Editorial Policy and Methodology

This page describes how Evallume produces two different kinds of content: the interpretation reports generated for your uploaded lab results, and the articles published on this site. It also states plainly what the service does not do.

Published by EVALLUME LLC. Questions and corrections: support@evallume.com.

1. Where reference intervals come from

This is the single most important thing to understand about our method.

We use the reference interval printed on your own laboratory form as the primary source. Not a textbook range, not an average from the internet — the range your lab printed next to your result.

The reason is practical: reference intervals are not universal. They depend on the analyser, the reagent lot, the measurement method and the population the laboratory calibrated against. Two labs in the same city can legitimately report different ranges for the same marker, which is why a value flagged "high" by one lab can be normal at another.

Consequently:

If your form contains a reference interval, the interpretation is made against that interval.
If the form splits reference values by sex or age, the interval matching your profile is selected.
If the form contains no reference interval (common for imaging descriptions or free-text clinician notes), generally accepted clinical norms are used — and the report explicitly says so with the note "Reference taken from generally accepted norms — not present on the form". That marking is mandatory, so you always know which of the two happened.

Units are cross-checked as part of the same step (g/L against g/dL, µmol/L against mg/dL), because a unit mismatch is the most common way an automated reading goes wrong.

2. How an interpretation is produced

Reading the document. Uploaded PDFs and photographs go through optical character recognition. Where the form contains a results table, the table is extracted as structured data and used as the primary source; the surrounding raw text is used only for context such as the collection date and laboratory comments.
Extraction. Marker name, value, units and reference bounds are extracted per row. Values that cannot be read reliably are not guessed.
Interpretation. A large language model processes the extracted table under a fixed system instruction, taking your sex and age into account, and produces the report.
Delivery. You receive the report in your account and by email as a PDF.

Rules the model operates under

These constraints are part of the instruction the model receives on every request:

Never invent a value. If a figure is covered by a stamp, handwritten illegibly, or otherwise unreadable, the report states "Illegible" instead of a number. A wrong digit is more dangerous than a missing answer.
Never state a diagnosis. The model is instructed to write "may indicate impaired carbohydrate metabolism", not "you have diabetes". Naming a disease as an established fact is out of scope.
Never reproduce personal data. Names, dates of birth, addresses, insurance and ID numbers, phone numbers and email addresses present on the form are ignored and are not carried into the report. Physician and clinic names are not reproduced either.
Never ignore what you told us. Complaints and chronic conditions you enter in the comment field are taken into account.

3. Limitations you should know about

We would rather state these than let you discover them.

OCR can misread. Faint scans, photographs at an angle, stamps over figures and handwritten corrections all reduce accuracy. Always compare the summary table in the report against your original form.
Language models can err. The interpretation is generated automatically. It is a reading aid, not an authoritative clinical judgement.
A single result is weak evidence. Lab values move with time of day, fasting state, recent exercise, medication and acute illness. One out-of-range figure is a reason to look further, not a conclusion.
Context we do not have. We do not see your history, your examination, your imaging or your prescriptions. Your physician does, and that is why the report is a starting point for a conversation rather than a substitute for one.
No medical review of individual reports. Reports are not reviewed by a clinician before you receive them. Evallume is an informational service, not a medical service, and does not present itself as one.

4. How we verify articles

Every article is checked by the Evallume editorial team against the primary sources listed at the end of it. Where a source does not support a statement, the statement is corrected or removed.
Sources are listed at the end of each verified article. We cite primary and institutional sources — MedlinePlus and the U.S. National Library of Medicine, the National Institutes of Health (NIDDK, NHLBI, NIAAA), the World Health Organization and comparable national health authorities — rather than secondary aggregators.
We are working through the catalogue in order of readership. If an article does not yet show a source list, it has not yet completed that check — treat it as general orientation and rely on the cited articles and on your physician for anything that matters.
Numeric ranges quoted in articles are illustrative and describe typical intervals. They do not override the reference values on your own form.
Articles carry a publication date and, where they have been revised, a revision date.

Corrections

If you find a factual error, write to support@evallume.com with the article URL and the disputed statement. We check the claim against primary sources, correct the text where the objection holds, and update the revision date on the page. We do not silently delete criticised material.

Independence

We publish no sponsored articles, no paid placements and no advertising for medications, laboratories or clinics. Nobody pays to be mentioned, and no external party has editorial control over what appears here. Our only revenue is what users pay for interpretation reports.

5. What Evallume is not

Not a medical service. No diagnosis, no treatment, no prescriptions, no medical opinions.
Not a substitute for a physician. The report is material to bring to an appointment, not a replacement for one.
Not an emergency service. Chest pain, difficulty breathing, severe bleeding, loss of consciousness or acute deterioration require emergency care immediately — not a lab-report interpretation.
Not a laboratory. We do not perform tests and are not responsible for the analytical accuracy of the figures on your form.
Not a screening or triage tool. The absence of flagged deviations does not mean the absence of disease.

Important. Evallume provides informational services only. Interpretation results do not constitute medical advice, a diagnosis, or a treatment recommendation. For diagnosis and treatment, consult a qualified physician.

Policy last updated: 26 July 2026. Operated by EVALLUME LLC, 30 N Gould St Ste N, Sheridan, WY 82801, United States.