Epidemiology Flashcards
What does a measure of effect do?
It summarises the strength of the relationship between exposure and outcome
Are prevalence ration, risk ratio and odds ratio absolute or relative measures of effect?
Relative
How to sample in a cross-sectional study?
Sample from study population
How to sample in a case-control study?
Sample from cases as well as controls
How to sample in a cohort study?
Sample a cohort of exposed and unexposed
Measures of effect in a cross-sectional study?
Prevalence ratio
Measures of effect in a case-control study?
Odds ratio
Measures of effect in a cohort study?
Risk ratio
Formula for prevalence ratio
What type of study does this apply to?
Prevalence in exposed group/Prevalence in unexposed group
= [a/(a+b)]/[c/(c+d)]
Cross-sectional
Formula for odds ratio
What type of study does this apply to?
Odds in exposed group/Odds in unexposed group
= { [a/(a+b)/[b/(a+b)] } / { [c/(c+d)]/[d/(c+d)] } = ad/bc
Case-control
Formula for risk ratio
What type of study does this apply to?
Risk in exposed group/Risk in unexposed group
= [a/(a+b)]/[c/(c+d)]
Cohort, randomised control trials
What does a relative measure of association do?
Examples?
Determines the strength of a variable’s relationship to an outcome. They give the relative effect of an exposure on disease occurrence.
Odds ratio, risk ratio, rate ratio
What does an absolute measure of association do?
Examples?
Determines how much of the disease can be attributed to the exposure.
Risk difference, rate difference, attributable fraction, number needed to treat
What does a risk ratio measure?
The risk of developing the disease (or event occurring) amongst exposed participants compared to the risk of developing the disease (or event not occurring) amongst the unexposed participants
What does the risk ratio mean?
RR = x, then those who are exposed are x times as likely to develop disease compared to those who are unexposed
RR = 1, there is no difference in risk for exposed and unexposed
RR>1, risk of disease is greater amongst the exposed than non-exposed (harmful exposure)
RR
What does an odds ratio measure?
The risk of having an outcome compared to the risk of not having an outcome. When performing a case-control study you start with the known outcomes and establish exposure. Case control studies can’t calculate incidence or prevalence.
Important points to remember about case-control studies
- Percentage of population with disease can’t be calculated with this study design
- -> Individuals with the disease have been oversampled
- -> Try to get 1:1 ratio of cases:controls for comparison
- -> Therefore, percentage of diseases participants is higher than in population, hence risk overestimating by generalising
- The OR approximates the RR when the outcomes is rare
- ->a Therefore, the lower the prevalence of a disease, the more similar the OR and RR will be in magnitude
What does OR/RR show us?
1 - exposure is harmful
Additional tidbits on OR and RR
- OR and RR will always estimate in the same direction of association e.g. if RR >1, then OR will also be >1
- The OR will always overstate the effect compared to the RR e.g. OR will be smaller when effect is 1
Formula for sensitivity
a/a+c
Formula for specificity
d/b+d
Formula for positive predictive value
a/a+b
Formula for negative predictive value
d/c+d
What is sensitivity?
The true positive rate. This measures the proportion of positive patients who were correctly identified (e.g. disease+test+ )
= true positive/condition positive
Useful for ruling out disease reliably
Low Type II error rate
“of those who were positive, how many were picked up as positive?”
What is specificity?
The true negative rate. This measures the proportion of negative patients who were correctly identified (e.g. disease-test-)
= true negative/condition negative
Useful for ruling in disease reliably
Low Type I error rate
“of those who were negative, how many were picked up as negative?”
What is false positive rate?
1-specificity
Represents Type I error e.g. false positive
What is false negative rate?
1-sensitivity
Represents Type II error e.g. false negatives
Statistical power
1-B
= sensitivity
What is Positive predictive value?
Number of true positives/Number of positive results
“Of those who tested positive, how many truly were positive?”
What is Negative predictive value?
Number of true negatives/Number of negative results
“Of those who tested negative, how many truly were negative?”
What is pre-test probability?
Prevalence of the disease
What is a likelihood ratio?
A tool used to assess the value of performing a diagnostic test. It uses sensitivity and specificity to determine whether a test result usefully changes the probability that a disease exists.
Type I error
1 - specificity
False positive rate
Type II error
1 - sensitivity
False negative rate
What is post-test probability?
Probability of the condition being present after the test
Calculated by multiplying Pre-test probability by likelihood ratio
When test is positive, post-test = PPV
When test is negative, post-test = 1-NPV
Positive likelihood ratio
sensitivity/(1-specificity)
Negative likelihood ratio
(1-sensitivity)/specificity
Definition of an infectious disease
An illness due to a specific infectious agent or its toxic products that arises through transmission of that agent or its products from an infect person, animal or reservoir to a susceptible host, either directly or indirectly through an intermediate plant, host, vector of the inanimate environment
Prevalence =?
