Biostatistics is the subject medical students skip, then regret skipping. It looks small next to pathology and pharmacology, so it gets pushed to the last week of dedicated. Then the exam hands you a 2x2 table, a Kaplan-Meier curve, and a paragraph-long abstract, and the points you assumed were free turn out to be the ones you never practiced.
The math itself is not hard. There are about a dozen formulas, and most of them are one line. The hard part is retrieving the right one, with the right numbers in the right places, while a question stem tries to distract you with a clinical story. That is a retrieval problem, which is exactly what Anki is for.
How Much Biostatistics Is Actually on the Exam
Biostatistics and epidemiology make up 4 to 6 percent of Step 1, according to Lecturio's breakdown of the exam content. Step 2 CK is similar: the official USMLE content outline gives biostatistics, epidemiology, and interpretation of the medical literature 3 to 5 percent of the exam. Step 2 CK has up to 318 questions, so that is roughly 10 to 16 questions. Not a huge share, but these are some of the most gettable points on the test. A biostatistics question either clicks or it does not, and there is no partial credit for almost remembering the formula.
The other reason to take it seriously is that biostatistics keeps showing up after the exam. Step 3 tests it, shelf exams test it, and every journal club and attending pimp session in residency will assume you can read an abstract. The cards you make now stay useful for years, which is more than you can say for most preclinical minutiae.
The Best Resources Before You Make Cards
Biostatistics punishes students who try to learn it from flashcards alone. You need one pass through the concepts first, then cards to keep them. The good news is that one pass is short.
First Aid
First Aid has a dedicated biostatistics chapter for Step 1, and First Aid for Step 2 CK has an epidemiology chapter. Both are dense and formula-heavy, which makes them good card-making material. Read the chapter once, work the example problems by hand, and make cards from the formulas and definitions as you go. If you only use one resource, use this one.
Randy Neil
Randy Neil's biostatistics videos are the standard recommendation for a reason. His first two-part series for Step 1 runs under 60 minutes total and walks through the commonly tested topics as USMLE-style questions, with the arithmetic shown step by step. Students with almost no statistics background start here. His later videos cover drug ads, abstracts, Kaplan-Meier curves, and bias and confounding, each with practice questions. Watch at normal speed the first time. The point is to follow the calculation, not to collect another video watched.
UWorld and AMBOSS
UWorld has a dedicated biostatistics question set, and AMBOSS has detailed biostatistics articles with Anki integration that lets you pull explanations straight into your cards. Do the questions after your First Aid pass, not before. When you miss one, make a card for the exact fact you missed, and paste the explanation into the Extra field. A missed-question card on odds ratios is worth ten generic definition cards.
Mehlman Medical
Mehlman Medical publishes a biostatistics review built entirely around practice questions with worked answers. It is a good second pass if you have finished First Aid and Randy Neil and still feel shaky on abstracts and study design. Use it as a question source, not as reading material.
The Card Types That Actually Work
Biostatistics needs four kinds of cards. Most students only make the first kind, then wonder why they still miss the questions.
1. Formula cards with real numbers
A formula card that only asks for the formula is half a card. You also need to practice plugging numbers into it. Make both versions:
- Front: "Sensitivity formula?" Back: "TP / (TP + FN). True positives divided by everyone who actually has the disease."
- Front: "A test catches 80 of 100 people with the disease. Sensitivity?" Back: "80%. 80 / 100."
Do the same for specificity (TN / (TN + FP)), positive predictive value (TP / (TP + FP)), and negative predictive value (TN / (TN + FN)). The exam gives you the 2x2 table and asks you to compute, so your cards should train the computation, not just the recitation.
Add the two likelihood ratios the same way. Positive likelihood ratio is sensitivity divided by (1 minus specificity). Negative likelihood ratio is (1 minus sensitivity) divided by specificity. A positive likelihood ratio above 10 is strong evidence for disease, and a negative likelihood ratio below 0.1 is strong evidence against it. Those two cutoffs show up in questions constantly.
2. The 2x2 table as a single image occlusion
Draw one 2x2 table, label the four cells, and occlude them one at a time. True positive is top left, false positive is top right, false negative is bottom left, true negative is bottom right. Occluding the table trains the spatial layout, and the layout is what lets you answer a question in ten seconds instead of thirty.
Put the derived formulas in the Extra field of each cell card so that every review reinforces the whole table. When you see a question stem with a table of numbers, your first move should be to label the cells. If you have drilled the table, that move is automatic.
