Key principles of epidemiology Flashcards
define frequency
number of occurrences of an outcome within a given time period
two main approaches to epidemiological study
descriptive and analytical
what information does descriptive epidemiology provide?
distribution of health outcomes by age, population type, geography or over time.
what information does analytical epidemiology provide?
investigates which factors may be responsible for increasing or decreasing the probability of an outcome
difference between sufficient and component cause
sufficient cause refers to a factor or sort of factors that produce the outcome. the factors that form a sufficient cause are called component causes
for an association to be causal, the …. must occur before the ….
exposure then outcome
what can effect causality? (alternative explanation for an association)
chance, bias and confounding
what is chance and how do you reduce it?
possibility that there is a random error and reduced by increasing sample size, using random selection or randomisation
what is bias
refers to systematic differences between comparison groups which may misrepresent the association being investigated
what is confounding
caused when another factor, independently associated with both the outcome and exposure, influences the association being investigated
two approaches to analytical epidemiology
observation or intervention
difference between observational and intervention study
observational compares the frequency of an outcome in groups or individuals with and without the exposure of interest. invervention is an experiment - evaluates the effect of reducing a risk factor or increasing a protective factor on the frequency of an outcome.
define prevalence
the number of existing cases in a defined population at a defined point in time divided by the total number of people in that population at the same point in time - it is a proportion
how do you measure prevalence?
population surveys or cross sectional studies
define incidence
frequency of new cases in a defined population during a specified time period
how do you measure incidence?
ecological or cohort studies
three ways to consider incidence
risk, odds and incidence rate
define risk
total number of new cases in a specified time period divided by the total number of individuals at risk in the population at the start of the time period
how do you work out secondary attack rate
number of new cases among contacts in a specified time period divided by total number of contacts of a primary case in that time period
define odds
ratio of two proportions so number of new cases in a specified time period divided by number who did not become a case during that time period
what do you use if it is an open population rather than a closed population to work out incidence rate?
population time at risk
examples of relative measurers that use ratios
prevalence ratio, risk ratio, odds ratio, incidence rate ratio
what does relative risk measure
how much more likely it is that an exposed individual will develop the outcome compared with the unexposed individual
what studies do you use to calculate risk ratio?
ecological, cohort or intervention
how do you calculate odds ratio?
odds of outcome in exposed group divided by odds of outcome in unexposed group
what studies do you use to obtain incidence rate ratio?
ecological, cohort or intervention
when would you use incidence rate ratio over the odds ratio of exposure?
where people are entering and leaving the study population or have changing levels of exposure
if the outcome is rare, are the measures of association more similar or different?
more similar
when would you use attributable risk?
when you want to know the excess incidence of the outcome that we an attribute to the exposure (if we assume a causal relationship)
how do you work out attributable fraction?
attributable risk divided by incidence in exposed OR relative risk - 1 divided by relative risk
what is the opposite of attributable fraction? (i.e. if there’s a protective factor)
preventable fraction
how to work out the population attributable fraction?
the incidence in the population minus the incidence in unexposed divided by the incidence in population
define p value
the probability that an observed value or association from a sample occurred by chance alone and that it does not exist in the population from which the sample was selected
define confidence interval
the range of values, estimated from a sample, within which the true population value is likely to be found
if confidence interval includes 1 what can we say about the association?
it is not significant
how do you reduce bias?
through random selection of study participants or random allocation of individuals to comparison groups
different types of bias
selection
information
when does selection bias occur
when there is a systematic difference between the characteristics of individuals samples and the population from which the sample is taken or a systematic difference between comparison groups
what factors affect whether selection bias occurs?
the definition of the study population or comparison group
the inclusion and exclusion criteria
the rate of loss to follow up
healthy worker
when does information bias occur
when there is a systematic difference between comparison groups in the way that data is collected
how can information bias be introduced?
