How laboratory intelligence works

Available to

Laboratory users, reviewers and directors.

Laboratory intelligence watches results as they arrive. It checks that each value is fit to interpret, applies approved rules, and, where your laboratory allows it, offers a plain-language suggestion for a named reviewer to accept or reject. The laboratory works exactly as before when this layer is idle, switched off or unavailable.

The rules it follows

Rules come first. Critical value alerts, plausibility, change-since-last-time and specimen checks are fixed rules from an approved, versioned rule set. They run the moment a value is entered or received, with the model switched off.
The model is optional and never decides. Each facility is set to Off: rules only, Shadow: model runs, nothing is shown to reviewers, or Supervised: suggestions go to named reviewers. Model text is unverified until a reviewer accepts it, and it cannot release, change or hide a result or a rule alert.
It holds back what it cannot interpret. A result with an unclear unit, no reliable reference interval, a failed or missing quality-control run, or a specimen problem is not interpreted, and says why.
Everything leaves a trail. Each case freezes the facts it saw, each decision names its reviewer, and a chained audit log can be checked from the screen.
Nothing identifying leaves the system. The model sees structured, de-identified facts only.

The Laboratory dashboard has a review row: cases waiting for a reviewer, rule alerts, overrides, held-back cases and open incidents.

Where things are

Menu under Laboratory	What it is for
Intelligence	The daily review work: the review worklist, suggestions to review, rule alerts, comments on reports and reviewer decisions.
Quality and safety	Incidents, model performance, the audit chain and feedback datasets.
Instruments and interfaces	Instruments, quality-control runs, instrument messages and sample identifier mapping.
Rules and AI configuration	Policies, rule sets, reflex protocols, unit conversions, AI models, prompts, references and routing. Under the + menu.

Roles

Role	Can
Laboratory Reviewer	Decide on suggestions, add or withdraw report comments, submit configuration.
Laboratory Director (rules and AI)	Approve, activate, retire and roll back rule sets, models, prompts, references and unit conversions; approve feedback datasets. A different person must approve what someone submitted.
Laboratory Quality Officer	Read the audit, incidents and performance screens.
Laboratory Interface Operator	Submit instrument messages and manage instruments, mapping and quality control.

Was this helpful?

Report an issue with this guide