Chapter 7 Flashcards
observational research
naturally occurring variation observed to identify patterns and associations
natural history od disease
type of observational study
case series ecological surveys and cross sectional case control cohort qualitative
experimental research
investigator controls 1 factor and measures variation in outcome
clinical trials
impact of preventative vaccines or behaviour change programs
type of experimental research
RCT
cluster RCT
cross over trials
factorial trials
modelling and stimulation research
development of theories
validity dependant on data used to set up model parameters
study disease transmission and empirical data
hierarchy of evidence
Systematic review and meta-analysis of trials. • RCTs. • Cohort studies. • Case–control studies. • Ecological studies. • Cross-sectional studies. • Case reports and case series. • Expert opinion.
exposure
describe something that might affect an outcome
explanatory or independent variables
exposure of interest -= one in the hypothesis
outcome
response or dependant variables
can have more than 1 outcome per study
confoiunder
independantkly associated with the exposure and outcome of the study
can lead to bias
explanatory or independent variables
bias
deviation from the truth that occurs in studies
systematic error is different from random error - increasing the sample size can reduce random error but not bias
selection bias
recruit participants based on characteristic linked to exposure and outcome
occurs in allocation to intervention
non-responders different to responders - healthy participant effect
measurement bias
systematic errors in measurement including errors in allocation to different groups
systematically wrong - recall bias in case control
association
statistical dependence between 2 variables - indicates the degress to which the outcome is different in those with(out) exposure
chance
inference from samples
repeat in different sample and results would be different
use CI to determine likelihood of chance being a factor
observer bias
RCT
researxher aware of the treatment that the person is getting
information bias
difference in the way that the information is collected and so different qualities between groups of the study
non-differential misclassification
chance of misclassification is the same regardless of the disease status/exposure
random misclassification
bias association towards null - masks true differences
differential misclassification
probability of exposure being misclassified depends on the disease status/vice versa
can bias estimates in either direction
confounding
false association because it distorts the observed association
randomisation
adjust for confounding at the design stage
distribution of confounding factors should be the same in both groups
matching
case control
select cases and controls so match on confounding factors
adjusting for confounding at analysis stage
stratification
standardisation
multiple regression
association/causation
exclude chance bias and confounding = true association - doesn’t necessarily mean causal
Bradford hill criteria
strength of association consistency of association specificity of association temporal sequence of association dose response relationship biological plausibility of association coherence reversibility analogy
strength
measured by RR
stronger = more likely causal
consistency
repeated demonstration in different populations and study designs
specificity
1:1 relationship between cause and outcome
temporal
risk factor before outcome
dose response
gradient of risk
plausibility
known biological mechanism
coherence
absence of conflict with other knowledge about the natural history and disease
reversibility
remove the risk factor prevent the outcome
analogy
analogy with other similar causal associations
causality and public health
absolute proof rarely attainable in empirical sciences
how important are causal relationships
knowledge of causal mechanism not essential for effective preventative strategies