15 points every clinical researcher should verify before protocol submission.
By Coefficients Health Analytics · Free to use & share
Phase 1 — Research Question & Causal Structure
01PICO/PECO is explicit and falsifiableVague questions → vague designs. Define population, exposure/intervention, comparator, and outcome precisely.
02Causal diagram (DAG) drawn before analysis planIdentifies confounders, mediators, and colliders. Prevents over-adjustment and M-bias. Not optional for observational studies.
03Target estimand stated ATE / ATT / CATEDifferent estimands → different methods → different interpretations. Decide before touching data.
04Study design matches the causal questionNot all questions need an RCT. Not all questions can be answered with a cohort. Match design to the estimand.
Phase 2 — Bias & Validity Assessment
05Immortal time bias evaluatedIf exposed group must survive to receive exposure, you've biased toward survival. Use landmark analysis or time-varying exposure.
06Selection bias mechanism identified (or ruled out)Who enters the study? Who drops out? Collider stratification bias lurks in case-control and hospital-based designs.
07Measurement error plan documentedNon-differential misclassification biases toward null. Differential misclassification can go either way. Both need assessment.
08Sensitivity analysis for unmeasured confounding planned E-valueReport E-value with confidence interval bound. No observational study should omit this.
Phase 3 — Analysis & Power
09Sample size justified with clinically meaningful effect sizeStatistically significant ≠ clinically meaningful. Power for the effect size that would change practice.
10Primary analysis pre-specifiedFishing for significance is not analysis. One primary outcome, one primary method, registered in advance.
11Missing data mechanism assessed MCAR / MAR / MNARComplete case analysis assumes MCAR (almost never true). Multiple imputation assumes MAR. MNAR requires sensitivity analysis.
12Competing risks addressed (if time-to-event)Kaplan-Meier overestimates risk when competing events exist. Use cumulative incidence function (CIF) + Fine-Gray or cause-specific hazards.
Phase 4 — Reviewer Preparedness
13DAG + balance diagnostics includedLove plot with SMD < 0.1 threshold. Reviewers will ask — have it ready.
14Limitations section anticipates methodological critiquesDon't wait for Reviewer 2 to find the weakness. Name it yourself, explain mitigation, cite precedent.
15Reporting guideline followed STROBE / CONSORT / PRISMAMandatory for most journals. Missing items = desk rejection. Use the correct checklist for your design.