Biostats Flashcards
How is the sensitivity of a test defined? What are highly sensitive tests used for clinically?
SNOUT SN detects Dz True POS/(true+false POS) Screening Sensitivity is defined as the ability of a test to detect disease, and mathematically as the number of true positives divided by the number of people with the disease. Tests with high sensitivity are used for disease SCREENING. False positives occur, but the test does not miss many people with the disease (low false-negative rate). One way to remember this is the word snout, written “Sn-N-out,” meaning with high sensitivity (Sn) a negative (N) test rules out (out) the disease.
How is the specificity of a test defined? What are highly specific tests used for clinically?
SPIN SP detects Health True NEG/(false +true NEG) Confirmation Specificity is defined as the ability of a test to detect health (or nondisease), and mathematically as the number of true negatives divided by the number of people without the disease. Tests with high specificity are used for disease confirmation. False negatives occur, but the test does not identify anyone who is actually healthy as sick (low false-positive rate). The ideal confirmatory test must have high sensitivity and high specificity; otherwise, people with the disease may be identified as healthy. One way to remember this is the word SPIN written “Sp-P-in,” meaning that with high specificity (Sp) a positive (P) test rules in (in) the disease.
Explain the concept of a trade-off between sensitivity and specificity.
The trade-off between sensitivity and specificity is a classic statistics question. For example, you should understand how changing the cutoff glucose value in screening for diabetes (or changing the value of any of several screening tests) will change the number of true- and false-negative and true- and false-positive results. If the cutoff glucose value is raised, fewer people will be identified as diabetic (more false negatives, fewer false positives), whereas if the cutoff glucose value is lowered, more people will be identified as diabetic (fewer false negatives, more false positives). As an example, if the diagnostic threshold for a fasting blood sugar for diabetes were raised from ≥125 mg/dL to ≥300 mg/dL, most people with diabetes would be missed (low sensitivity because a patient with blood sugar of 285 mg/dL would be negative for diabetes according to this criterion). In addition, the test would be very specific for patients with blood sugar ≥300 mg/dL (a patient would certainly have diabetes if he had a positive test).
Define positive predictive value (PPV). On what does it depend?.
How likely diseased when test is positive NPV=True Pos/ All Pos Depends on prevalence and SN/SP When a test is positive for disease, the PPV measures how likely it is that the patient has the disease (probability of having a condition given a positive test). PPV is calculated mathematically by dividing the number of true positives by the total number of people with a positive test. PPV depends on the prevalence of a disease (the higher the prevalence, the higher the PPV) and the sensitivity and specificity of the test (e.g., an overly sensitive test that gives more false positives has a lower PPV).
Define negative predictive value (NPV). On what does it depend?
How likely healthy when test is negative NPV=True Neg/ All Neg Depends on prevalence and SN/SP When a test is negative for disease, the NPV measures how likely it is that the patient is healthy and does not have the disease (probability of not having a condition given a negative test). It is calculated mathematically by dividing the number of true negatives by the total number of people with a negative test. NPV also depends on the prevalence of the disease and the sensitivity and specificity of the test (the higher the prevalence, the lower the NPV). In addition, an overly sensitive test with many false positives leads to a higher NPV.
Define attributable risk. How is it measured?
Attributable risk is the number of cases of a disease attributable to one risk factor (in other words, the amount by which the incidence of a condition is expected to decrease if the risk factor in question is removed). For example, if the incidence rate of lung cancer is 1:100 in the general population and 10:100 in smokers, the attributable risk for smoking in causing lung cancer is 9:100 (assuming a properly matched control group).
Given the 2 × 2 table in the following table, define the formulas for calculating the test values indicated. + - + A|B - C|D
SN = A/A+C SP = D/B+D PPV = A/A+B NPV = D/C+D Odds Ratio = AD/BC Relative risk = (A/A+B)/(C/C+D) Attributed Risk = (A/A+B)-(C/C+D)
Define relative risk. From what type of studies can it be calculated?
