healthcare system is one of the reasons this project exists. I spent nearly ten years trying to understand my own options, while repeatedly finding myself having to become my own researcher just to know what care was possible, what the evidence said, and what the system required.
That experience is not the whole project, but it is where the urgency comes from.
Transammans Mitt
I served on the board of Transammans Mitt in 2022. I came with ideas and limited capacity, newly employed, and still dealing with too many things that took most of my energy.
I did not deliver what I said I would, and I think about that a lot.
This is what I am doing instead: something with no hedged commitment, no partial effort, and no version of showing up that lets me off the hook. Something I will finish.
AI and tools
AI is used as part of building HeldToStandard, and that is stated openly because hiding it would contradict the transparency this project depends on.
The tools that collect and archive sources are crawlers and academic APIs, automated retrieval infrastructure of the kind that has underpinned search engines for decades. They fetch and index what already exists. They do not generate content.
A generative AI coding assistant is also used to build and maintain the tooling and to draft documentation. What it cannot do is author a treatment profile or generate a medical assertion that passes validation. Every treatment profile must cite an archived primary source. The validation layer refuses anything that cannot show its receipts.
AI does the labour here. It does not get to decide what is true.
Scope and status
Sweden first. Not yet public.
HeldToStandard is not yet in its first public release.
Sweden first, with a Scandinavia rollout planned. The crawler, validator, academic ingestion, and pre-commit hook are built and tested. An initial source retrieval run completed but is paused pending a legal review of the redistribution model; source data will be repopulated once that review is resolved. EMA bulk download and source integrity vetting are still in progress.