3. Discrimination cards for study designs
Study design questions are discrimination questions. The stem describes a study, and four answer choices are four designs. Your cards should train the distinction, not the definition:
- Front: "Researchers measure disease and exposure at a single point in time. Design, and what can it measure?" Back: "Cross-sectional study. Measures prevalence. Cannot establish causation."
- Front: "Study starts with people who have the disease and looks backward at exposure. Design, and what does it calculate?" Back: "Case-control study. Calculates odds ratio. Best design for rare diseases."
- Front: "Study starts with exposed and unexposed people and follows them forward. Design, and what does it calculate?" Back: "Cohort study. Calculates relative risk and incidence. Strongest observational design for causal inference."
- Front: "Investigators assign the exposure themselves, usually by randomization. Design?" Back: "Randomized controlled trial. The gold standard for causal inference."
Add a card for each design's main weakness. Case-control studies are vulnerable to recall bias and selection bias. Cohort studies lose people to follow-up. Cross-sectional studies cannot tell you which came first. These weaknesses are the answer to a large share of study design questions.
4. Bias cards written as vignettes
Bias is the most missed biostatistics topic because students memorize definitions and the exam tests recognition. Build your bias cards as short scenarios, the way the exam presents them:
- Front: "A screening program finds cancers earlier, so survival looks longer even though the date of death does not change. Bias?" Back: "Lead-time bias."
- Front: "A screening program finds slow-growing cancers more often, so the screened group looks like it survives longer. Bias?" Back: "Length bias, also called length-time bias."
- Front: "People with a disease remember past exposures better than healthy controls. Bias?" Back: "Recall bias."
- Front: "Hospital-based cases and controls differ systematically from the general population. Bias?" Back: "Selection bias. Berkson bias is the classic example."
- Front: "A third factor is associated with both the exposure and the outcome, distorting the relationship. What is it, and what fixes it at the design stage?" Back: "Confounding. Randomization controls for it in the design stage. Stratification and multivariate analysis adjust for it after the fact."
Notice the pattern: the front is a situation, the back is a name. That is the direction the exam tests. If your cards only ask "What is lead-time bias?" you are training the wrong direction.
5. Image occlusion for curves
Two graphs show up repeatedly, and both are easier to learn as pictures than as text. Screenshot them from First Aid or your lecture slides and occlude the labels one at a time.
The ROC curve plots sensitivity against 1 minus specificity. The diagonal line is a useless test. The further the curve bows toward the top left corner, the better the test discriminates. Occlude the axis labels first, then the curve itself, and ask yourself what the shape means before you flip the card.
The Kaplan-Meier curve plots survival over time, with each step down marking an event. The key facts to card: the curve estimates the probability of surviving past a given time, censoring is shown as tick marks for patients lost to follow-up, and the hazard ratio compares the event rate between two groups. A hazard ratio of 0.6 means the treatment group has 40 percent lower risk of the event at any given time. Students confuse hazard ratio with relative risk, so make a card that forces the distinction: hazard ratio is about the rate of events over time, relative risk is about the proportion who have the event by the end.
The Formulas Worth Memorizing Cold
If your time is short, these are the ones to get to automatic recall. Everything else in the subject is definitions and study design.
- Sensitivity = TP / (TP + FN). High sensitivity is good for screening. Remember it as SNOUT: a sensitive test, when negative, rules out disease.
- Specificity = TN / (TN + FP). High specificity is good for confirmation. Remember it as SPIN: a specific test, when positive, rules in disease.
- Positive predictive value = TP / (TP + FP). Negative predictive value = TN / (TN + FN). Both depend on prevalence. As prevalence rises, PPV rises and NPV falls.
- Absolute risk reduction = event rate in the control group minus event rate in the treatment group. Number needed to treat = 1 / ARR. Always round NNT up to the next whole person.
- Number needed to harm = 1 / attributable risk, where attributable risk is the incidence in the exposed group minus the incidence in the unexposed group.
- Relative risk = incidence in exposed divided by incidence in unexposed. Odds ratio = odds of exposure in cases divided by odds of exposure in controls. Use relative risk for cohort studies and trials, odds ratio for case-control studies.
- Standard deviation and the 68-95-99.7 rule: about 68 percent of values fall within one standard deviation of the mean, 95 percent within two, 99.7 percent within three.
- A confidence interval that crosses the null value is not statistically significant. For a relative risk or odds ratio, the null value is 1. For a difference between means, the null value is 0.
That list is roughly a dozen cards. A dozen cards is a rounding error in a deck of thousands, and it covers most of what the exam asks about formulas.
