Study Design
Measures of Effect
Screening
Interpretation and Causation
Bias and Confounding
100
In this type of study, participants are selected on the basis of their disease status.
What is a case control study?
100
The risk ratio describing the following situation: 100 kids live in homes with poor indoor air quality. 20 develop asthma. 200 kids are not exposed. 10 develop asthma.
What is RR=(20/100) / (10/200) = .2/.05 = 4
100
3 criteria that make a disease appropriate for screening.
What are: 1. serious consequences if untreated 2. detectable before symptoms appear 3. better outcomes if treatment begins before clinical detection?
100
Interpretation of a relative risk of 2.4, with a p-value of 0.08.
What is: There is an 8% chance of observing this risk ratio of 2.4, or a more extreme result by chance alone, given that there is no relationship in the source population?
100
Explanation of recall bias and an example.
What is Differential accuracy of recall between cases and controls. Infant mortality: mothers may be more likely to recall prenatal exposures if infant dies
200
A study that compares the relationship between unemployment and infant mortality by county is an example of this type of study.
What is an ecologic study?
200
The interpretation of the following study: Case control study in men over 60 OR = 3.5 Exposure: high cholesterol Outcome: stroke
What is: Among men over 60, the odds of high cholesterol among those who had a stroke were 3.5 times higher than the odds of high cholesterol among those who did not have a stroke?
200
Definition of sensitivity, and how to calculate it.
What is Probability (proportion) of ?correct classification of cases Sensitivity = true positives / all cases ?
200
Interpretation of the following 95% confidence interval: RR=2.4 (1.7-2.9)
What is: If the study were repeated 100 times, 95 out of 100 times the CI of 1.7-2.9 would contain the true population relative risk.
200
Description of and example of selection bias in a case control study.
What is Cases and controls have a different probability of being selected according to exposures or outcomes of interest, creating a biased measure of association (i.e. odds ratio)? -response bias, exclusion bias
300
Definition of the study base.
What is a defined population whose disease experience during some period of time is the source of the study data? What is the population that gave rise to the cases?
300
Definition of attributable risk.
What is the excess occurrence of a disease among the exposed that can be attributed to the exposure? Example: The excess occurrence of cardiovascular disease among male smokers attributable to their smoking is 43 per 100,000.
300
Calculation and interpretation of specificity in the following scenario: True negatives: 150 False positives: 25
What is specificity = 150/(150+25) = .86 86% of non-cases are correctly identified using this test.
300
Rankings of common study designs, in order of their ability to determine causation, from best to worst.
What is: Randomized controlled trials Cohort studies Case control studies Cross sectional studies Ecologic studies
300
The three criteria for classical confounding, and an example.
What is 1. a confounder must be associated with the exposure in the source population 2. a confounder must be a risk factor for disease 3. a confounder must not be an intermediate cause between exposure and disease Example?
400
Description of a retrospective cohort study.
What is a cohort that is reconstructed with data on the predictor variable (measured in the past) and data on the outcome collected (measured in the past after some follow-up period)? Important: data on the exposure was recorded before the outcome occurred (may be in medical records, school records, etc. etc.)
400
Incidence rate of myocardial infarction (MI) in this study, in cases per 100 person-days: participant # days in study MI? 1 19 Yes 2 70 No 3 24 Yes 4 70 No 5 53 Yes
What is: IR = 3 cases/236 person days = .0127 cases/person-day IR = 12.7 cases per 1000 person-days
400
Relationship between prevalence and predictive value of a test.
What is: The same test (with the same sensitivity and specificity) when applied under scenarios with different prevalence can give very different positive and negative predictive values. Higher prevalence --> higher positive predictive value
400
A causal graph for HIV treatment adherence including at least 6 risk factors.
????
400
Assessment and interpretation of the following situation: You conducted a cohort study looking at the association between gambling and cancer. The crude rate ratio is 4.3. Among smokers, the rate ratio is 2.1 Among non-smokers, the rate ratio is 1.9
What is: Smoking is a confounder, because the crude RR is different from the stratified RRs, and the stratum-specific RRs are the same as each other.
500
Three disadvantages of a case control study.
What are: 1. susceptible to bias 2. restricted to a single outcome 3. cannot calculate incidence
500
Predicted weight of a 3-year old child (y) who weighed 3000g at birth, whose mother weighed 120 pounds before pregnancy, if: child's weight = 21.1 + 0.0380(mother's weight) + 0.00227(birth weight)
What is 32.5 pounds? 32.5 = 21.1 + 0.0380(120) + 0.00227(3000)
500
Sensitivity and specificity of the following screening test: Cases Non-cases Positive 125 1000 Negative 30 15,000
What is: Sensitivity = 125/155 = 80.6% Specificity = 15000/16000 = 93.75%
500
Interpretation of these multiple linear regression results: Outcome: maternal mortality rate in Iranian provinces Coefficient p-value Human Development Index -247.6 <0.001 Gini coefficient (inequality) 51.0 0.22 Female literacy rate -2.1 0.03
The mortality rate in Iran is inversely related to the HDI and literacy rate among women. A one unit increase in HDI is associated with a 247.6 unit decrease in the maternal mortality rate. A one unit increase in the literacy rate among women is associated with a 2.1 unit decrease in maternal mortality. The Gini coefficient does not predict maternal mortality in the adjusted model.
500
Comparison of the types of bias common in case control vs. cohort studies, and ways to prevent them.
What is: Case control - selection bias (make sure that cases and controls come from same study base), recall bias (memory aids, calendars) Cohort - response bias (careful recruitment procedures), loss to follow up (appropriate incentives, ease of participation) Misclassification is possible in both types of studies
M
e
n
u