Study Design Flashcards
What are the 2 main classes of epidemiological studies? What are the main aims/features of these classes of study?
Descriptive
- determine frequency of disease in particular population to indicate those most at risk
- can use to plan healthcare for particular region
- generates hypothesis for analytical studies
Analytical
- explicit comparison between 2 (+) groups of people over a period of time
- assess for cause and effect
What is an ecological study?
Why is it performed?
What are the advantages + disadvantages of this form of study?
How would you interpret results of a study?
Measure rate of death or disease in populations and the population rate of a risk factor
Use:
- comparison of same population at different time points
- comparison of different population at same time point
Adv:
- quick + inexpensive
- hypothesis generating
Disadv:
- disease and exposure groups might be heterogenous in terms of other baseline characteristics i.e. risk of confounding
- unclear relationship between exposure and disease i.e. unclear timing between the exposure and the disease
- cannot assume relationship at population level will hold at individual level
Interpreting:
-Population with increased/decreased of X exposure/risk factor per capita, tend to have lower/higher proportion of population with X outcome
What is a cross-sectional study? Why is it performed? What analysis can be done? What are the advantages + disadvantages of this form of study? How would you interpret the results?
Information collected from individuals to measure point prevalence of certain disease from a defined population/geographical region in planned way at one point in time (snapshot)
Use:
- determine need for health services in population in particular area
- hypothesis generating for risk factors for disease
Analysis
- T-test= compares the means between the groups
- Chi-squared test= compared proportions between groups i.e. mortality
- Correlation
- prevalence ratio/odds ration= calculate the strength of association
Adv:
-inexpensive + quick
Disadv:
- causality cannot be assessed because cause and effect measured at same time
- range of ages in population means that will have different levels of exposure
- baseline characteristics different between the groups
- difficult to interpret prevalence due to being mixture of incidence and survival
Interpreting:
-Individuals with increase of X exposure/risk factor, tend to have higher frequency of outcome/disease etc
What cannot be commented on in descriptive studies?
Causal relationships between possible exposure and outcome
I.e. need to speak in terms of altered frequency or rate of outcome in association with certain level of exposure
What are the 2 classes of analytical studies? Which study designs are included in each?
Observational
- case-control
- cohort
Interventional/experimental
-RCTs
What is a case-control study? What is done to minimise confounding? How are the results analysed? Why is it performed? What are the advantages + disadvantages of this form of study? How would you interpret the results?
Retrospective comparison of historical exposures between disease and control from a defined underlying cohort such that the controls in the study could be cases if they developed the disease
Confounding:
- Matching= cases and control matched on possible confounding factors (age/sex etc)
- regression model
- unmatched= logistic
- matched= conditional
OR used
- OR used when outcome is rare which is when CC studies are used
- cannot used RR because CC don’t be incidence rates so cannot calculate the risk
Use:
-determining if there are common exposures which are associated with cases more than controls and could be indicated to be risk factors
Adv:
- explore multiple exposures for disease
- no follow up because people recruited based on having the outcome of interest
- can be used in context of rare outcomes over cohort as people can be recruited based on having rare outcome
- useful in disease with long latency period
Disadv:
- only investigate on disease/outcome at time
- lack of certainty about causality due to exposure not guaranteed to come before outcome
- cannot generate AR/RD due to not being able to calculate RR as incidence not possible to assess
- multiple bias:
- recall
- recording
- interviewer
- response
- sampling
Interpreting:
- association between exposure and development of disease based on difference in likelihood of exposure between cases and controls
- expressed as OR i.e. interpreted in terms of likelihood
What is a cohort study? What is done to minimise confounding? How are the results analysed? Why is it performed? What are the advantages + disadvantages of this form of study? How would you interpret the results?