by those measuring the outcome (observer bias), the study participants (responder bias) or measuring tools (measurement bias) or misclassification
when does non differential misclassification occur?
occurs when both comparison groups are equally likely to be misclassified - unable to measure effectively. causes comparison groups to appear more similarly than they actually are and may lead to underestimation of the strength of association. independent of exposure or outcome
when does differential misclassification occur?
when classification of the exposure is dependent on the outcome or vice versa - due to observer or respondent bias
how do you reduce the effect of observer bias?
blinding, objective vs subjective measures
three ways to avoid confounding
randomisation - of individuals to exposure and control groups
restriction - limits the study to people who are similar in relation to the confounder
matching - selects two comparison groups to have the same distribution of potential confounders
how to control for confounding in the analysis
stratification - extension of frequency matching as it measures the association between exposure and outcome separately for each category of confounder
statistical modelling - allows us to adjust simultaneously for several confounders using multivariable regression analyses.
bradford hill criteria for building up evidence for a causal relationship
strength - the stronger the association between the exposure and outcome, the less likely the relationship is due to another factor
consistency - repeatability of the result
temporality - exposure comes before the outcome
dose-response - increased risk of outcome with increased exposure
plausibility - existence of a reasonable biological mechanism
reversibility - whether an intervention to remove the exposure results in the elimination of the outcome
coherence - consistency with other information
analogy - similarity between other established cause-effect relationships helps support the argument
specificity - the relationship is specific to the outcome of interest
ecological studies: what do they do
compare frequency of outcome an exposure at a population level
cross sectional studies: what do they do
compare prevalence of outcome with exposure status at one time point from a random sample of individuals
cohort studies: what do they do
compare the incidence of outcome in individuals with recorded differences in exposure
case control: what do they do
select individuals on the basis of their outcome status and analyse whether they differ in relation to previous exposure
intervention studies: what do they do
allocate a protective exposure and compare outcome between those exposed and unexposed
Ecological study: what measures do you use? advantages and disadvantages?
prevalent or incident cases
easy to collect data, rapid, inexpensive
cannot show causality, high probability of confounding, ecological fallacy
cross sectional studies: what measures do you use? advantages and disadvantages?
prevalent cases, prevalence ratio
easy to collect data, rapid
difficult to know if exposure preceded outcome, selection bias, confounding
cohort studies: what measures do you use? advantages and disadvantages?
incident cases - risk ratio, odds ratio or rate ratio
low probability of selection bias, recall bias and confounding
loss to follow up, more expensive, more difficult as large number of participants required
case control: what measures do you use? advantages and disadvantages?
no measures of outcome frequency, odds ratio of exposure
low probability of loss to follow up
high probability of selection and information bias, confounding
intervention: what measures do you use? advantages and disadvantages?
prevalent or incident cases, preventable fraction
low probability of selection and information bias and confounding, satisfy temporality consideration, satisfies reversibility, can radnomize subjects to reduce effects of chance, reduce information bias
loss to follow up, more complex, expensive
difference between statistical power and precision
statistical power = probability of detecting an effect if it is real
statistical precision = probability of detecting an effect if it is not real
different types of random sampling
simple random sampling
systematic sampling - regular intervals
stratified sampling - equal proportion from each group
multi stage sampling - sampling frame required for high level unit and for primary sampling units within higher levels sampled
cluster sampling - also uses organisational structure but samples all or most of the primary sampling units
if you need to exclude certain individuals what do you use when sampling?
exclusion and inclusion criteria
what is informed consent?
participants need to give their consent to participate
medical records, census data, health surveys, schools records are examples of what type of data collection method?
indirect data collection methods
questionnaires, structured interviews and clinical examination are examples of what type of data collection method?
direct data collection methods
why do we use unique identifiers?
so we can link all necessary data without identifying participant
three types of quantitative variables
binomial - two values
categorical
continuous
how do you usually present a meta analysis
forest plot
what does a meta analysis do?
