Public Health Flashcards
What are the different kinds of research questions?
Descriptive and analytical
What are the characteristics of non-longitudinal studies?
No follow up of subjects. Includes ecological, cross-sectional and case control studies
What are characteristics of longitudinal studies?
Follow up of subjects, information progressively collected over a period of time
What is prevalence?
Number of existing cases of an outcome in a defined population at one point in time. Expressed as proportion or percentage
What is incidence?
Number of new cases of an outcome in a defined population during a time interval. Expressed as a rate, and can only be determined from a longitudinal study
What is risk?
Probability of disease occurring during a time period.
risk=n/P per time
What is rate?
Probability of disease occurring during sum of individual follow-up periods
rate=n/(total person-time of follow-up)
Considers how long each person was in the study
What is hazard?
An instantaneous rate derived from longitudinal studies
What is relative risk/rate?
Relative magnitude of change in risk/rate of outcome, associated with exposure
RR=R(e)/R(u)
What is attributable risk/rate?
Absolute magnitude of change in risk/rate of outcome, associated with exposure
AR=R(e)-R(u)
What is the attributable risk/rate percent?
Proportion of outcome among those exposed, that is related to the exposure and not some other factor
AR as a proportion of those exposed
AR%=AR/R(e)100%=[R(e)-R(u)]/R(e)100%
What is the population attributable risk?
The additional risk/rate of the outcome in the whole population that is due to the exposure
PAR=R(t)-R(u)
What is the population attributable risk percent?
The proportion of the outcome in the whole population that is due to the exposure. Also known as preventable fraction
PAR%=PAR/R(t)100%=[R(t)-R(u)]/R(t)100%
What are the Bradford Hill Criteria for Causality?
Temporal relationship, Strength, Dose-response relationship, Consistency, Plausibility, Exclude alternative, Experimental evidence, Specificity, Coherence
Describe cross-sectional studies
Take a cross section of the population and obtain information from them.
Tends to be descriptive
Non-longitudinal
Relatively cheap & easy
Weak evidence of causality
Describe case control studies
Recruit subjects with an outcome, and then match them with controls without that outcome by confounders. Compare previous exposures between these subjects.
Useful for studying rare outcomes
Non-longitudinal
Can calculate odds ratio
What is an odds ratio?
An approximation of relative risk
OR=[outcome(exposure):outcome(non-exposure)]/[no outcome(exposure):no outcome(non-exposure)]
Describe cohort studies
Follows a population without an outcome, to determine who then develops that outcome
Restrospective cohort studies only begin following subjects after an outcome, and look back to consider the exposure
Longitudinal - can collect incidence
Can consider multiple outcomes and exposures
Difficult to study rare outsomes
Expensive
Describe selection bias
A systematic difference in the people recruited to participate in a study compared to the general population, meaning that the observed result may not reflect the true situation
A systematic difference between the groups being compared. Can be due to the way groups are assigned, due to drop out
How can selection bias be reduced?
Maximise response
Minimise lost to follow-up
Intention-to-treat analysis
Randomisation of study groups
Describe information bias
Systematic difference in the way information is collected between compared study groups
How can information bias be reduced?
Make sure collection methods are uniform
Objective assessment
Standardised tools
Blinding
What is a confounder?
A factor that mediates the relationship between an exposure and an outcome of interest, such that the association between the two is distorted. This confounding factor independently predicts the outcome, and is not in the causal pathway between the two.
Describe clinical trials
Longitudinal studies to determine if an intervention vs placebo has an effect on an outcome.
How can confounding be reduced?
Randomisation of study groups
How can randomisation be used in a clinical trial?
All subjects are randomly placed into treatment groups so that every treatment group is made up of the same kind of people. This should distribute confounders evenly into all groups, even for confounders that we are unaware of
The effectiveness of randomisation can be assessed by profiling the treatment groups and considering their similarity
How can blinding be used in a clinical trial?
Ensure that neither subjects nor investigators are aware of which treatment subjects are receiving. If there are independent assessors of outcomes, they should also be blinded.
How can cross over in clinical trials lead to selection bias?
Subjects who cross over tend to be systematically similar, and thus it is a source of self selection
How does intention-to-treat analysis reduce function?
FIND SOME KIND OF EXPLANATION
What is hazard/hazard rate?
An instantaneous rate derived from a longitudinal study with close follow up.
What is hazard ratio?
A measurement similar to risk ratio that applies to the whole period of follow-up.
What is the number needed to treat?
Number of people needed to undergo the treatment in order to prevent the outcome in just one of those people. It is a marker for the efficiency of the intervention. It is affected by RR as well as the underlying likelihood of the outcome.
NNT=1/(AR)
What is the purpose of PICOT?
PICOT is used to frame a research question
What does PICOT stand for?
Population, Intervention, Comparator, Outcome, Timing
What is internal validity?
The extent to which the results of a study are valid for the patients who were studied
What does internal validity depend upon?
Appropriate study design, data collection and data analyses. Whether the study appropriately controlled for bias and confounding
Consider: Randomisation Blinding Objective Outcome Assessment Intention-to-treat Analysis
What is stratified randomisation?
A way to ensure that key confounders are spread across treatment groups evenly, and so further reduce the potential for confounding effects
What does the p-value indicate?
The probability that the results observed arose by chance
What is the confidence interval?
The interval in which there is a 95% confidence of the true value lying
The width of the CI also indicates precision of the result.
What p-value and confidence intervals are considered significant?
p<0.05
CI not spanning the null value
What is the null value?
The value that indicates no differences between the groups being compared.
When the comparison is a ratio, the null value is 1. When the comparison is an absolute difference, the null value is 0
What is external validity?
Whether an internally valid study is abled to be applied to a certain patient or situation
How does one consider whether a study is externally valid?
Assess similarity via PICOT
