What is epidemiology?
The study of frequency, distribution + determinants of diseases and health related states in populations in order to prevent + control disease
What is the difference between prevalence + incidence?
Incidence (cumulative incidence):
- New cases
- Denominator (number of disease free people at the start of the study)
- Time
Prevalence:
- existing cases
- denominator (the population at risk of having that disease)
- point in time (point prevalence)
What is person-time?
A measure of time at risk - i.e. the time from entry to a study to (i) disease onset, (ii) lost follow-up or (iii) end of study
Used to calculate INCIDENCE RATE which uses person-time as a denominator (new cases/person-time)
(can be person-years)
What does cumulative incidence use as the denominator?
The number of disease free people at start of study
What is incidence rate useful for?
When study participants are followed up for varying lengths of time
How do you calculate incidence rate? (SBA exam Q)
(Number of new cases in a given time period) / (total person-time at risk during that time)
What are the usual headings used when describing the epidemiology of a disease?
- time
- place
- person (age, gender, class, ethnicity)
What is (absolute) risk?
(the number of events) / (the total population at risk over a given time period)
What is the difference between absolute and relative risk? (Exam)
(Absolute) risk: gives a feel for actual numbers involved (has units) –> e.g. 50 deaths / 1000 population
Relative risk: risk in one category relative to another (i.e. no units)
What is attributable risk vs relative risk?
Attributable risk: The rate of disease in the exposed that may be attributed to the exposure (i.e. incidence in exposed minus incidence i unexposed. Also called ‘absolute excess risk’ (it is a type of absolute risk)
Attributable risk = absolute risk one group - absolute risk of other group
Relative risk: ratio of risk of disease in the exposed (e.g. treatment group or smokers…) to the risk in the unexposed (e.g. placebo group) (i.e. incidence in exposed divided by incidence in unexposed)
both are used to COMPARE two groups (e.g. see if smoking carries greater risk than non-smoking to lung cancer)
Relative risk tells us about the STRENGTH of association between a risk factor and a disease
Attributable risk is about the size of effect in absolute terms (i.e. gives a feel for the public health impact - if causality is assumed)
Calculate the attributable risk + relative risk between these two groups:
Incidence of Disease A in smokers, 1/1000 person-years
Incidence of Disease A in non-smokers, 0.05/1000 person-years
Attributable risk: 1/1000 - 0.05/1000 = 0.95/1000 person-years
Relative risk: (1/1000) / (0.05/1000) = 20 (no units)
Calculate the attributable risk + relative risk between these two groups:
Incidence of Disease B in smokers, 8/1000 person-years
Incidence of Disease B in non-smokers, 4/1000 person-years
Attributable risk: 8/1000 - 4/1000 = 4/1000 person-years
Relative risk: (8/1000) / (4/1000) = 2 (no units)
In the previous examples, why was the attributable risk smaller for Disease A compared with Disease B even though the relative risk is much larger?
Because Disease B is more common
How do you calculate absolute risk difference/reduction?
Highest cumulative incidence of one group MINUS the cumulative incidence of the other group (or either way but take away the negative sign)
How do you calculate number needed to treat (to avoid one case of disease X)?
How do you calculate relative risk reduction?
NNT = 1 / (absolute risk reduction)
Relative risk reduction = 1 - relative risk (remember if relative risk is % convert to decimal)
From the following (RCT w/ 5-yr follow-up), calculate the absolute risk reduction, relative risk, number needed to treat, relative risk reduction:
Cumulative incidence of Disease X in people given a new treatment is 6/1000
Cumulative incidence of Disease X in people on placebo is 10/1000
Absolute risk reduction/difference = 4/1000 (over 5 years)
i.e. 1000 people treated and four cases if disease avoided
Relative risk = 0.6
i.e. Incidence in treatment group / Incidence in placebo group
Number needed to treat (to avoid one case of Disease X) = 1 / (absolute risk reduction: 4/1000) = 250
Relative risk reduction = 40% (or 0.4) i.e. relative risk reduced by 0.4 (1-0.6)
Summary of some risk terms
AR (absolute risk) = the number of events (good or bad) in treated or control groups, divided by the number of people in that group.
ARC = the AR of events in the control group.
ART = the AR of events in the treatment group.
ARR (absolute risk reduction) = ARC – ART.
RR (relative risk) = ART / ARC
What is bias?
A systematic deviation from the true estimation of the association between exposure and outcome
What are the 3 types of bias?
- selection bias
- information (measurement) bias
- publication bias - studies with -ve results less likely to be published
Give an example of bias
E.g. comparing mean BP of people in Sheffield and Rotherham -> there may be measurement bias in that two different nurses (dif skills) are measuring BP and a different BP machine is used. Or selection bias
What is selection bias?
A systematic error in:
- the selection of participants
- the allocation of participants to different study groups
What is information (measurement) bias?
A systematic error in measurement of:
- exposure
- outcome
What are the sources of information bias?
- Observer (e.e. observer bias)
- Participant (e.g. recall bias or reporting bias)
- Instrument (e.g. wrongly calibrated instrument)
What is confounding?
The situation where a factor is associated with the exposure of interest + independently influences the outcome (but doesn’t lie on the causal pathway)
Give an example of a cofounding variable in a study
Study looking at drinking coffee and lung cancer link
Smoking is a confounding variable in this
What must you consider when looking at association and causation?
Bias, chance, confounding, “criteria” for causality
What is the Bradford-Hill criteria for causality (i.e. what factors to consider when assessing causality)? - know 4
Dose-response: the higher the exposure, the higher the risk of disease
Consistency: similar results from different researchers using various study designs
Temporality: does exposure precede the outcome?
Reversibility (experiment): removal of exposure reduces risk of disease
Biological plausibility: biological mechanisms explaining the link
others: analogy, coherence, strength of association, specificity
If an association/correlation is not causal, how can it be explained?
(- causal)
- bias
- chance
- confounding
- reverse causality
What is reverse causality
give an example of reverse causality
Situation where where an association between an exposure + outcome is due to the outcome causing the exposure
study examining idea that stress causes HTN: Study shows HTN pts have higher stress levels (consistent with the hypothesis) BUT HTN could have caused ^ stress levels (reverse causality)
What is lead time bias?
lead time is the length of time between detection of the disease and its usual clinical presentation - apparent ^ in survival time
What is length-time bias?
Disease that progress more slowly is more likely to be picked up by screening (i.e. symptom free and around for longer) than more rapidly progressing diseases
What types of bias can be seen in screening?
lead time bias
length time bias
selection bias
What is regression?
Looking at predictors of an outcome (looks at the strength of a relationship between two variables)