Restricted access

This page is intended for invited clinician reviewers. Please enter the access code your study coordinator sent you.

No personal information is sent to a server. Your responses save in your browser locally and are exported by you as a JSON file at the end.

Clinician review of missingness policies

0 / 29 reviewed Saving…
Show:

You're reviewing a first draft of the rules our trial-matching system uses to decide what to do when a patient's note doesn't explicitly mention something a trial criterion asks about — e.g. the trial requires "history of MI" but the note never mentions cardiac history. The wrong default either rejects patients who should have matched, or sends too many for needless in-person checks.

Your judgment goes straight into the next version of the matcher — the end product is a system you'd trust enough to use on your own patients. Disagree freely; that's the point of this pass.

What the three options mean (click to expand)
Treat as the criterion is satisfied   assume the criterion IS met. Use when an unmentioned condition probably means the patient has it (e.g. "patient can walk", "patient has no history of major bleeding" — the note would mention impairment if present).
Treat as the criterion is not satisfied   assume the criterion is NOT met. Use when the condition would be in the chart if the patient had it ("cystic fibrosis", "MI in last 30 days", "on dialysis").
Needs in-person check   don't decide from the chart, send to the visit. Use when an unmentioned condition is uninformative — common or unreliably-documented conditions. When in doubt, pick this; sending an eligible patient to a needless visit is far less costly than rejecting them up front.

Each rule has two parts because criteria can be phrased either way ("patient must have X" vs "patient must NOT have X"). The two parts are set independently — pick the right rule for each side. They don't have to be inverses. For a safety-first category (a serious acute condition, say), silence on both sides may reasonably reject; for an informative-absence category, silence on the "must-have" side may reject while silence on the "must-NOT-have" side may pass through.

⊕ Propose a brand-new category

Use this only if you see a kind of clinical criterion that doesn't fit ANY of the existing subcategories below — even with a stretch.

⌭ Extra rule: criteria with attached detail (severity, certainty, documentation, etc.)

Some trial criteria add a detail to a diagnosis — e.g. "severe pneumonia," "Kawasaki disease diagnosed per CDC criteria," "confirmed COVID-19." We sort these into two buckets and apply a meta-rule on top of the per-diagnosis rules above. Review each bucket the same way you reviewed the cards above.