# Content-Validity Reviewer Guide v0.1

**Purpose:** Independent review of the four pilot estimands. This review evaluates whether the questions are well defined and meaningful. It does not ask reviewers to rank drugs or invent numerical risks.

## Reviewer roles sought

Aim for a panel that collectively includes epidemiology, toxicology, addiction medicine, psychiatry, emergency medicine, biostatistics/causal inference, harm reduction, health economics/policy, and lived experience. No person is expected to represent every domain.

## Before discussion

1. Read `PILOT_ESTIMANDS_V0.2.md` and `DATA_FEASIBILITY_REVIEW_V0.3.md`.
2. Complete `reviewer-conflict-disclosure.csv`.
3. Independently complete every assigned row in `estimand-content-validity-form.csv`.
4. Do not view other reviewers' ratings until Round 1 closes.

## Rating scale

- **1 — not relevant/clear:** material redesign required.
- **2 — somewhat relevant/clear:** major revision required.
- **3 — quite relevant/clear:** minor revision only.
- **4 — highly relevant/clear:** retain as written.

A numeric rating without a short rationale is incomplete.

## Required written judgments

For each estimand, identify:

- the most important omitted population or context;
- any overlap or double counting among outcomes;
- whether exposure and comparator can be operationalized;
- whether numerator and denominator can come from compatible data;
- wording that could be misinterpreted by the public;
- whether the intended use should be narrower.

## Independence and adjudication

Original ratings remain visible after discussion. The author team may revise wording but cannot replace independent ratings with a consensus average. Accepted and rejected suggestions receive a public rationale. Material revisions increment the estimand version and trigger a second review round.

## Conflict handling

A disclosed conflict does not automatically exclude a reviewer. The panel lead decides whether disclosure, partial recusal or full recusal is appropriate for each estimand. Industry, advocacy and personal experience should be represented transparently rather than hidden.

## What counts as passing

Numerical thresholds must be fixed before ratings are opened. Regardless of threshold, the gate fails if reviewers identify an unresolved construct overlap, non-operational exposure/comparator, incompatible numerator/denominator, or major omitted context that changes intended use.