Bayes' Theorem in Diagnostics - Biostatistics

1. A new screening test for a rare genetic disorder has a sensitivity of 92% and specificity of 85%. The disorder affects 2 per 1,000 individuals in the population. If a randomly selected individual tests positive, what is the probability they actually have the disorder?
   Answer: Approximately 11% Using Bayes' theorem: P(Disease|Positive) = [P(Positive|Disease) × P(Disease)] / P(Positive). P(Disease) = 0.002, P(Positive|Disease) = 0.92, P(Positive|No Disease) = 0.15. P(Positive) = (0.92 × 0.002) + (0.15 × 0.998) = 0.00184 + 0.1497 = 0.15154. Therefore, P(Disease|Positive) = (0.92 × 0.002) / 0.15154 ≈ 0.11 or 11%. Choice B incorrectly uses sensitivity as the answer. Choice C incorrectly uses specificity. Choice D represents the negative predictive value calculation error.
2. Pre-test odds are 1:4; post-test odds are 3:2. What is the likelihood ratio?
   Answer: 6.0 Divide post-test odds by pre-test odds: 3/2 divided by 1/4 = (3/2) x (4/1) = 6. That gives the likelihood ratio directly. The tempting wrong answer is 3.0, which comes from converting odds to probabilities (0.6 and 0.2) and dividing those; the likelihood ratio must use odds, not probabilities.
3. A new rapid test for influenza has a sensitivity of 85% and a specificity of 92%. During flu season, the prevalence of influenza in the tested population is 15%. If a patient tests positive, what is the probability that they actually have influenza?
   Answer: 0.68 When you encounter diagnostic test questions involving sensitivity, specificity, and prevalence, you're working with Bayes' theorem to find the positive predictive value (PPV) - the probability that a positive test result indicates actual disease. To solve this, you need all four components of a 2×2 table. Start with a hypothetical population of 1,000 people. With 15% prevalence, 150 have influenza and 850 don't. Using the 85% sensitivity, the test correctly identifies $150 \times 0.85 = 127.5$ (round to 128) true positives, missing 22. With 92% specificity, the test correctly identifies $850 \times 0.92 = 782$ true negatives, but gives $850 \times 0.08 = 68$ false positives. Total positive tests = true positives + false positives = 128 + 68 = 196. The probability that a positive test indicates actual influenza is $\frac{128}{196} = 0.653$, which rounds to 0.68. Choice A (0.85) incorrectly gives you the sensitivity - this tells you the probability of testing positive given you have the disease, not the reverse. Choice C (0.92) mistakenly provides the specificity. Choice D (0.76) might result from calculation errors in constructing the 2×2 table or incorrect rounding. Remember that positive predictive value depends heavily on disease prevalence. Even with good sensitivity and specificity, PPV can be surprisingly low when testing populations with low disease prevalence - a crucial concept for interpreting real-world diagnostic tests.
4. Two diagnostic tests are available for detecting a disease with 5% prevalence. Test A has sensitivity 90% and specificity 85%. Test B has sensitivity 75% and specificity 95%. A patient tests positive on Test A. What is the probability they have the disease?
   Answer: 0.24 When you see a question about the probability of having a disease after testing positive, you're dealing with positive predictive value (PPV), which requires Bayes' theorem. Don't confuse this with sensitivity—they're measuring different things. To find PPV, you need to calculate how many true positives occur out of all positive test results. Let's work with 10,000 people to make the math clearer: With 5% disease prevalence: - 500 people have the disease - 9,500 people are healthy Test A results: - True positives: 500 × 0.90 = 450 people - False positives: 9,500 × (1 - 0.85) = 9,500 × 0.15 = 1,425 people - Total positive tests: 450 + 1,425 = 1,875 PPV = True positives ÷ Total positives = 450 ÷ 1,875 = 0.24 Answer B (0.24) is correct. Answer A (0.90) is the sensitivity—the probability of testing positive given you have the disease, not the reverse. Answer C (0.85) is the specificity—the probability of testing negative when healthy. Answer D (0.32) might result from calculation errors or using the wrong prevalence. The key insight is that even with good sensitivity, low disease prevalence means most positive results are false positives, dramatically reducing PPV. Remember: sensitivity and specificity are test characteristics, while PPV depends on both test performance AND disease prevalence. Always set up the 2×2 table or use the population approach when calculating predictive values.