weights the individual estimates in relation to the sample size of each study
when do we use direct standardisation?
in ecological studies to reduce confounding when we know the group specific outcome rates in the study population. the age specific rate is the incidence rates calculated separately for each group - can then add them up to give a standardised rate
when do we use indirect standardisation?
calculates the expected number of cases in the study population if the age specific rates of a standard population had applied. used if we don’t know the age specific rates. the expected number of cases can then be used to calculate a standardised mortality ratio which compares the observed number of deaths to that expected in the same population.
what is ecological fallacy?
in ecological studies when a mismatch arises from trying to draw conclusions about individual level epidemiological associations from a group level study
what is mixing in ecological study interpretation?
geographical comparisons may suffer from migration of populations between groups
if 95% confidence interval does not include 1, what can you say about the p value?
p value will be less than 5 (less than 5% chance that this association is not true)
difference between point and period prevalence
point prevalence is at a specific point in time and period prevalence is over a period of time. period prevalence increases possibility of recall bias
what are cross sectional studies at high risk of when assessing causation?
reverse causality
are analytical cohort studies retrospective or prospective?
can be either as long as cohort defined on exposure and not outcome
in a cohort study, if follow up times are similar what frequency measure do we use? if follow up times are different what do we use?
risk if similar and rate if follow up times differ
what is the issue with case control studies and bias?
susceptible to information and selection bias
how can you reduce confounding in case control studies?
identify controls with the same characteristics as the cases - matching
what are nested case control studies?
cases are members of a cohort that have developed the outcome and control are those without the outcome. can automatically match on factors that are common to all cohort members
in a case control study what do you need to ensure when sampling the population?
that they are representative of the target population in relation to the frequency of exposure, selection of cases and control are not influenced by their exposure
why might inclusion of prevalent cases in a case control study be an issue?
inclusion of prevalent cases, especially chronic outcomes, can create problems for determining exposure to certain risk factors that may change over time which may lead to reverse causality. may also lead to underrepresentation of severe cases
why can’t case control studies directly estimate prevalence or incidence of the outcome or frequency of the exposure in the general population?
because they don’t randomly sample the population, instead they select individuals based on their outcome status
what is the ideal intervention study?
randomised controlled trial
difference between efficacy and effectiveness?
efficacy = measures the effect of the intervention under "experimental" conditions, where maximum effort is put into intervention delivery. once proven to be efficacious... effectiveness = measures the effect of the intervention of the intervention under routine conditions.
what is a plausibility study?
evaluates the effectiveness of interventions by comparison with a historical, geographical or opportunistic control group but without randomisation.
when might you use a plausibility study?
when an intervention is so complex that RCT results will be unacceptably artificial.
when an intervention is known to be efficacious or effective in small scale studies but effectiveness on a large scale must be demonstrated
when ethical concerns prevent the use of an RCT
what is a plausibility study with a historical control group?
not possible to have a contemporary comparison arm so it may compare outcome frequency before and after the intervention is introduced. makes it difficult to distinguish between the effect of the intervention and any other changes that may have affected the outcome over the study period.
what is a plausibility study with a geographical control group?
the use of communities or individuals as contemporary control outside the trial area may adjust for the effects of temporal changes during the course of the evaluation. however, there are likely to be differences between intervention and control. can use stepped wedge design (by introducing into population in phases)
what is a plausibility study with a opportunistic control group?
involves the use of individuals or communities that should have reached the intervention but did not because the programme was unable to reach them.
define RCT
the existence of a contemporary comparison group of study subjects who do not receive the intervention known as the “control” and the random allocation of study subjects to the intervention and control arms
how is a cluster randomised controlled trial different?
groups of individuals are randomly allocated to the intervention and control arms
what happens in a factorial trial?
two or more interventions are compared individually and in combination against a control comparison group
what happens in a crossover design RCT?
