Show:
Biases
Case-Control
Cohort
Experimental
Potent Potables
100
What method can one use to minimize investigator bias?
What is masking/blinding?
100
Relative to a cohort study, what is the cost of a case-control study?
What is Low?
100
What happens to the results of a cohort study with large levels of drop out? Why?
What is can potentially bias the results, as people who drop out or are not compliant are different than those who do comply. May result in selection bias.
100
What is the difference between observational and experimental designs?
What is only experimental design is RCT, researcher can manipulate and assign exposure states. Allows for full randomization. All other study designs are observational.
100
In which study designs can you calculate relative risk?
What is cohort and randomized trials?
200
In an exercise trial with two arms: an 8 week intervention or a no-treatment "usual care" group. Patients who enroll in the study are seeking an exercise program. However, those who are randomized to the control group register for an exercise program at their local YMCA. What type of threat is this to your study?
What is internal validity?
200
Case control studies are good studying what types of outcomes or what types of exposures?
What is good for studying rare diseases?
200
Does a retrospective cohort need to have the same number of persons in the cohort and a concurrent cohort study?
What is yes....?
200
Name three reasons for using randomization?
What is attempting to have groups comparable at baseline, inability to predict next assignment, elimination of selection bias in treatment allocation
200
What is a type 2 error, and how does one calculate power for a given study design?
What is the error when researchers conclude that treatment groups are not different, when in reality they are different. The probability of committing a type 2 error is beta. To calculate power it's 1-beta.
300
Regarding recall bias, what study design is it likely to occur and how does it influence the estimation of exposure?
What is case control study, and often over-estimates exposure.
300
When designing a case control study, is one using incident or prevalent cases? What's the problem with each choice?
What is using incident cases is excluding patients who died prior to disease diagnosis and having to wait for new cases; and prevalent cases is it's unknown if the exposures being studied are related more to survival with the disease than disease development.
300
Name two differences between a nested case-control study and a case-cohort design?
What is controls are only individually matched with nested CaCo, and only with nested design are cases and controls matched on calendar time and length of follow up.
300
What are the two ways to assess interaction between factors?
What is looking at the observed verses the expected joint effects, and looking at the heterogeneity of effects across levels of another variable.
300
What is the problem with the interpretation of attributable risk as "the amount of disease reduction if exposure were removed"?
What is it implies that the exposure is necessary and sufficient, and that the exposed and unexposed groups have equal exposures to other not necessary but sufficient causes.
400
What is the difference between efficacy and effectiveness?
What is efficacy is under clinical/ideal conditions, and effectiveness is in less-controlled and real-world situations.
400
Selection of controls in a case control study is one of the great banes of epidemiology. Using all hospital controls from one hospital threatens what about the study results?
What is there is a lack of generalizability, and a threat to external validity as hospitalized patients are not representative of the population as a whole especially from a singular source.
400
When cross-over occurs during clinical trials, there are two ways to analyze: intent to treat vs as treated. Which one is better and what is the problem with the lesser?
What is ideally one would use intent to treat as is least biased and most conservative. The as-treated analysis as it has violated randomization and may be biased as to selection of most compliant and lowest risk population.
400
What is the difference between quanitative and qualitative interaction?
What is qualitative shows effect in different directions with addition of factors, or in the (rare) case of no interaction to an increase. Quantitative shows the magnitude increases and varies across strata (basically number goes up).
400
When assessing for interaction, what does the type of scale used tell you about the type/presence of interaction?
What is there are two types of interaction: additive and multiplicative, and if present they will be seen graphically. If the interaction is additive, one should use an arithmetic scale and if the interaction is multiplicative one would use a logarithmic scale. If an effect is multiplicative, it will also be additive too.
500
What is the issue with controlling for confounding variables? (Bonus point: other than adjusting for counfouding, what is the other way to handle confoudning variables in analysis?)
What is once a variable is adjusted for, it is no longer able to be studied. (Bonus answer: stratification)
500
In a case control study design, you find that the stratified OR between exposures are significantly different. You've ruled out chance and bias as an explanation. Is there interaction, if so what kind, and why?
What is yes, interaction and not confounding because we don't know more information. The interaction is multiplicative, by nature, because we have calculated OR.
500
What are some ways to express incidence in cohort studies?
What is depends on your denominator used. Numerator is always defined, but could use population at beginning of study (probability or risk) population at time of event (hazard) population at midpoint of follow up (rate/average) or rate per person time/density.
500
What is the difference between non-probability samples and random samples? Give an example of each
What is non-probability samples include elements of the population that have no chance of selection (convenience sample, purposive sampling). Random sampling is scheme for each part of population has chance of being selected (simple random sample, systematic sampling, stratified sampling and two stage sampling)
500
Go to the board and draw the following example. Tell me all you can about it.
See Jess' example