Clinical Evidence Assistant – correlation vs causation
Clinical Evidence Assistant
Quick bedside/desktop tool for HCPs to think through correlation vs causation and decide how far they can go with an association before changing practice or reacting to a safety signal. (Educational support – not a substitute for guidelines or labels.)
Step 1
What kind of data are you looking at?
Choose the main design that best fits your study or dataset.
Step 2
Key questions about causation
Temporality – did the exposure clearly come first?
Example: drug started before effect / event, not after.
Randomization or strong control of confounders?
Random allocation or careful adjustment for age, sex, comorbidities, severity, etc.?
Effect size and consistency?
Is the effect moderate–large and fairly consistent across subgroups or studies?
Biological / clinical plausibility?
Is there a sensible mechanism or prior evidence to support the causal story?
Multiple studies?
Are there several independent studies that point in the same direction?
Step 3
Your decision context
What decision are you mainly thinking about for this data?
Overall evidence strength (auto-updated)
–
Start by answering questions or load a solved example →
No assessment yet.
How to read this:
- 0–2: Very weak – hypothesis only.
- 3–4: Weak – correlation; treat as a "maybe."
- 5–6: Moderate – possible causation; more work needed.
- 7–8: Strong – likely causal under study conditions.
- 9–10: Very strong – consistent causal signal (still not absolute).
Study-type snapshot
Select a study type or load an example to see a brief comment.
Decision guidance for your context
Choose a context (practice change, safety signal, etc.) or load an example to see tailored advice.
Solved examples
See the tool in action
Click any example below to auto-fill the inputs with a realistic clinical scenario. Then read the score and decision guidance.
Example 1 – Observational survival benefit (Drug A registry)
A cohort/registry shows patients on Drug A have 30% lower mortality than non-users, but Drug A is given mainly to younger, less comorbid patients.
You are considering changing practice.
Example 2 – Large RCT reducing stroke
A well-conducted RCT of 1,500 patients shows Drug B reduces stroke by ~20% vs standard of care with similar safety.
You are deciding whether to update local practice/guidelines.
Example 3 – Early thrombosis safety signal (new mAb)
Post-marketing, 25 serious thrombosis cases are reported with a new monoclonal antibody among 5,000 exposed; expected background is 2–3.
Mechanism plausibly affects coagulation. You are in a safety signal meeting.
Example 4 – Single case report of severe rash
One patient develops a severe rash shortly after starting a new drug. Timing fits, but no other cases are known yet.
You are considering how to handle this individual safety case.