each trial subject acts as its own control by receiving either the intervention or control at different points in the study, with a washout period to avoid contamination between the study periods
different types of random allocation
simple randomisation systemic randomisation blocked randomisation stratified randomisation matched pair randomisation
what is simple randomisation?
uses number tables or a computer generated random number list
what is systematic randomisation?
allocates participation to each group alternately but subject to selection bias
what is blocked randomisation?
restricts the allocation list to ensure equal numbers in all study arms.
what is stratified randomisation?
divides participants into subgroups or strata based on key risk factors and equal numbers of subjects from each stratum are randomly allocated to each study arm. this ensures that suspected confounders or effect modifiers of the association between intervention and outcome are equally distributed between comparison groups
what is matched pair randomisation?
form of stratified random allocation that matches individuals or communities into pairs with similar baseline risks of the outcome
what is the primary analysis of an intervention study?
intention to treat
what is the efficacy of an intervention?
the proportion of cases that can be prevented by the intervention - also known as protective efficacy or preventable fraction and calculated by doing 1 - relative risk
what is primary prevention and example?
aims to stop an outcome developing by preventing or reducing exposure to a risk factor
vaccination against an infectious disease
what is secondary prevention and example?
aims to interrupt progression from early to mid-stage of the outcome by early detection and prompt treatment. e.g. blood lead screening in USA
what is tertiary prevention and example?
aims to reduce complications or severity once outcome is established and symptomatic by offering appropriate treatment and intervention.
e.g. by monitoring blood glucose of those with diabetes to prevent renal disease and glaucoma
what is screening used for?
to identify small numbers of individuals at high risk of an outcome
how do you work out sensitivity?
the proportion of people who have the disease who are correctly identified.
how do you work out specificity?
the proportion of people who do not have the condition that are correctly identified.
what is the positive predictive value?
the likelihood of an outcome based on the result
what is the negative predictive value?
the likelihood of no outcome based on a result
disadvantages of screening
false positives and false negatives potential harm (X-ray in mammography) good quality control needed specialist equipment may be required staff may require additional training
WHO criteria for screening
the condition should be an important health problem
natural history of the condition should be well understood
there should be a recognisable latent or early stage
there should be a suitable method for detection
screening method should be acceptable to population
should be an accepted treatment for those with the condition
should be an agreed policy on who to treat
facilities for diagnosis and treatment should be available
costs of detection should be balanced inc elation to overall healthcare spending
screening should be ongoing and not one off
what is lead time bias?
screening identifies an outcome earlier than it otherwise would have been identified but has no effect on outcome (so looks like prolonging)
what is length time bias?
outcomes that take longer to develop to a stage where they threaten health - more likely to pick up less aggressive tumours with better prognosis leading to over-estimate of screening success.
what is the prevention paradox?
where the majority of cases of a disease come from a population at low or moderate risk of that disease, and only a minority of cases come from the high risk population (of the same disease - target whole population
what is a case definition?
set of standard criteria for determining whether or not a person has a particular disease or health related event
what effect does an emerging new treatment that increases the life expectancy of HIV cases have on prevalence/incidence?
increased prevalence
what effect does the discovery of a cure for DM have on prevalence/incidence?
decreased prevalence
what effect does the introduction of a new malaria vaccine have on prevalence/incidence?
decreased incidence
different types of bias
information selection recall sampling non response
difference between descriptive and analytical cross sectional studies
descriptive - collect info on frequency and distribution of health related exposures or outcomes in a defined population
analytical - investigate the association between exposure to risk factors and outcome of interest
examples of when selection bias can occur in cross sectional, cohort, case control and intervention studies
cross sectional: non response
cohort: healthy worker effect, loss to follow up
case control: selection of cases and controls
intervention: systematic selection for intervention/controls
what is validity?
the degree to which a measurement measures what it purports to measure
what is reliability?
the degree to which the results obtained by a measurement procedure can be replicated