2- Sampling Variation, Bias and Confounding Flashcards
When sampling a population what do we need it to be?
- Unbias
- Representative
- Precise
What are the two types of error that may occur in a study?
- Random (chance): decreases as sample size increases
- Bias (systematic): does not decrease. quantified by difference between true value and actually value
How do you increase precision?
Increase sample size. Decreases uncertainty
What are sources of bias?
- Selection Bias
- Information Bias
What are some selection biases?
1. Non-representative study sample
2. Groups within study may not be comparable
3. Healthy worker effect (people who aren’t working may be due to illness and disease so not necessarily the fact they aren’t working that is causing them to die)
What are some information biases?
- Recall error
- Interviewer/Observer error (predispositions)
- Measurement error (measurer or equipment)
- Misclassification (some people without disease diagnosed as having disease, mainly due to measurement or observational error)
What does no bias, high precision look like?
Does a large study mean less bias?
NO!
but means more precise
Define prevalence.
The proportion of people who have a disease at a given point in time
NOT A RATE - per 1000 people then made into a percetage
Define incidence and state how you can calculate incidence rate?
Incidence is the number of new cases of a diseases within a given time frame.
Useful when monitoring epidemics
What is person years and how is it calculated?
Sum of the total time of everybody followed up in a study
e. g 3 people followed for 5 years and two people followed for 3 years =
(3x5) + (2x3) = 21 py’s
How else can prevalence be measured?
Incidence x Duration of Disease
What is the IRR and how is it calculated?
Incidence Rate Ratio
Used to compare rate of incidence in two groups
What is relative risk? (a.k.a risk ratio)
Absolute risk of one group / Absolute risk of another group
The comparison of the probability of an event occuring in one group compared to another
What factors affect the length of a disease?
- Mortality
- Curability
How do you work out the odds of an event occuring?
It is a ratio
Diseased/Non-diseased for example
or
P/1-P
What is the odds ratio?
Odds ratio is ratio of odds of outcome in groups defined by levels of exposure at a particular time
Comparison of odds of disease in two groups
When can relative risk and odds ratios be used?
- When all patients in the groups have been consistently followed up
What is the absolute risk of disease?
Disease/Diseased+Non-diseased
How do we interpret OR?
- Ratio risk of the group of interest is less than the risk of the comparison group
- Ratio of 1: risk of the group of interest is the same as the risk of the comparison group
- Ratio > 1: risk of the group of interest is greater than the risk of the comparison group
What is risk difference?
Absolution difference in risk between two groups. If zero there is no difference in risk
What is a confounding variable?
An unobserved factor that impacts both exposure and can cause disease
or
When comparing groups the association between exposure and outcome is distorted by another variable
e.g alcohol, gender, mouth cancer
How can we account for confounding variables?
- Standardisation: e.g weight, age, sex
- Some confounding variables unknown so difficult to address