Relative risk compares the disease risk in people exposed to a certain factor with the disease risk in people who have not been exposed to the factor in question. Relative risk can be calculated only after prospective or experimental studies; it cannot be calculated from retrospective data. If a Step 3 question asks you to calculate the relative risk from retrospective data, the answer is “cannot be calculated” or “none of the above.”
What is a clinically significant value for relative risk?
Any value for relative risk other than 1 is clinically significant. For example, if the relative risk is 1.5, a person is 1.5 times more likely to develop the condition if exposed to the factor in question. If the relative risk is 0.5, the person is only half as likely to develop the condition when exposed to the factor; in other words, the factor protects the person from developing the disease. RR >1 Risk factor RR <1 Protective
Define odds ratio. From what type of studies is it calculated?
The odds ratio attempts to estimate relative risk with retrospective studies (e.g., case-control). An odds ratio compares two factors—(1) the incidence of disease in persons exposed to the factor and the incidence of nondisease in persons not exposed to the factor and (2) the incidence of disease in persons unexposed to the factor and the incidence of nondisease in persons exposed to the factor—to see whether there is a difference between the two. As with relative risk, values other than 1 are significant. The odds ratio is a less than perfect way to estimate relative risk (which can be calculated only from prospective or experimental studies).
What do you need to know about standard deviation (SD) for the USMLE?
68-95-99.7 You need to know that for a normal or bell-shaped distribution, the mean ± 1 SD contains 68% of the values, the mean ± 2 SD contains 95% of the values, and the mean ± 3 SD contains 99.7% of the values. A classic question gives the mean and SD and asks what percentage of values will be above a given value. For example, if the mean score on a test is 80 and the SD is 5, 68% of the scores will be within 5 points of 80 (scores of 75 to 85) and 95% of the scores will be within 10 points of 80 (scores of 70 to 90). The question may ask what percentage of scores are over 90. The answer is 2.5% because 2.5% of the scores fall below 70 and 2.5% of the scores are over 90. Variations of this question are common.
Define mean, median, and mode.
The mean is the average value, the median is the middle value, and the mode is the most common value. A question may give several numbers and ask for their mean, median, and mode. For example, if the question gives the numbers 2, 2, 4, and 8: The mean is the average of the four numbers: (2 + 2 + 4 + 8)/4 = 16/4 = 4. The median is the middle value. Because there are four numbers, there is no true middle value. Therefore take the average between the two middle numbers (2 and 4), so the median = 3. The mode is 2, because the number 2 appears twice (more times than any other value). Remember that in a normal distribution, mean = median = mode.
What is a skewed distribution? How does it affect the mean, median, and mode?
A skewed distribution implies that the distribution is not normal; in other words, the data do not conform to a perfect bell-shaped curve. Positive skew is an asymmetric distribution with an excess of high values; in other words, the tail of the curve is on the right (mean > median > mode) (Fig. 1-3). Negative skew is an asymmetric distribution with an excess of low values; in other words, the tail of the curve is on the left (mean < median < mode). Because such distributions are not normal, the SD and mean are less meaningful values.
Define test reliability. How is it related to precision? What reduces reliability?
From a practical perspective, the reliability of a test is synonymous with its precision. Reliability measures the reproducibility and consistency of a test. For example, if the test has good interrater reliability, the person taking the test will get the same score if two different people administer the same test. Random error reduces reliability and precision (e.g., limitation in significant figures).
Define test validity. How is it related to accuracy? What reduces validity?
From a practical perspective, the validity of a test is synonymous with its accuracy. Validity measures the trueness of measurement; in other words, whether the test measures what it claims to measure. For example, if a valid IQ test is administered to a genius, the test should not indicate that he or she has an intellectual disability. Systematic error reduces validity and accuracy (e.g., when the equipment is miscalibrated).
