Cohort studies Flashcards
Explain basic overview of a cohort study [1]
You start by identifying a cohort (big group of people) and measure your exposures of interest in those people (e.g. smoking, cholesterol, diet, blood pressure, etc) and follow them up over time, seeing who is exposed and who isn’t, and who does / doesn’t have the disease in these groups.
Name three biases in cohort studies [3]
- Loss to follow up (i.e. following individuals up for many years, individuals can move away, not respond to contact, etc)
- Exposure is usually only measured at just one time point. Can not be sure that there is significance if you ask at only one time point with big gaps between these. However, in cohort studies you can ask multiple times yet it is often done at the beginning
- Selection of cohort: are the representative of the general population? If not, could you really apply these results to the general population?
What is an issue with cohort studies? [1]
Confounding issues: associated with both factor and disease
E.g. if heavy smokers are also drinkining - which is causing cancer?
Name three disadvantages of cohort study [3]
- Take a long time
- Need a lot of people (for example: 120,000 in a cohort study about red meat consumption and risk of cancer)
- Very expensive
In cohort studies, risk refers to WHAT? [1].
Incidence: no. who die per 100, 000
Explain how you usually calculate incidence in cohort studies [1]
Incidence is usually taken to be a measure of risk, and is the number of new cases or deaths of a disease per 100,000 people per year
How do you calculate relative risk of cohort study? [1]
Relative risk: this is calculated by:
incidence of disease in exposed population / incidence of disease in unexposed population
How can you tell from this evidence that smoking does increase the risk of lung cancer and IHD? [1]
In these examples, the 95% CIs for both lung cancer and ischemic heart disease do not include 1
If the risk of lung cancer for smokers is the same as the risk for non-smokers, the relative risk would be WHAT VALUE? [1]
1
State three methods you can adjust for confounders by creating an adjusted relative risk? [3]
Stratify: calculate the relative risk for alcohol and lung cancer among smokers and non-smokers separately. Then take a weighted average of the two relative risks, with weights based on the number of smokers and non-smokers
Log-binomial regression
Poisson regression
How do you assess how long someone lives? [2]
Kaplan-Meier curves: plot the proportion of people surviving without an event over time
In survival analysis hazard ratios are calculated
They take time to death/event into account
Whats the difference in interpretation between relative risk and hazard ratios?
Interpretation of hazard ratio similar to that for relative risks - but give the risk of dying at any time point in one group compared to the other.
How do you calculate absolute excess risk? [1]
What does absolute risk tell you? [1]
Absolute excess risk = Risk in Exposed – Risk in Unexposed
Shows your absolute risk of dying.
What is the following a definition of?
proportion of deaths are actually attributable to the exposure you are looking at [1]
Attributable proportion
How do you calculate attributable proportion? [1]
Incidence in population attributable to exposure / Incidence in population
= p (relative risk - 1) / 1 + p (relative risk - 1)
where p = proportion exposed in population
