Epidemiology 1-3 Flashcards
(41 cards)
Definition of prevalence
Cross-section
Number of cases/sample at a given time point
Expressed as % or fraction
Incidence risk
Number of new cases / sample of people at risk
in a given time period
Expressed as % or fraction per unit time
aka cumulative incidence aka risk
Utility of prevalence estimates
Idea of risk factors
Allows planning of services
Incidence density
Number of new cases / person-years of observation
Only consider duration person is at risk
More accurate for larger studies with different followup times
Factors affecting disease prevalence
Reporting rate
Length of disease (shorter due to cure or longer due to better treatment)
Incidence
Relationship between incidence and prevalence
Prevalence = incidence * duration of disease
Sources of routine health information
Mortality - death registries
Morbidity - hospital episode data (coded)
morbidity- primary care registries
Incidence - notifiable diseases
Definition of ecological study
Study of aggregates - not individuals!
Units can be e.g. school, region, healthcare facility
Correlation between exposure/disease at aggregate level
What is the ecological fallacy?
Not possible to draw conclusion about individual risk based on aggregate data, as within-unit association may be reerse (e.g. CVD risk and wealth)
Infant mortality rate
Deaths under 1 year old / number of live births
Neonatal mortality rate
Number of deaths under 28d / number of live births
Perinatal mortality rate
Number of stillbirths and deaths <7d / total number of live and still births
Age-standardisation process
Multiply population age structure by age-specific mortality rate in a reference population –> calculate expected deaths for age
Divide observed mortality by expected –> standardised mortality ratio
This is the indirect method
Features of cohort studies
Populations defined on basis of exposure
Exposure measurement precedes disease measurement - free from disease at this point!
Population followed up (regular) over time (may be retrospective)
Assess for changes in outcome between different exposure groups
Can measure multiple outcomes (c.f. case-control)
Features of a case-control study
Define a set of cases with outcome of interest
Define population of controls without outcome of interest
Compare exposures in the past/future between groups
Useful in rare outcomes
Can study multiple exposures
CANNOT CALCULATE RISK - artificially inflated cases
Features of ideal control in case-control study
Drawn from same population as cases
Would be eligible to be cases if they had the outcome of interest
Not overmatched - weakens study
- but matched enough* - e.g. age, sex
2: 1 or more ratio of controls to cases
Similarly accurate measurements to those in cases
Features of a cross-sectional study
Snapshot measure of population - can look at prevalence of outcomes or exposures
Difficult to interpret relationship between outcome and exposure
Not suitable for rare outcomes/exposures (need very large sample)
Difference between odds and risk
Odds is a ratio
Risk is a proportion
Risk can only be obtained from cohort studies that follow patients up over time (ones that give an incidence!)
Odds indicate association, do not interpret magnitude as risk (except if rare disease assumption)
What odds ratios can be calculated from cohort and case control studies?
Case-control: Odds of exposure given disease (a/c divided by b/d)
Cohort: Odds of disease given exposure (a/b divided by c/d
Same mathematically!
What is attributable risk?
Excess risk in a population caused by exposure
Equal to risk in exposed group - risk in unexposed group
Only from cohort studies!
What is attributable fraction
Attributable risk / incidence in exposed individuals
Assumes causal relationship (c.f. ratio measures)
How to calculate population attributable risk
Need to know either true population incidence
OR
population exposure prevalence
PAR = population incedence - unexposed incidence from study
OR
PAR = AR x population prevalence of exposure
What is power?
Probability of a study detecting a true difference if there is one
Usually set at 80%
Precision
Degree of random error in a study
Described by confidence intervals
Reduced by larger sample sizes
