Correlation vs Causation – Interactive Practice
Correlation vs Causation – Interactive Practice
Click through the short scenarios below. Your job: decide what the data really tell you – and avoid jumping from “linked” to “causes”.
For each situation, decide if the result is just a correlation, likely causation, or if there isn’t enough information.
Q1. A large randomized clinical trial shows Drug X lowers blood pressure more than placebo over 6 months.
Q2. A cohort study finds people who walk 10,000 steps per day have lower blood pressure than those who walk 2,000 steps.
Q3. In summer, ice cream sales and drowning incidents both increase.
A confounder is a hidden third factor that affects both sides of an association. Pick the most plausible confounder.
Q1. Coffee intake is associated with higher rates of heart disease. Which confounder fits best?
Q2. People who take daily multivitamins live longer. Which factor is the best confounder?
For each correlation, decide which direction of causality is most plausible.
Q1. Patients on higher doses of pain medication have higher pain scores. What is most likely?
Q2. Depression scores are higher in patients with uncontrolled diabetes. What can you say?
Choose the weakest study design that would still be reasonable to support the claim.
Q1. Claim: “Drug X reduces 1-year mortality after myocardial infarction.” What is the minimum reasonable design?
Q2. Claim: “There may be a new rash side effect with a recently approved drug.” What is the minimum design needed to raise concern?
Click “Show suggested headline” to compare a sensational claim with a more honest, evidence-based version.
Q1. Study: case series of 10 mild diabetic foot ulcers treated with a new cream. Most healed in 2 weeks. No control group.
Q2. Study: observational data show people with higher vitamin D levels had lower COVID-19 rates. No randomization.
Decide if each pattern looks like a potential safety signal or more like background noise.
Q1. Drug A: 50,000 patients exposed. Two cases of a very rare autoimmune disease. Expected background: 1–2 per 50,000 per year. No known mechanism.
Q2. New monoclonal antibody: 5,000 exposed, 25 serious thrombosis events (expected 2–3), and the mechanism affects coagulation.
Read the mini-timeline. Decide if a causal link is plausible and then reveal the suggested reasoning.
Q1. A patient develops a pulmonary embolism 5 days after starting a new hormonal therapy. Mild shortness of breath was present 1 week before the drug started but worsened after.
- Did the drug start clearly before the clot?
- Could the clot have been forming already?
- What extra tests or history would you want?
First, mentally put these designs in order from weakest to strongest for showing causation. Then reveal a suggested ranking.
Q1. How would you order these for showing that a drug reduces mortality?
Case report → Uncontrolled case series → Case–control study → Cohort study → Single randomized controlled trial → Meta-analysis of RCTs.
Decide what you would do with each study result, then compare with a suggested answer.
Q1. Small open-label study, 30 patients with venous ulcers, new cream, no control group, high healing rate.
Q2. Large RCT (1,500 pts) shows a new oral drug reduces stroke by 20% vs standard of care, with similar safety.
Answer each question correctly to unlock the next step. When all are correct, you’ll see the final takeaway.
Step 1. A registry finds that patients on Drug Y have better survival than those not on it, but treated patients are younger with fewer comorbidities. What is this?
Step 2. What is the best name for this problem?
Step 3. Which design is best to move from correlation toward causal evidence for Drug Y’s benefit?
Final message: Correlation is a clue. Causation is a verdict. Good clinical decisions depend on knowing the difference.