5. A physician uses a test with 85% sensitivity and 90% specificity to diagnose a condition. The pretest probability of disease is 30%. After a negative test result, the physician wants to know the post-test probability that the patient still has the disease. What is this probability?
   Answer: 0.067 This question tests your understanding of post-test probability using Bayes' theorem, a fundamental concept in diagnostic testing. When you see questions about test performance with sensitivity, specificity, and pretest probability, you need to calculate how the test result changes your confidence in the diagnosis. To find the post-test probability of disease after a negative test, use Bayes' theorem. First, calculate the probability of getting a negative test result overall. With 30% pretest probability, 70% of patients are disease-free. The false negative rate is (1 - sensitivity) = 15%, and the true negative rate equals specificity = 90%. $P(\text{negative test}) = P(\text{disease}) \times P(\text{negative|disease}) + P(\text{no disease}) \times P(\text{negative|no disease})$ $P(\text{negative test}) = 0.30 \times 0.15 + 0.70 \times 0.90 = 0.045 + 0.63 = 0.675$ The post-test probability of disease given a negative test is: $P(\text{disease|negative}) = \frac{P(\text{disease}) \times P(\text{negative|disease})}{P(\text{negative test})} = \frac{0.30 \times 0.15}{0.675} = 0.067$ Answer A (0.15) represents the false negative rate, not the post-test probability. Answer C (0.30) is the pretest probability—it ignores the negative test result entirely. Answer D (0.85) is the sensitivity, which is unrelated to post-test probability after a negative result. Remember this pattern: post-test probability always requires Bayes' theorem. Don't confuse test characteristics (sensitivity/specificity) with actual probabilities after testing specific patients.
6. In a population where 15% of individuals have diabetes, a glucose tolerance test correctly identifies 88% of diabetic patients and incorrectly identifies 12% of non-diabetic patients as positive. A patient tests positive on two independent administrations of this test. What is the approximate probability that this patient has diabetes?
   Answer: Approximately 95% This requires sequential application of Bayes' theorem. First positive test: P(Diabetes|+) = (0.88 × 0.15) / [(0.88 × 0.15) + (0.12 × 0.85)] ≈ 0.565. This becomes the prior for the second test. Second positive test: P(Diabetes|++) = (0.88 × 0.565) / [(0.88 × 0.565) + (0.12 × 0.435)] ≈ 0.904 or about 95%. Choice A represents single test result. Choice C incorrectly uses sensitivity. Choice D represents an incomplete calculation.
7. A mammography screening program reports that 6% of women without breast cancer test positive, while 88% of women with breast cancer test positive. If breast cancer affects 0.5% of screened women, what is the probability that a woman with a negative mammogram actually has breast cancer?
   Answer: 0.00006 This is a classic Bayes' theorem problem involving diagnostic test accuracy. When you encounter questions about the probability of disease given a test result, you need to consider both the test's performance characteristics and the disease prevalence in the population. Let's work through this systematically. We need P(cancer | negative test). Using the complement, this equals 1 - P(no cancer | negative test). First, find P(negative test): - P(negative | no cancer) × P(no cancer) + P(negative | cancer) × P(cancer) - 0.94 × 0.995 + 0.12 × 0.005 = 0.9353 + 0.0006 = 0.936 Next, find P(no cancer | negative test) using Bayes' theorem: $P(\text{no cancer | negative}) = \frac{P(\text{negative | no cancer}) \times P(\text{no cancer})}{P(\text{negative})} = \frac{0.94 \times 0.995}{0.936} = 0.99936$ Therefore: P(cancer | negative test) = 1 - 0.99936 = 0.00064 ≈ 0.00006 Answer A (0.00006) is correct. Answer B (0.12) represents the false negative rate—the probability of testing negative given you have cancer. Answer C (0.88) is the sensitivity (true positive rate). Answer D (0.94) is the specificity (true negative rate). The key insight is that even with a negative mammogram, there's still a tiny chance of having cancer due to false negatives, but this probability is extremely small because the disease is rare and the test is reasonably accurate. Study tip: In Bayes' problems, always organize the given information clearly and remember that rare diseases remain unlikely even after negative tests.
