Results & evaluation
Results in from every lab. Statistics you can defend.
Labs enter results in their own portal, with every value checked as it is entered. EqasOnline then scores the round with ISO 13528 robust statistics, and your team reviews it in an analysis workspace before anything is released.
The problem
Most errors in a round happen before the statistics start.
- A unit slip or a misplaced decimal is only spotted when it has already pulled the consensus.
- Robust statistics live in macros nobody wants to touch, and each round is calculated slightly differently.
- When an assessor asks how an assigned value was reached, the answer takes days to reconstruct.
With EqasOnline
From submission to a released verdict
Online submission
Labs fill in a form built from your test, with their instrument, method and reagent pre-filled, and can attach the raw instrument output to a result. Results can be edited until the deadline.
Plausibility checks at entry
Values outside your plausible range or in the wrong unit are queried as they are entered, and the lab’s confirmation is stored with the value for you to see.
Analysis workspace
A labs-by-analytes matrix, Grubbs outlier flags, Youden plots, method comparisons and a detailed evaluation report with every Algorithm A iteration.
Release checklist
Incomplete submissions, values awaiting review and inadequate assigned values are flagged before you release, so nothing slips through unnoticed.
Trends across rounds
Westgard-style rules (1-3s, 2-2s, 4-1s, runs and drift) look back over up to 24 rounds and flag labs whose performance is shifting, in plain English.
Methods reference
What the engine calculates, and how
Every verdict is computed on the server. The portals display the result; they never recalculate it.
| What | Method | Reference |
|---|---|---|
| Assigned value, 12 or more results | Algorithm A robust mean and standard deviation | ISO 13528, Annex C |
| Assigned value, 3 to 11 results | Median and MADe (nIQR when the MAD is zero) | ISO 13528 |
| Uncertainty of the assigned value | u(x_pt) = 1.25 · s* / √p, judged adequate when u ≤ 0.3 σpt | ISO 13528 |
| Performance scores | z = (x − x_pt) / σpt and % bias for every result; z′ shown beside z in your review matrix and the parameter-wise report | ISO 13528 |
| Action limits | |z| ≤ 2 satisfactory · 2 < |z| < 3 questionable · |z| ≥ 3 unsatisfactory | ISO 13528 |
| Minimum group size | No score from fewer than 3 results; the analyte is marked not evaluable with a reason | |
| Outliers | Grubbs’ test, two-sided, α = 0.05, flagged for review | |
| Peer groups | All participants, and within method, instrument and reagent groups | |
| Qualitative results | Single most common answer from ≥ 60% of labs (≥ 80% is strong), or acceptable answers you define | |
| Semi-quantitative results | Ordered grades, such as dipstick results, are ranked and scored by their distance from the consensus grade | |
| Specialist schemes | Concordance scoring for HLA typing, crossmatch, HLA antibodies and immunophenotyping |
One-page overview
Get the one-page overview by email.
What EqasOnline covers for a proficiency testing provider, on one page you can pass to your team.
- Rounds, test items, results, ISO 13528 statistics, reports and certificates in one system
- Participants enter results in their own portal, and implausible values are queried at entry
- An ISO 13528 Annex B homogeneity and stability check before test items ship
- Reports under your branding, versioned once issued, with an append-only audit trail
FAQ
Common questions
What happens when an analyte has fewer than 12 results?
EqasOnline switches from Algorithm A to the median and MADe, which hold up better in small groups. Below three results no z-score is given, and the analyte is marked not evaluable with the reason.
Can labs change a result after submitting?
Yes, until the round’s deadline or a 24-hour extension you open. After that the result is locked for the lab, though your staff can still correct it.
How is the statistics engine checked?
It has its own test suite: results are compared with an independent implementation of the ISO 13528 formulas and with real rounds whose results are pinned so that any change shows up.
See a full proficiency testing round on EqasOnline.
In a 30-minute demo we run one full round in both the provider console and the lab portal, from item check to issued report, on your own scheme structure.