Incidence x duration
Incubation period
The time between exposure to an infectious agent and the onset of symptoms/signs of infection
Latent period
Time period from successful infection until the development of infectiousness
Infectious period
The time when infection can be transmitted to another susceptible host
Serial interval
Time period between successive generations of a disease being spread from person to person
Infectivity
The ability of an agent to cause infection in a susceptible host
Measured by: minimum number of infectious particles required to establish infection; proportion of susceptible people who develop infection after exposure
Basic reproductive ratio (R0)
Average number of secondary cases from one primary case. This tells us how fast the infection is spreading
Primary case: individual who brings disease into population
Secondary case: people infection by primary case
Different R0 values
1 then there will be an epidemic
Assumes everyone in population is susceptible
Determinants of R0
P: Probability of transmission in a contact between infected and susceptible individuals
C: Frequency of contacts in the population/number of exposures of susceptible people to infectious partners per unit time
D: How long an infected person is infectious for
R0 = PCD
Determinants of disease outbreaks or epidemics
- Number of people in that population who are susceptible
2. Number of people who are immune
Pathogenicity
The ability of a microbial agent to induce disease
Measured by attack rate
Attack rate
people at risk who get sick/total # at risk
Virulence
The severity of disease after infection occurs
Measured by case fatality ratio/proportion that develop severe disease
Immunogenicity
The ability of an organism to induce specific immunity
Premunition
Host response that protects against against high numbers of parasites (e.g. P. falciparum) without clearing the infection
Herd immunity
Threshold of number of immune people in community at which the likelihood of transmission is small enough to prevent an epidemic
Outbreak investigation
The study of a disease cluster or epidemic in order to control or prevent further spread of disease in a population
What is a cause?
An event, condition, characteristic or combination of these factors which plays an important role in producing the disease
Causality is the relationship between an event (cause) and a second event (effect), where the second event in understood as a consequence of the first
Sufficient cause
The set of minimal conditions and events that inevitably produce disease in at least some people
== complete causal mechanism
Component cause
One cause amongst a constellation of causes that make up a sufficient cause together. Factors that work together with the necessary cause to produce disease
Necessary cause
A component cause which is a member of every sufficient cause for a given disease. This MUST be present for a disease to occur
Necessary AND sufficient cause
= Only one component cause e.g. rabies virus
Most causes of non-communicable diseases are neither necessary nor sufficient
3-step process in determining causation
- Is there an association? (measures of association)
- Is it a true association? (validity - exclude other reasons: chance, bias, confounding)
- Is the observed association likely to be a causal one?
Confounder
A third variable which confuses our assessment of the true association between exposure and outcome
How to address confounding
- Study design:
- -> Randomization
- -> Matching
- -> Restriction
- Data analysis:
- -> Measure confounders
- -> Adjust for them through stratification or multivariate regression
Strength of association
A small association doesn’t mean that there isn’t a causal effect, although the larger the association, the more likely that it is causal
Consistency
Consistent findings observed by different persons in different places with different samples strengthens the likelihood of an effect
Temporality
The effect has to occur after the cause!
Bradford-Hill criteria
1 Strength of association 2 Biological plausibility 3 Consistency (reproducibility) 4 Dose-response relationship 5 Coherence (between epidemiological and lab findings) 6 Specificity 7 Temporality 8 Experiment 9 Analogy
Strengths of case-control study
Cheap Easy to manage/arrange Quick Useful for studying rare diseases Useful as a preliminary study when little is known about the relationship between a risk factor and a disease
Weaknesses of case-control study
Observational - don’t have the same level of evidence as RCT
Prone to confounding
May be difficult to establish timeline of exposure to disease outcome
Strengths of cross sectional study
Use of routinely available data - cheap
Quick
Useful preliminary study to identify factors and populations for further investigation with better method
Weaknesses of cross sectional study
Routine data may not be designed to answer the set question
Data may not describe which variable is case and which is effect
Prone to confounding
Strengths of RCTs
Reliable form of evidence
Can correct for bias, confounding, chance
Weaknesses of RCTs
Limited external validity i.e. findings may not be generalisable
High time and cost
Potential for conflict of interest
Potential for ethical breaches
Strengths of cohort studies
Can be used to identify causal relationships between variables
Can help identify risk factors for developing a disease
Recall error is reduced
Reliable data is produced - rank high in hierarchy of evidence
Can measure large variety of exposures
Weaknesses of cohort studies
Expensive to run
Requires many years
Participants are prone to loss-to-follow-up
What is recall bias?
Systematic error caused by differences in the accuracy or completeness of recollections retrieved by study participants regarding past events
Snowball sampling
Existing study subjects recruit future subjects from their acquaintances
Purposive sampling
Have certain criteria
Theoretical sampling
Using emerging theory in project to drive selection of new participants
Triangulation
Facts are supported by more than a single source of evidence. If you have used multiple sources but not triangulated the data, then you have analysed each source of evidence separately
Methodological triangulation
Different data collection techniques
Investigative triangulation
Different researchers
Data triangulation
Different data sources
Theory triangulation
Different theories
Why run a pilot study?
Is question measuring what it is intended to measure?
Is wording understood?
How do respondents feel about/react to questionnaire?
Time taken
Threats to validity of questionnaire
Cognitive factors: comprehension, literacy, recall bias
Situation and setting in which the measurement takes place: concerns over social desirability, non-participation, social norms, interviewer, privacy etc
Factors related to study design: detection bias (prospective study), bias in measuring exposure (case control), period effects (RCT and prospective)
Face validity
The relevance is obvious
Content validity: all the relevant elements are included
Consensual validity: experts agree a measurement is valid
Some degree of face validity is required, but it can be deceptive, and where possible needs to be supported by criterion validity
Criterion validity
Correlation between the measure and another variable, which is suitable for use as a criterion of validity e.g. scale measuring level physical activity can be validated by test of fitness
Ideal criterion is the true value
In real life, we use a measure which has been tested before and shown to have high criterion validity
Definition of epidemiology
The branch of medicine that deals with the incidence, distribution, and possible control of diseases and other factors relating to health
Definition of demography
The study of statistics such as births, deaths, income, or the incidence of disease, which illustrate the changing structure of human populations
Environmental burden of disease
Number of deaths and DALYs that can be attributed to environmental factors
How to calculate DALY/environmental burden of disease
DALY = # of people with disease x duration of disease (or loss of life expectancy) x severity (0=perfect health, 1=death)
