Lecture #12 - Cohort studies Flashcards
What do cohort studies set out to answer?
Why a pattern of distribution/health outcome might be occurring
What is a cohort study?
- Allows for what kind of epidemiology?
- What kind of study design?
- “Individuals are defined on the basis…..”
- So what is a cohort study in essence?
- Allows for analytic epidemiology: exposures and outcomes
- Observational study design: oversee people’s exposures and what happens to them (as opposed to intervention where people assigned exposures)
- “Individuals are defined on the basis of the presence or absence of exposure to a suspected rick factor.”
- Just look at whether people exposed to a particular risk factor or not - follow them overtime and see if they develop the outcome you’re interested in.
What are the six steps in a cohort study?
- Identify a source population
- Recruit your sample population (the sample population must not already have the outcome of interest)
- Assess exposure to identify which group participants belong in (exposed/comparison)
- Follow up overtime
- Observe whether/not participants develop the outcome (fill in relevant numbers in GATE)
- Calculate measures of association
What can you measure using cohort studies?
What are their formulae?
Measures of occurrence (couldn’t with CSS bc ppl don’t have outcome at start of study so we can measure incidence of new cases)
- Cumulative incidence
- (no. of ppl develop in time period)/(no. of ppl at risk at start) - Incidence rates (incidence density)
- (no. of ppl develop in period)/(no. of person-years at risk of developing)
- takes into account the diff lengths of time ppl in study
What else can you measure using cohort studies?
What are their formulae?
Measures of association (since can calculate the incidence)
- RR
- rate ratio: Ie/Ic
- risk ratio: CIe/CIc
- “as likely” - RD (attributable risk)
= Ie - Ic
= CIe - CIc
-“how many more cases”
Example of cohort study?
Dunedin multidisciplinary health and development study
Go to book and look through the example of a cohort study
-
What might you need to consider carefully with these steps:
- Identify a source pop
- Recruit your sample pop
Under this, what is the healthy worker effect?
Ideal: random selection independent of exposure status
-so some have exposure and some don’t etc
Sometimes: Selection based on exposure status, consider appropriate comparison group
I.e. If wanna study rare exposure or something that only occurs in particular group of people then recruit exposed group from particular setting where exposure occurred else might not have enough people who’ve been exposed to tell you much about anything in study (recruit comparison group separately too but make sure they’re similar to exposed in terms of all aspects that could relate to exposure except exposure
Health worker effect:
- Can occur if exposed group is an occupational group - comparison group selected from general population
- People who’re working are generally healthier than general population because in general pop, you have people can’t work because unwell etc - this is a form of selection bias
Another thing to consider regarding sample population and outcome?
Can you be sure that the sample pop doesn’t already have the outcome
- In general, cohort studies can determine temporal relationship bw exposures and outcomes
- excludes those who have already got the outcome (can’t follow up, obi) so can see temporal sequence
- There can be preclinical stage for some diseases - disease might’ve started but can’t detect it (exposure might not be nicely proceeding outcome as you think)
What might you need to carefully with this step:
- Asess exposure to identify which group participants belong in (exposed or not exposed i.e. comparison)
Have the participants been correctly classified?
What might you need to carefully with this step:
- Follow up over time
-Have the participants CHANGED EXPOSURE status over time?
e.g. runner who used to run recreationally are now competing in marathon so training way more
So follow up overtime esp for exposures susceptible to change over regular intervals to detect changes happening to exposure status
- Has everyone been followed up over the LENGTH OF THE STUDY?
e. .g lost to follow up - some might move away/can’t contact or decide they don’t wanna be part of study - How LONG do participants need to be followed up?
e. g. if looking at something that takes long time to develop - need to have study go on for that long. Or if study rare outcome - need to accumulate enough people with outcome to make study meaningful
What might you need to carefully with this step:
- Observe whether or not participants develop outcome
Has the outcome been correctly classified?
What are the 4 strengths of cohort studies?
- Determines temporal sequence bw exposure and outcome (CSS can’t!)
- Can examine multiple outcomes from exposure (can study more than one exposure too)
- Can calculate incidence (therefore can calculate RR and RD)
- Good for studying rare exposures
What are the 5 limitations of cohort studies?
- Loss to follow up
- Can lead to bias if related to exposure and the outcome - Potential for misclassification of exposures/outcomes
- Generally not good for studying rare outcomes (might be following 1000 people but if rare - only a few get outcome so not much you can say about exposure and outcome)
- Time consuming e.g. Dunedin study
- Can be expensive (esp if long)
What are the two types of cohort studies and explain the second type in detail by answering these questions:
- What is it? (2)
- Advantages (3)
- Disadvantages (3)
What’s the basic difference between the two types?
- Prospective cohort study
- do the whole, “classify exposures, follow up, assessing which people do/don’t develop outcome” but to overcome the problem of time - can consider doing
- case control
- historical cohort - Historical cohort studies
- use existing data
- reconstruction follow-up period in the past
Advantages:
- less time consuming compared with perspective
- good for outcomes that take a long time to develop
- less expensive
Disadvantages:
- Use existing data (collected for other reasons) - quality?
- May not know about all relevant factors
- Selection bias?
- The basic difference is that historical cohort studies start and end and perspective start from start.
