This topic is worth a small number of questions, so it deserves a small amount of your time. Everything the USMLE has ever asked about Relative Risk and Odds Ratio fits on these pages, and like most high-yield Step 1 topics, the exam does not test the theory. It tests whether you can name the study design and pick the matching measure before you calculate anything.
Draw the table this way every single time, with exposure down the side and outcome across the top.
| Disease present | Disease absent | Row total | |
|---|---|---|---|
| Exposed | a | b | a + b |
| Not exposed | c | d | c + d |
| RR = [ a / (a+b) ] / [ c / (c+d) ] OR = (a/b) / (c/d) = ad / bc |
| THE ONE SENTENCE TO MEMORIZE Relative Risk should be used when incidence can be measured, whereas Odds Ratio is primarily used in case control studies where incidence cannot be directly calculated. |
This table decides the answer to most questions on this topic before any arithmetic begins. Biostatistics and epidemiology are a small but reliable slice of every form, our USMLE Step 1 content outline shows exactly where they sit and how much to prioritize them.
| Study design or analysis | Correct measure | Why |
|---|---|---|
| Prospective or retrospective cohort | Relative Risk | Groups defined by exposure and followed forward, so incidence exists |
| Randomized controlled trial | Relative Risk, plus ARR and NNT | Randomization creates the exposure groups |
| Case control study | Odds Ratio only | The investigator fixes how many cases are enrolled, so incidence is meaningless |
| Logistic regression, any design | Adjusted Odds Ratio | The model produces odds by construction |
| Cross sectional study | Prevalence Odds Ratio | Exposure and outcome measured at the same moment, no temporality |
| Survival or time to event analysis | Hazard Ratio | Accounts for when events occur and handles censoring |
That’s how USMLE Impact Theory teaches every high-yield topic, recognize the pattern first, then the answer follows. Built by IMGs, for IMGs.
These two cohorts share an identical Relative Risk yet produce very different Odds Ratios.
| Scenario | Exposed | Unexposed | RR | OR |
|---|---|---|---|---|
| Common outcome | 50 of 100 affected | 25 of 100 affected | 2.0 | 3.0 |
| Rare outcome | 20 of 1000 affected | 10 of 1000 affected | 2.0 | 2.02 |
The Odds Ratio always lies further from 1 than the true Relative Risk. When the outcome is rare, roughly under 10 percent, the two are nearly identical, which is the rare disease assumption. When the outcome is common, the Odds Ratio overstates the association. Whenever a stem hands you both an RR and an OR from the same data, this is what it is testing.
A prospective cohort follows 2000 adults for five years. Of 400 with untreated obstructive sleep apnea, 60 develop atrial fibrillation. Of 1600 without OSA, 80 develop atrial fibrillation.
For comparison, the odds ratio here is (60 x 1520) divided by (340 x 80), which equals 3.35, further from 1 than the true RR. Attributable risk is 15 percent minus 5 percent, or 10 percent.
A case control study enrols 120 patients with esophageal adenocarcinoma and 240 controls. Long standing reflux is reported by 84 cases and 60 controls, giving a equal to 84, b equal to 60, c equal to 36 and d equal to 180.
These memory tricks stick best when you review them on a schedule instead of cramming once. Drop the study-to-measure pairings and null values into spaced repetition, our guide on how to use Anki for Step 1 shows how to set that up so they never leak out on test day.
1. RR = Real Risk = Real incidence. Real people followed forward.
2. OR = Odds = Often Retrospective. Odds ratios live where you look backward.
3. Cases get Odds. Cohorts get Risk.
4. Rows for RR, Cross for OR.
5. One is None. A ratio of 1 means no association.
6. If the CI spans one, the finding is undone.
7. Rare means Share. Rare outcome, OR and RR share a value.
8. Common is a Con. Common outcome, the OR exaggerates.
9. LOG gives ODDS. Logistic regression always reports odds ratios.
10. Less than one, protection has begun.
1. RR equals incidence in exposed divided by incidence in unexposed.
2. OR equals ad divided by bc.
3. Risk Ratio and Relative Risk are the same measure.
4. Both are null at 1. Differences are null at 0.
5. Neither can be negative, since both are ratios.
6. Case control studies cannot yield RR, ARR, attributable risk or NNT.
7. Logistic regression reports odds ratios in every design.
8. Cox regression reports hazard ratios for time to event data.
9. The OR always lies further from 1 than the true RR.
10. Under roughly 10 percent event frequency, OR approximates RR.
11. Significant when the 95% confidence interval excludes 1.
12. Wider intervals mean smaller samples and less precision.
13. Statistical significance is not clinical significance.
14. Recall bias threatens case control studies, loss to follow up threatens cohorts.
15. A large RR on a tiny baseline risk is still a tiny absolute risk.
Book a free call with mentors who matched as IMGs. We’ll build a plan that drills these traps across biostats, pharm, micro, and path.