A Study Schedule That Fits in One Weekend
Biostatistics is one of the smallest Step 1 subjects by card count, which is why it is so easy to finish and so easy to skip. A realistic plan:
- Saturday morning: Watch Randy Neil's two-part biostatistics series, under 60 minutes total. Pause and work each calculation yourself before he reveals the answer.
- Saturday afternoon: Read the First Aid biostatistics chapter and make your formula, 2x2, study design, and bias cards as you go. Aim for 60 to 80 cards total. If you already use AnKing, unsuspend its biostatistics tags instead and add your own cards only for gaps.
- Sunday: Do 30 to 40 UWorld biostatistics questions. For every miss, add one card for the missed fact. Do not add a card for the whole question, just the fact.
- Daily for two weeks: Keep the biostatistics cards in your regular review queue. They take under five minutes a day once mature, and that is the whole point. The subject is small enough that two weeks of reviews makes it permanent.
If you are studying for Step 2 CK, the same plan works with the First Aid Step 2 CK epidemiology chapter swapped in, and the question count is higher because abstracts and drug ads get more emphasis. Step 2 CK also tests interpretation of the medical literature directly, so add cards for hazard ratios, Kaplan-Meier curves, and intention-to-treat analysis.
Common Mistakes to Avoid
- Memorizing formulas without numbers. A formula you cannot compute under time pressure is a formula you do not know. Every formula card needs a numbers version.
- Making definition cards for bias. The exam gives you a scenario and asks for the name. Your cards should do the same, or you will recognize the definition and still miss the question.
- Confusing odds ratio and relative risk. Odds ratio comes from case-control studies. Relative risk comes from cohort studies and trials. If your card does not say which design goes with which measure, add that to the card.
- Forgetting that PPV and NPV depend on prevalence. Sensitivity and specificity are properties of the test. PPV and NPV change with the population. This distinction is tested directly, often with a low-prevalence screening scenario.
- Treating p-values as the whole story. A p-value under 0.05 means the result is unlikely to be due to chance alone. It does not tell you the effect size, the clinical importance, or whether the study was well designed. The exam tests this confusion.
- Skipping the abstract and drug-ad questions. Step 2 CK and Step 3 both include questions built around abstracts and drug advertisements. They look long and intimidating, which is why students skip them in practice and then face them cold on exam day. Do them in UWorld, and make cards for the traps: relative risk reduction without absolute numbers, surrogate endpoints, and subgroup analyses.
Biostatistics FAQ
How many biostatistics cards do I need? Far fewer than any other subject. The whole topic is roughly 60 to 100 cards if you build your own, or the AnKing biostatistics tags if you already use AnKing. The small size is the trap: students assume small means unimportant, and it means the opposite. Small and formula-based is the easiest material to turn into guaranteed points.
Is biostatistics high yield for Step 1? Yes, relative to its size. At 4 to 6 percent of the exam, it is worth more questions than most single organ systems, and the questions are faster to answer once the formulas are automatic. It is also one of the few subjects where a weekend of work covers nearly everything testable.
Should I learn biostatistics from videos or from First Aid? Both, in that order. Watch Randy Neil first to understand the calculations, then read First Aid to consolidate and make cards. Videos alone leave you unable to recall formulas cold. First Aid alone leaves you unable to follow a worked problem. The combination takes about three hours total.
What about Step 2 CK biostatistics? It is the same core with more emphasis on abstracts, drug ads, and interpreting the literature. The content outline gives it 3 to 5 percent, which is 10 to 16 questions on a 318-question exam. Add cards for Kaplan-Meier curves, hazard ratios, intention-to-treat versus per-protocol analysis, and the phases of clinical trials. Phase 1 is safety and dose finding in a small group, phase 2 is efficacy and side effects, phase 3 is a large comparison against standard treatment, and phase 4 is post-marketing surveillance.
How do I handle biostatistics on in-house exams? In-house exams test whatever your professor emphasized, which is often not the same as the boards. If your lecturer spent an hour on a specific study design or a local screening program, make cards from those slides. That is the one place where turning your own lecture slides into cards beats a premade deck, because no premade deck knows what your professor cares about.
Bottom Line
Biostatistics is a small, formula-driven subject that rewards a weekend of focused work and a few minutes of daily review. Learn the calculations from Randy Neil, consolidate them from First Aid, and build cards that train computation and recognition instead of definitions. Drill the 2x2 table until labeling it is automatic, write bias cards as scenarios, and keep the study design distinctions sharp.
The students who lose biostatistics points are not the ones who find it hard. They are the ones who never got around to it. Sixty to a hundred cards, two weeks of reviews, and the subject stops being a weakness.