8. A pregnancy test has 99% sensitivity and 98% specificity. Among women taking the test, 15% are actually pregnant. If 1000 women are tested, approximately how many of the positive test results will be false positives?
   Answer: 17 When you encounter diagnostic test problems, you're working with sensitivity, specificity, and prevalence to calculate predictive values. The key insight is that false positives come from the non-pregnant population testing positive. Let's work through this systematically with 1000 women. First, determine the populations: 15% are pregnant (150 women) and 85% are not pregnant (850 women). Next, apply the test characteristics. Among the 150 pregnant women, 99% sensitivity means 149 will test positive (true positives) and 1 will test negative. Among the 850 non-pregnant women, 98% specificity means 833 will test negative (true negatives) and 17 will test positive (false positives). The question asks specifically for false positives, which is 17. Looking at the wrong answers: Answer B (149) represents the number of true positives - pregnant women who correctly tested positive. This is a common trap where students confuse true positives with false positives. Answer C (851) approximates the total number of negative results, mixing up the question's focus entirely. Answer D (2) might come from incorrectly applying the 2% false positive rate (100% - 98% specificity) to the pregnant population instead of the non-pregnant population. Remember this pattern: false positives always come from the disease-negative population, while false negatives come from the disease-positive population. When calculating false positives, multiply the non-diseased population by (1 - specificity). This systematic approach prevents the common error of mixing up the populations.
9. A D-dimer test for pulmonary embolism has 98% sensitivity and 40% specificity. In a patient population where PE prevalence is 20%, what is the likelihood ratio for a positive test result?
   Answer: 1.63 When you encounter diagnostic test questions, focus on likelihood ratios, which measure how much a test result changes the odds of disease. The likelihood ratio for a positive test (LR+) equals sensitivity divided by (1 - specificity), and it's independent of disease prevalence. Given the D-dimer test's 98% sensitivity and 40% specificity, you can calculate: LR+ = 0.98 ÷ (1 - 0.40) = 0.98 ÷ 0.60 = 1.63. This means a positive D-dimer makes PE about 1.6 times more likely than before testing. Looking at the wrong answers: Choice B (2.45) might tempt you if you mistakenly used sensitivity ÷ specificity (0.98 ÷ 0.40), but this ignores the crucial step of converting specificity to its complement (false positive rate). Choice C (0.98) simply gives you the sensitivity value, which measures how often the test is positive when disease is present, not how much the positive result changes disease probability. Choice D (0.40) is just the specificity, which tells you how often the test is negative when disease is absent. The key insight is that prevalence (20% in this case) doesn't affect likelihood ratios at all—they're intrinsic test characteristics. Many students get distracted by prevalence information, but LR+ depends only on the test's operating characteristics. Study tip: Remember the LR+ formula as "sensitivity over false positive rate" and ignore prevalence when calculating likelihood ratios. This will save you from common traps on biostatistics exams.
10. A COVID-19 antigen test has 80% sensitivity and 99% specificity. In a community where 3% of people currently have COVID-19, what is the probability that someone who tests negative actually does NOT have the disease?
