GuidelineVahti — panel consensus for guideline groups (preview)

What you get

A frozen sample of a 12-panelist guideline round — this is exactly what the tool renders.

Sample output

Panel ordinal agreement (Krippendorff’s α, interval): 0.74 — tentative; read it with the per-statement medians/IQR. Chance-corrected across all statements on the 1–9 scale.

Use a slice thickness ≤ 1.5 mm for nodules < 10 mm. Consensus · FORP1

Importance

IQR 0.75

Feasibility

IQR 1

% for (7–9)

100%

Votes

1–3 · 0

4–6 · 0

7–9 · 12

GRADERecommend (strong, high certainty)

Discharge a stable solid nodule after 12 months of follow-up. No consensus↻ continue — re-vote

Importance

IQR 7

Feasibility

IQR 6

% for (7–9)

50%

Votes

1–3 · 6

4–6 · 0

7–9 · 6

GRADENo recommendation — consensus not reached — the panel is split (extreme-disagreement rule)

Importance × feasibility (median ± IQR · cut at 7)

Implementation priority

Implement first — high importance · feasible

S1 — imp 9 · feas 8 — slice thickness ≤ 1.5 mm

High importance · low feasibility — enable, then do

S3 — imp 8 · feas 4 — volumetry as the primary growth metric

Feasible · lower importance — optional quick win

S2 — imp 6 · feas 8 — report micro-nodules without measuring

From diagnostic algorithms to treatment recommendations

The same engine, the full GRADE range — direction (for / against) and strength (strong / conditional) follow the panel’s consensus and the evidence certainty.

Sample output · treatment guideline

Offer adjuvant osimertinib after complete resection of stage IB–IIIA EGFR-mutated NSCLC. Consensus · FOR

Median

IQR 0.75

% for (7–9)

100%

Certainty

high

Votes

GRADERecommend (strong, high certainty)

Routinely add adjuvant immune-checkpoint inhibitor to osimertinib in EGFR-mutated disease. Consensus · AGAINST

Median

IQR 1

% in 1–3

86%

Certainty

moderate

Votes

GRADERecommend against (strong, moderate certainty)

Extend adjuvant osimertinib beyond 3 years in high-risk resected disease. Consensus · FOR

Median

IQR 2

% for (7–9)

79%

Certainty

low

Votes

GRADESuggest (conditional, low certainty) — the panel agrees, but low certainty holds it to conditional

A treatment guideline also makes the impact quantitative — projected NNT, survival, toxicity, or cost per recommendation — which can feed the importance/priority ranking. (Two reference guidelines are in progress: a diagnostic pulmonary-nodule guideline and a lung-cancer treatment guideline.)

What it does

RAND/UCLA two-region consensus — median, IQR, distribution, and a state (consensus for/against · uncertain · no consensus) with a configurable threshold and a disagreement rule.
Delphi rounds — round-by-round trajectory, stability between rounds, and a stop rule (consensus, stable “won’t converge”, or round cap).
Importance × feasibility priority map — a 7-cut quadrant plot that ranks an implementation priority list (do-first → deprioritise).
Chance-corrected ordinal agreement — interval Krippendorff’s α across the statement set, more honest than IQR alone.
GRADE / EtD recommendations — consensus + evidence certainty become a recommendation: strong “recommend” or conditional “suggest”.
Between-round feedback — a per-panelist anonymous feedback file (their prior vote + the group distribution) to run the next round.
Anonymous & private by design — aggregates only, no panelist ever named; the AI is never a panelist; nothing uploads (no network calls at all).
Part of the evidence ladder — keyed to the same protocol the rest of the Vahtian house uses, so screening (ReviewVahti) and consensus (GuidelineVahti) line up.

Who it’s for

Specialty societies & guideline groupsRun a modified-Delphi/RAND consensus with auditable numbers, not a spreadsheet.

Guideline methodologistsThe consensus, priority, agreement, and GRADE analysis in one place — in the browser.

Medical-communications teamsA defensible, reproducible consensus record for a client guideline.

Living-guideline programmesRe-vote affected statements each cycle; the data file is the versioned source of truth.

Early access

GuidelineVahti is complete and in early access. Pilot partners welcome — pricing on request.

Request early access →