Lilac Module 2 (L9 - L16) Flashcards
Lecture 9: What is the purpose of the comparison group in analytic epidemiological studies
The purpose of the comparison group in analytic epidemiological studies is to estimate how common the outcome is in people without the exposure you’re looking at. It is important to know this because most health outcomes have multiple causes. For example, even though lung cancer is strongly associated with smoking, people who have never smoked will still get lung cancer. SO THE COMPARISON GROUP IS OUR WINDOW INTO HOW OFTEN WE WOULD EXPECT THE OUTCOME TO OCCUR IN THE ABSENCE OF THE EXPOSURE WE’RE INTERESTED IN.
To determine whether an exposure is associated with an outcome (i.e. changes the risk of developing the outcome), you need to compare to occurrence of the outcome in a group with the exposure and the occurrence of the outcome in a group without the exposure. If the occurrence of the outcome is different in the two groups, this suggests that the exposure may be associated with a change in the occurrence of the outcome, and therefore potentially a determinant (we cover the things you need to think about before considering whether exposure actually is a determinant of an outcome in Module Three). If the measure of occurrence for the outcome is higher in the exposed group, for example, that would suggest that the exposure might be a risk factor for the outcome.
Lecture 10: The NNT is just a number and not the only thing you need to consider when evaluating a treatment. Other things you might want to think about include:
- What are the HARMS associated with the treatment? It may save lives, but all treatments have risks as well. If you are giving everyone in the population a treatment, you need to make sure the risks are minimal.
- What OTHER TREATMENTS are there? Maybe there are better treatments available, or ones that are safer.
- What is the COST of this treatment? Even a really low NNT may not be compelling if the treatment costs so much that the health system can’t afford to run hospitals after buying it!
Lecture 11: What is the measure of occurrence calculated in a cross-sectional study design?
Prevalence
Lecture 11: Explain why establishing temporal sequence is a problem for cross-sectional studies.
Temporal sequence = when the exposure occurs before the outcome develops
Cross-sectional studies measure exposures and outcomes at the same point or period in time. It is therefore often not possible to clearly establish that an exposure occurred before an outcome developed. This is required to demonstrate that there is temporal sequence, and is crucial to determining whether an exposure is causally associated with an outcome.
Lecture 11: What is meant by the term ‘ecological fallacy’, and what study design most commonly has this problem?
School and breakfast example
The ecological fallacy is most commonly associated with ecological studies. The fallacy occurs when an association found to occur at the group level does not occur at the individual level. For example, it might be found that schools that provide breakfasts for their students have better pass rates (a group level association). This might lead you to conclude that school breakfasts improve student performance. But it might be found that, within these schools, the students who eat breakfast at home are the ones passing (because of more stable home environments, for example) (an individual level association). So the school (group) level association does not apply at the student (individual) level.
“Means you can’t ascribe characteristics of that group to the individuals within that group.”
Lecture 12: What are the strengths (3) and limitations (6) of cohort studies?
Strengths
• Able to establish TEMPORAL SEQUENCE
• Enables the direct calculation of INCIDENCE rates and cumulative incidence (measures of
occurrence), which means relative risks and risk differences (measures of association) can
be calculated
• The study design allows for the investigation of MULTIPLE EXPOSURES AND OUTCOMES at the
same time
Limitations
• Not efficient for investigating outcomes that take a LONG time to develop, as this means you would need to follow participants for a very long period of time
• Not efficient for investigating extremely RARE OUTCOMES, as this means you would need to include a very large number of people in your study
• Not efficient for investigating TRANSIENT/INTERMITTENT EXPOSURES
• Issues with BIAS – cohort sample selection, loss to follow-up, and misclassification of
exposure and outcome
• May be extremely EXPENSIVE
• Management of the study is COMPLEX, especially across long periods of time
Lecture 12: Are cohort studies suitable if the outcome is rare and it takes a long time to develop? Justify your answer.
Cohort studies may not be the most suitable study design if the outcome of interest is rare and/or takes a long time to develop. In these situations, you may require an extremely large number of participants (potentially hundreds of thousands), and to follow-up these people over a very long period of time (possibly decades). These conditions increase the expense, time, and complexity of the study, and may cause problems with loss to follow-up bias.
Lecture 12: Are cohort studies suitable if the outcome is already present in the study population? Justify your answer.
No, because one of the basic pre-requisites of the cohort studies is that the participants are free of the outcome at the beginning of the study (i.e. they are ‘at risk’ of developing the outcome). This pre-requisite strengthens the cohort study’s ability to demonstrate causation, as the exposure is measured in the participants first, and then they are followed through time to observe if there is a difference in the occurrence of the outcome between the exposed and comparison groups (i.e. temporal sequence can be established).
Lecture 13: In epidemiology, we compare groups to identify possible determinants of a health outcome. As a group, what do the controls represent?Lecture 13:
The control group represents an estimate of the odds of the exposure in the source population from which the cases were selected. If a harmful exposure under investigation is a risk factor for the outcome then you would expect that the odds of exposure would be higher for cases compared to controls.
Lecture 13: What issues might using prevalent cases, as opposed to incident cases for case-control studied, introduce?
We are interested in exposures that happened prior to the outcome. In a case-control study a prevalent case may have had the outcome for a long period and might not clearly remember if they had the exposure prior to their original diagnosis, especially if it is a transient or intermittent exposure. Also, because we are looking at prevalent cases (rather than incident cases), the probability of a person being included in the study as a case depends in part on how long they have the outcome (remember, prevalence is dependent on an outcome’s incidence and average duration). If the exposure we are looking at is related to how long someone has the outcome, the odds ratio for the study will partly reflect the exposure’s impact on whether someone develops the outcome and partly reflect how long they have it.
Lecture 13: Example of using prevalent vs incidence cases in case-control study
Consider the following example: Suppose we wanted to know whether people from high socioeconomic status areas are more likely to develop heart disease than people from low socioeconomic status areas (so our exposure is higher socioeconomic status and our comparison is low socioeconomic status). For our cases we choose people who had diagnosed heart disease on the 31st December 2014. These are prevalent cases, and they will be a mix of people who have been newly diagnosed with heart disease and people who have had it for a long time. It is this latter group of people who are going to cause us problems, because how long you survive with heart disease (i.e. duration) is associated with your socioeconomic status – people of higher socioeconomic status tend to survive longer with it. Including these people in our cases is going to mean that the number of exposed people in the cases is greater than it should be (i.e. the odds of the exposure in the cases will be greater than it should be). So the resulting odds ratio will tell us the association between socioeconomic status and both developing heart disease and surviving with it, when we just wanted to know about the association with developing heart disease.
Lecture 13: Why can’t you calculate prevalence or incidence when using a case-control study design?
1) The incidence or prevalence of the outcome in the cases will always be 100% (by definition, they all have/develop the outcome), while the incidence or prevalence of an outcome in the controls will always be 0% (again, by definition controls do not have the outcome).
2) The prevalence or incidence of an outcome in all study participants (both the cases and the controls) also cannot be calculated. In a case-control study, you are recruiting a certain number of cases and then a specific ratio of controls (usually from one to four controls per case). The prevalence or incidence of the outcome is therefore a function of the number of controls you recruit, not the actual prevalence or incidence of the outcome. So, for example, if you recruit one control for every case, and you recruit 100 cases, your prevalence and incidence of the outcome will be 50% (100/(100+100)), even if the incidence of the outcome in the source population is really low.