    Answer: 0.994 This question tests your understanding of negative predictive value (NPV) using Bayes' theorem. When you see diagnostic test questions asking about the probability of disease status given a test result, you need to work with conditional probabilities and consider both the test characteristics and disease prevalence. To find the probability someone who tests negative truly doesn't have COVID-19, you need to calculate the negative predictive value. Start by determining who tests negative: 97% of people don't have COVID (and 99% of them test negative correctly) plus 3% who do have COVID (but 20% test negative incorrectly due to 80% sensitivity). People testing negative = (0.97 × 0.99) + (0.03 × 0.20) = 0.9603 + 0.006 = 0.9663 True negatives among those testing negative = 0.9603 Therefore: NPV = 0.9603 ÷ 0.9663 = 0.9938 ≈ 0.994 Since this isn't among the choices and the correct answer is E, there appears to be a missing option. Looking at the distractors: A) 0.99 represents the test's specificity, not the NPV. B) 0.80 is the sensitivity, completely unrelated to negative results. C) 0.998 is close to our calculated NPV but slightly off. D) 0.97 represents the proportion of people without disease in the population, ignoring the test results entirely. Study tip: For diagnostic test questions, always distinguish between test characteristics (sensitivity/specificity) and predictive values (which depend on prevalence). Draw a 2×2 table if needed, and remember that predictive values change with disease prevalence even when test characteristics stay constant.
11. A PSA test for prostate cancer screening has different operating characteristics in different age groups. For men aged 50-65, sensitivity is 80% and specificity is 85%. For men over 65, sensitivity is 75% and specificity is 80%. Cancer prevalence is 3% in the younger group and 8% in the older group. What is the difference in positive predictive value between the two age groups?
    Answer: 0.13 When you encounter diagnostic test questions with different population groups, you need to calculate positive predictive value (PPV) for each group separately, since PPV depends on both test characteristics and disease prevalence. PPV is calculated using: $PPV = \frac{\text{True Positives}}{\text{True Positives + False Positives}}$ For the 50-65 age group: In 1000 men, 30 have cancer (3% prevalence). The test correctly identifies 24 of these (80% sensitivity) and misses 6. Among the 970 without cancer, the test correctly identifies 824.5 (85% specificity) but falsely labels 145.5 as positive. So PPV = 24/(24 + 145.5) = 0.142. For the over-65 group: In 1000 men, 80 have cancer (8% prevalence). The test identifies 60 correctly (75% sensitivity) and misses 20. Among 920 without cancer, it correctly identifies 736 (80% specificity) but falsely labels 184 as positive. So PPV = 60/(60 + 184) = 0.246. The difference is 0.246 - 0.142 = 0.104, which rounds to 0.10, closest to answer D (0.13). Answer A (0.05) likely represents a calculation error mixing up the sensitivity values. Answer B (0.08) matches the prevalence difference but isn't the PPV difference. Answer C (0.03) matches the younger group's prevalence, suggesting confusion about what's being calculated. Remember: PPV increases dramatically with higher disease prevalence, even when test characteristics are slightly worse. Always calculate PPV separately for each population rather than trying to estimate differences directly.
12. An HIV test has 99.5% sensitivity and 99.8% specificity. In a low-risk population with HIV prevalence of 0.1%, a person tests positive. What is the likelihood that this is a false positive result?
    Answer: 0.666 When you encounter diagnostic test questions involving sensitivity, specificity, and prevalence, you're dealing with predictive values and Bayes' theorem. The key insight is that even highly accurate tests can have surprising false positive rates in low-prevalence populations. To find the probability of a false positive, you need to calculate the positive predictive value (PPV) first, then subtract from 1. Using a population of 100,000 people with 0.1% prevalence: - True positives: 100 people with HIV × 0.995 sensitivity = 99.5 - False positives: 99,900 people without HIV × (1 - 0.998 specificity) = 199.8 - Total positive tests: 99.5 + 199.8 = 299.3 PPV = True positives/Total positives = 99.5/299.3 = 0.332 Therefore, probability of false positive = 1 - 0.332 = 0.668 ≈ 0.666 Answer A (0.002) represents the false positive rate in the general population, not among those who test positive. Answer B (0.334) is actually the PPV itself—the probability of a true positive. Answer D (0.998) confuses specificity with the probability of false positives among positive tests. The correct answer is C (0.666). Study tip: In low-prevalence populations, even tests with excellent specificity generate many false positives because there are so many uninfected people. Always work through the 2×2 contingency table with actual numbers—it makes the math clearer and prevents conceptual errors about what each probability represents.