A prospective cohort follows 300 hospitalized patients. Among 100 catheterized beyond 72 hours, 40 develop a urinary tract infection. Among 200 catheterized for less than 72 hours, 20 develop infection. What is the relative risk?
A. 0.25 B. 2.0 C. 4.0 D. 6.0 E. 30 percent
Answer: C. 4.0. Risk in the exposed is 0.40 and in the unexposed 0.10, so RR equals 4.0. Option D is the odds ratio of 6.0, which is far above the true RR because this outcome is common. B ignores the denominators and E is the absolute risk difference.
Investigators identify 200 patients admitted with Guillain Barre syndrome and compare them with 400 matched hospitalized controls, then interview all participants about influenza vaccination in the preceding six weeks. Which measure can validly be calculated?
A. Relative risk B. Absolute risk reduction C. Number needed to treat D. Odds ratio E. Attributable risk
Answer: D. Odds ratio. Participants were selected by outcome and questioned about prior exposure, which defines a case control study. Every other option is a risk based measure requiring incidence, which cannot exist when the investigator fixes the number of cases.
A cohort study reports that prolonged anesthesia is associated with postoperative delirium, with a relative risk of 1.8 and an odds ratio of 2.6 from the same data. Delirium occurred in 32 percent of the cohort. What best explains the difference?
A. A computational error B. Unmeasured confounding C. The outcome is common, so the OR overestimates the RR D. Selection bias E. Only the OR was adjusted
Answer: C. The odds ratio always sits further from 1 than the relative risk, and the gap widens as the outcome becomes frequent. At a 32 percent event rate the rare disease assumption fails, so this divergence is expected rather than erroneous.
A trial assigns 3000 patients with heart failure to an SGLT2 inhibitor or placebo. Hospitalization occurs in 12 percent of the treatment arm and 20 percent of the placebo arm. How many patients must be treated to prevent one hospitalization?
A. 8 B. 12 C. 13 D. 25 E. 40
Answer: C. 13. The absolute risk reduction is 0.08 and the NNT is 1 divided by 0.08, which is 12.5, always rounded up. Option A is the ARR itself, B rounds down and E is the relative risk reduction of 40 percent.
USMLE Impact Theory covers every high-yield topic the same way, NBME-style vignettes, wrong-answer analysis, and the exact traps written into the stems.
Relative Risk compares probabilities using row totals, while Odds Ratio compares odds using the diagonal cross product. Relative Risk requires incidence data and Odds Ratio does not, which is why they belong to different study designs.
The investigator decides how many cases and controls to enrol, so the proportion with disease is a design choice rather than a natural frequency and incidence cannot be calculated. The odds ratio is unaffected by that sampling.
They converge when the outcome is rare, conventionally below about 10 percent, which is the rare disease assumption. As the outcome becomes common the odds ratio moves further from 1 and exaggerates the association.
The exposed group has twice the risk of the unexposed group, which is a 100 percent relative increase. Whether that matters clinically depends entirely on the baseline risk.
For any ratio measure, check whether the 95 percent confidence interval includes 1. If it excludes 1 the result is significant at an alpha of 0.05, corresponding to a p value below 0.05.
1. Risk divides everyone. Odds divide only by those without the outcome.
2. RR uses rows. OR uses diagonals, as ad over bc.
3. Cohort and RCT give RR. Case control gives OR.
4. Logistic regression always gives an adjusted OR.
5. Cross sectional gives a prevalence OR, survival analysis gives a hazard ratio.
6. Both are null at 1. Differences are null at 0.
7. Above 1 means increased risk, below 1 means protective.
8. The OR always lies further from 1 than the true RR.
9. Rare outcome means OR approximates RR. Common outcome means it exaggerates.
10. Significant when the 95% confidence interval excludes 1.
11. RRR is 1 minus RR. ARR is the difference in event rates. NNT is 1 over ARR, rounded up.
12. Identify the design first, choose the measure second, calculate last.
Biostatistics rewards a little time spent well, not hours of memorizing formulas. Fitting focused reps like this into your day is what makes them stick, our breakdown of how many hours a day to study for Step 1 helps you budget the biostat block without overspending on it.
This is exactly how USMLE Impact Theory works, built by IMGs, for IMGs, on the same recognize-then-solve engine you just used to master a topic most students dread. Here’s what that looks like across the whole exam:
References:
USMLE Content Outline (FSMB and NBME).
CDC, Principles of Epidemiology in Public Health Practice,
Lesson 3: Measures of Risk. Cochrane Handbook,
Chapter 6: Choosing effect measures. STROBE and CONSORT statements.
Guyatt et al, Users Guides to the Medical Literature. Tenny and Hoffman, StatPearls.
Articles published by IMG Helping Hands are prepared by our team using information from direct experience, publicly available resources, and educational references. AI tools may be used to assist with drafting, proofreading, and formatting; however, all content undergoes review and approval before publication.
The information provided is intended for educational purposes only. Requirements, policies, and processes may change over time. Readers should consult official sources for the most current information.
Get our most valuable tips right inside your inbox, once per month!
