How it works

Lab results, read the way a physician thinks.

Plainvue takes the numbers on a lab report, reads them together instead of one at a time, and tells both the doctor and the patient in plain English what they mean, what should happen next, and how urgently. Under the hood is a patent-pending AI engine trained on every current medical text book, journal article, and clinical guideline.
The Problem

Lab reports are printed one line at a time. Our bodies do not work that way.

A standard lab report lists each test with a number, a range, and a flag. That’s where it stops. Nobody tells you that a calcium value only makes sense next to a parathyroid hormone level, that a low potassium with high blood pressure points somewhere specific, or that a mildly high liver enzyme means one thing when the other liver tests are normal and something else entirely when they aren’t.

When we mapped the routine adult blood panel, 83 percent of the analytes couldn’t be interpreted correctly without knowing the values of other, interdependent tests. A physician carries that map in their head for their own specialty. No one knows them all, and a patient-facing web portal has no clue. Plainvue was built to close that gap.

Layer One

The Interdependency Engine. Patent Pending.

At the foundation of Plainvue is a directed map of how every common lab test relates to every other. For each analyte, in each direction it can go, the engine knows which other results influence it, which follow-up tests settle the question, and which of those are already answered elsewhere on the same draw. The map was written by physicians, not scraped from the internet: every rule is tied to published guidelines and can be inspected and edited by a physician.

Two things set it apart from a lookup table. It reads the whole report at once, so a pattern across several tests is recognized as one problem, not five. And when a value comes back abnormal, it assembles the smallest set of follow-up tests that resolves it, removes duplicates across the report, and hands it forward as a clean, orderable requisition, not a paragraph of suggestions.

Layer Two

Clarity AI. Software that thinks like a physician.

Clarity AI lives on top of the Interdependency Engine. Where the engine knows what relates to what, Clarity decides what a good doctor would actually do about it, and explains the reasoning in language anyone can follow.

A physician doesn’t treat every flagged value the same way. A cholesterol a few points high and a potassium a few points low are different problems with different urgency. Clarity encodes that judgment as physician action thresholds: for each test, in each direction, the level at which guidelines say repeat it, the level at which they say act now, and the level at which they say treat rather than test. Every threshold is cited to its source.

From those thresholds, every abnormal value is sorted into one of four tiers.

Tier 0

Normal or barely off: No new testing.

Within the normal range, or just outside it and below the level where a physician would order anything. Clarity explains what usually causes a small shift, and that nothing more is needed right now.

Tier 1

Normal or barely off: No new testing.

The guidelines call for a repeat in three months or more, or a treatment conversation rather than more blood work. Nothing is recommended today; Clarity spells out why waiting is the right call and outlines the conversation to have.

Tier 2

Modestly abnormal: Limited follow-up testing now.

The value is past the threshold where a physician, per clinical guidelines, would recheck that test along with one or two first-line tests that settle the most likely explanations.

Tier 3

Abnormal: Comprehensive workup now.

The value is high or low enough that most guidelines recommend a full evaluation. Clarity lays out the ranked list of possible causes and the tests that separate them.

The result behaves the way an experienced clinician does: calm about the small stuff, precise about the serious stuff, and never silent about which is which.

Physician-reviewed · sourced from clinical guidelines

Advanced View

All 58 analytes, Aug 28, 2026, with reference intervals.

Where each result falls on its own axis. Open Advanced View on any test for its full population histogram against men your age.

Fasting glucose Normal 70–99 mg/dL

Advanced View

108 Aug 28, 2026

Phosphorus Normal 2.8–4.1 mg/dL · Adult

Advanced View

4.6 Aug 28, 2026

Total cholesterol Desirable < 200 mg/dL

Advanced View

218 Aug 28, 2026

LDL cholesterol Optimal < 100 mg/dL

Advanced View

123 Aug 28, 2026

Mean cell volume Normal 79–97 fL

Advanced View

102 Aug 28, 2026

TWO AUDIENCES. ONE TRUTH.

Plain language for patients.

Clinical depth for physicians.

Every test in Plainvue carries two descriptions written by physicians, not generated on the fly by a chat box in the cloud. The patient version reads at about a tenth-grade level: what the test measures, what a value at this level usually means, what happens next, and why. The physician version carries the mechanism, the interpretation pearls, a ranked differential, and the next labs to order, each tied to the diagnosis it rules in or out.

The two never disagree, because they come from the same clinical logic. A patient and their doctor looking at the same report see the same conclusion at two different depths.

DIFFERENTIAL DIAGNOSIS AND CODING

From an abnormal value to a ranked differential to an ICD-10 code.

For each abnormal value, and each combination of them, Clarity produces a ranked differential: the possible explanations in order of likelihood, with the follow-up tests that tell them apart. That’s exactly the exercise a physician runs in their head before ordering anything.

Because follow-up testing has to be ordered against a diagnosis, Clarity carries a knowledge base of ICD-10 codes and how they relate: which diagnoses travel together, which follow from which, and which code justifies which test. When a requisition is generated, the right ICD-10 codes are attached automatically. The order is complete, coded, and ready to send, with no manual lookup.

Ordering

Follow-up testing with one click, one draw, one bill.

Commercial lab panels are built for screening, a CBC, a metabolic panel, a lipid panel: great at finding an abnormal value, useless at working one up. Plainvue adds a second class of panel, purpose-built for the secondary workup, so a patient with several abnormalities orders one panel instead of six tests, on a single draw, with the ICD-10 coding already attached. And the follow-up is deliberately rationed: Clarity only recommends testing when a physician would, and never turns a slightly-off value into a cascade.

Validation

Developed and validated on real populations.

The engine and the action thresholds were developed and validated against the national database of more than 3.2 million laboratory results from roughly 70,000 adults. That is how we know the tiers behave sensibly across a real population, how often each tier fires, and what a typical follow-up looks like for a typical adult, before a single patient sees it.

Step by Step

What happens when you upload a lab report.

  • Step 1

    Your results are read as a whole, every value placed on the interdependency map alongside the others on the report.

  • Step 2

    Each abnormal value is assigned a tier using physician action thresholds, separating the serious findings from the trivial ones.

  • Step 3

    You get a plain-language explanation of every value, what it means at this level, and what happens next. Your physician sees the same findings in clinical depth, with a ranked differential.

  • Step 4

    Clarity assembles the smallest set of follow-up tests that resolves the abnormalities, de-duplicated across the report, and matches it to a purpose-built workup panel.

  • Step 5

    The order is generated with ICD-10 coding attached, ready to send in one click, on a single draw.

What Plainvue is not

Plainvue is not a chatbot guessing at your labs. Every description was written by a physician, every threshold is cited, every rule can be inspected, and the clinical logic is frozen and versioned, so the same report gives the same answer today and next year. Four patents pending.