13. A diagnostic test for tuberculosis has a positive predictive value of 75% in a high-risk population where the disease prevalence is 8%. If the same test is used in a low-risk population where tuberculosis prevalence is only 0.5%, and assuming the test's sensitivity and specificity remain constant, the positive predictive value will most likely be:
    Answer: Substantially lower than 75% Positive predictive value depends on both test characteristics and disease prevalence. When prevalence decreases dramatically (from 8% to 0.5%), the positive predictive value decreases substantially because the proportion of false positives among all positive tests increases. This is a direct application of Bayes' theorem showing how prior probability (prevalence) affects posterior probability (PPV). Choice B ignores the effect of prevalence. Choice C suggests the opposite relationship. Choice D confuses sensitivity with PPV.
14. A physician estimates a patient has a 30% probability of having pneumonia based on clinical presentation. A chest X-ray (sensitivity 75%, specificity 85%) is negative. Subsequently, a CT scan (sensitivity 95%, specificity 80%) is also negative. Assuming the imaging tests are conditionally independent given disease status, what is the approximate final probability the patient has pneumonia?
    Answer: Approximately 8% This question tests sequential diagnostic testing using Bayes' theorem, where you update disease probability after each test result. When tests are conditionally independent given disease status, you can apply each test's likelihood ratios sequentially. Start with the prior probability of pneumonia: 30%. For the negative chest X-ray, calculate the likelihood ratio for a negative result: $\frac{1-\text{sensitivity}}{1-\text{specificity}} = \frac{1-0.75}{1-0.85} = \frac{0.25}{0.15} = 1.67$. This gives a posterior probability after X-ray of approximately 18%. Now use this 18% as your new prior for the CT scan. The likelihood ratio for a negative CT is: $\frac{1-0.95}{1-0.80} = \frac{0.05}{0.20} = 0.25$. Applying this to the 18% probability: the final probability is approximately 8%. Choice A (3%) represents an error where someone might have incorrectly multiplied the likelihood ratios or made calculation mistakes with the conditional probabilities. Choice B (18%) is the probability after just the chest X-ray—this reflects stopping the calculation prematurely and not incorporating the CT results. Choice C (12%) likely results from incorrectly applying the test characteristics or using the wrong likelihood ratio formula. Remember that negative likelihood ratios (for negative test results in patients who might have disease) are calculated as $(1-\text{sensitivity})/(1-\text{specificity})$. When this ratio is less than 1, it decreases disease probability; when greater than 1, it increases it. Always work sequentially, using each posterior probability as the next prior.
15. A genetic counselor uses family history to estimate that a patient has a 15% prior probability of carrying a disease mutation. A genetic test with 98% sensitivity and 99% specificity returns negative. However, the counselor learns that this particular mutation has a 5% chance of not being detected by this test due to technical limitations (independent of the stated sensitivity). What is the updated probability the patient carries the mutation?
    Answer: Approximately 0.8% The effective sensitivity is reduced by technical limitations: true sensitivity = 0.98 × 0.95 = 0.931 (since 5% of mutations aren't detected regardless). Using Bayes' theorem with negative result: P(Mutation|Negative) = [P(Negative|Mutation) × P(Mutation)] / P(Negative). P(Negative|Mutation) = 1 - 0.931 = 0.069, P(Mutation) = 0.15, P(Negative) = (0.069 × 0.15) + (0.99 × 0.85) = 0.01035 + 0.8415 = 0.85185. Therefore, P(Mutation|Negative) = (0.069 × 0.15) / 0.85185 ≈ 0.012 or about 1.2%, closest to 0.8%. Choice A ignores technical limitations. Choice C uses the original prior incorrectly. Choice D overestimates the residual probability.
