click below
click below
Normal Size Small Size show me how
WEEK 2:
An epidemiologist's toolkit:
| Question | Answer |
|---|---|
| observational studies | investigator passively observes as nature takes its course |
| interventional studies | investigator actively manipulates study participants |
| types of observational studies (4) | ecological, cross-sectional, case control, and cohort studies |
| example of interventional study | RCTs |
| interventional studies are used to | assess the safety, tolerability and effectiveness of interventions based on experiments on many individuals |
| observational studies are used to | show patterns of distribution of disease and factors which may be associated with disease in whole population and to collect data to make association between a disease and an exposure |
| when can we not use RCTs | ethnicity (as researcher cant randomly allocate it), smoking (its unethical), and questions that dont involve testing a hypothesis |
| order of hierarchy of evidence (top to bottom) based on quality of evidence | systematic reviews and meta analysis, RCTs cohort studies, case control studies, cross sectional surveys, and ecological studies |
| process of hypothesis testing | Set up the hypothesis (association). State the null hypothesis (no association) Sometimes referred to as H0 Summarise the data Interpret probability of no association |
| probability of 0 | results could not occur if H0 is true |
| probability of 1 | results exactly consistent with H0 |
| p value (in epidemiology) | what is the probability we would see results this extreme if null hypothesis was true |
| if the probability is less than 1 in 20 (0.05) | it is unlikely that the null hypothesis is true |
| if the probability is more than 1 in 20 (0.05) | it is likely that the null hypothesis is true |
| if p<0.05 | data shows statistically significant evidence against null hypothesis (H0 is unlikely to be true and we can reject null hypothesis) |
| H0 meaning | null hypothesis |
| p>0.05 | this does not prove that H0 is true but merely demonstrates that the observed findings are consistent with H0 being true |
| statistical significance is influenced by | sample size where very small differences may be statistically significant if sample large enough |
| statistical significance tells us | that the results are unlikely to be chance |
| cause and effect relationship | something happens that makes something else happen |
| epidemiology is based on assumptions include (2) | disease does not occur at random and that casual factors can be identified through systematic investigation |
| Koch's postulates includes (3) | necessary cause, sufficient cause and specific cause |
| necessary cause | disease cannot occur without cause being present |
| sufficient cause | cause on its own will always lead to disease |
| specific cause | cause is absent in other diseases/ agent must not be found in cases of any other disease |
| associations may be a result of | chance, bias and confounding |
| p value tells you | how likely it is that the results of a study would have occurred by chance if the null hypothesis was true |
| false positive findings | p<0.05 (means we accept 1 in 20 chance of concluding something is statistically significant when it isnt) |
| bias | any systematic error in a study resulting in an incorrect estimate of the association between exposure and risk of disease |
| selection bias | any error that arises in the process of identifying the study population |
| observation (information) bias | any systematic error in the measurement of information on the exposures or outcomes |
| types of observation (information) bias (4) | recall bias, interviewer bias, attrition bias, and misclassification bias |
| confounding bias | a mixing of effects between the exposure, the outcome and a third factor that is associated with the exposure and independently affects the risk of developing the disease |
| reverse causality | when you think X causes Y but really Y is causing X (other way) |
| simultaneity | X causes Y and Y causes X (both ways) |
| Bradford Hill's criteria is used for | inferring causality (whether an exposure/ outcome relationship is casual) but is usually applied when there is an absence of data from RCTs |
| Bradford Hill's criteria includes (8) | strength of association, consistency, specificity, temporality, dose response relationship (biological gradient), plausibility, coherence, and reversibility (experiment) |
| describe strength association as part of Bradford Hill's criteria | strong associations more likely to be casual BUT weak associations are NOT necessarily unlikely to be casual |
| describe coherence as part of Bradford Hill's criteria | does association conform with current knowledge and theory eg lab findings |
| describe specificity as part of Bradford Hill's criteria | many factors can cause multiple diseases |
| describe temporality as part of Bradford Hill's criteria | does the exposure precede (come before) the outcome |
| describe dose response relationship (biological gradient) as part of Bradford Hill's criteria | as exposure increases, does the risk of disease also increase |
| describe reversibility (experiment) as part of Bradford Hill's criteria | does removal of the risk factor reduce or remove the risk of developing the outcome |