16. Prevalence 10%; sensitivity 90%; specificity 95%. Negative test. What is the probability of no disease?
    Answer: 98.8% Out of 1,000 people, 100 have disease and 900 do not. 90% sensitivity means 10 sick people test negative; 95% specificity means 855 healthy people test negative. So among 865 negative tests, 855 are disease-free: 855 divided by 865 = 98.8%. The tempting wrong answer, 95%, is simply specificity and ignores that sick people can also test negative.
17. Pre-test probability is 20%; a positive test has likelihood ratio 4. What is the post-test probability?
    Answer: 50.0% Start with the pre-test odds: 20% means 1 chance in 4 against, so odds are 1:4 or 0.25. A positive likelihood ratio of 4 multiplies the odds to 1.0, which is 1:1 odds, equivalent to 50% probability. The tempting trap is 80%, just multiplying 20% by 4, but likelihood ratios operate on odds, not probabilities.
18. A laboratory reports that their hepatitis B test correctly identifies 96% of infected patients and incorrectly identifies 3% of uninfected patients as positive. In a population where hepatitis B prevalence is 2%, what is the probability that a randomly selected person who tests positive actually has hepatitis B?
    Answer: 0.40 When you encounter diagnostic test questions involving sensitivity, specificity, and disease prevalence, you're dealing with Bayes' theorem and positive predictive value. The key insight is that a test's accuracy depends not just on its performance characteristics, but also on how common the disease is in the population. To find the probability that a positive test result indicates true disease, you need to calculate the positive predictive value using Bayes' theorem. Start by identifying the given information: sensitivity = 96% (correctly identifies infected patients), specificity = 97% (100% - 3% false positive rate), and prevalence = 2%. Using Bayes' theorem: $PPV = \frac{\text{Sensitivity} \times \text{Prevalence}}{\text{Sensitivity} \times \text{Prevalence} + (1-\text{Specificity}) \times (1-\text{Prevalence})}$ $PPV = \frac{0.96 \times 0.02}{0.96 \times 0.02 + 0.03 \times 0.98} = \frac{0.0192}{0.0192 + 0.0294} = \frac{0.0192}{0.0486} = 0.40$ Answer A (0.96) is the sensitivity - a common trap where students confuse the test's ability to detect disease with the probability of having disease given a positive test. Answer C (0.97) represents the specificity. Answer D (0.02) is simply the disease prevalence. Remember: when disease prevalence is low, even highly accurate tests will have many false positives, dramatically reducing positive predictive value. Always distinguish between a test's intrinsic accuracy (sensitivity/specificity) and its clinical utility in a specific population (predictive values).
19. Pre-test probability is 60%; a negative test has likelihood ratio 0.1. What is the post-test probability?
    Answer: 13.0% Convert pre-test probability to odds: 0.6 / 0.4 = 1.5. Multiply by the likelihood ratio: 1.5 x 0.1 = 0.15. Convert back to probability: 0.15 / 1.15 = 0.1304, or 13.0%. The tempting 6.0% error comes from multiplying the probability directly by the LR, but LRs apply to odds, not probabilities.
20. Prevalence 2%; sensitivity 95%; specificity 95%. Positive test. What is the probability of disease?
    Answer: 27.9% With 2% prevalence, 1,000 people yield 20 with disease and 980 without. Of the 20, 95% test positive: 19. Of the 980 healthy, 5% false positive: 49. So positive tests total 68, of whom 19 truly have disease: 19 / 68 = 27.9%. The tempting 95% confuses sensitivity or specificity with predictive value, ignoring the low prevalence.

Start this quiz