Individuals recruited based on exposure and followed up over time to see if develop outcome
Confounding:
-Matching/stratification/regression curves i.e. adjustments made at analysis stage
Use:
-establish if there is relationship between specific exposure and development of outcome of interest
Adv:
- can estimate the absolute risk of disease
- no recall bias
- increased certainty about causal relationship due to exposure occuring before the outcome
- can look at multiple outcomes
Disadv:
- Expensive + time-consuming due to long follow up
- only able to look at one exposure
- not useful in rare disease as would have to recruit very large population to see rare disease in outcome
- diagnostic techniques might change over time of study
- bias:
- healthy worker effect
- loss to follow up bias
- responder bias
Interpreting:
- RR= number of times increased risk or percentage increase risk of outcome
- AR/RD= how many more people developed the outcome in exposure group compared with unexposed per 100 people
What are the different classifications of exposure? Where can exposure information be gathered from?
Dichotomous
-ever or never
Categorical
-no/low/medium/high
Quantitative
-exposure is continuous variable
Sources:
- occupational records
- questionnaires
- clinical investigations
What are examples of sources of individuals for the comparison/unexposed group?
General population
Another occupational group
Regional studies
Sub-cohort of originally exposed group
What are the different sources of gathering disease data?
National database records
Questionnaires
Clinical examinations
Note:
-would get more expensive and time consuming down the list
When would you want to use a case-control over cohort?
When the outcome is rare
-can recruit people with rare outcome in case-control where as would need to recruit large number of people for rare outcome in cohort and would not be enough to analyse
Long latency period
What is an RCT? What is done to minimise confounding? Why is it performed? What are the advantages + disadvantages of this form of study? How would you interpret the results?
Study investigating effectiveness through participants being randomised to intervention or control
Confounding:
-RCT study design is itself a method of minimising risk of confounders at study designs level
I.e. randomisation process results in baseline characteristics being equally/similarly distributed between groups such that any difference in baseline characteristics is said to be due to chance
Use:
-determine if intervention is effective at reducing outcome
Adv
- randomisation process reduces risk of confounding due to baseline characteristics similar between groups
- allocation concealment reduces risk of allocation bias by ensuring participants have equal chance of entering either arm of study
Disadv:
- bias:
- responder
- selection
- allocation
- performance
- detection
- attrition
- reporting
Results:
-AR= interpret in terms of how many individuals can be prevent from outcome intervention group compared to control per 100 people
I.e. looking at difference between intervention and control
-RR= expressed as increased risk of event
-OR= expressed as likelihood of event happening
-need to consider the precision of the results by looking at 95% CI
What must be present for a RCT to be ethically viable?
Clinical equipoise:
- genuine uncertainty of whether intervention or control is better
- intervention will do more good than harm
I.e. used to ensure that participants will not be intentionally disadvantaged by whether they receive the intervention or not
Incidence and prevalence of disease can be used as measurement outcome.
What is prevalence and incidence?
What is the difference between point prevalence and period prevalence?
Prevalence:
-number of exiting cases of disease in population at given time which indicates how wide spread a disease is (disease burden)
Incidence x duration of disease
Point= prevalence at point in time Period= prevalence during a specific period
Incidence:
-number of new cases of disease in population over period of time which indicates the risk of occurrence of new disease and the risk of developing disease
What can affect the prevalence of disease in population?
Survival rate of people with outcome of interest
Increased prevalence:
- non-curative treatment leads to increased survival time meaning fewer people dying with disease
Decreased prevalence:
-people leave the disease population due to death or being cured
Why is incidence prefered to prevalence?
Incidence useful for indicating the risk of developing disease= can be used to calculate RR or AR
Prevalent cases affected by survival:
- if interested in very severe disease then might have very small prevalence due to people passing away (short survival time)
- might lead to interpretation that disease not as significant due to not having as high a prevalence
- Can lessen strength of association between exposure and disease due to people with higher exposure might develop more severe disease and therefore have reduced survival time and might not be represented in the population
When would you assess for causality? How can you assess for causality in cause and effect studies?
When the following have been excluded as possible explanations for the association:
- bias
- confounders
- chance
Bradford Hill Criteria
Temporality:
-exposure precedes the outcome
Strength
-How big the effect is
Biological gradient/dose response
- Risk of outcome increases with exposure increase
Consistency
- Similar results in different populations and different types of study
Plausibility
- Support by known biological mechanism
Coherence
-Does not contradict existing knowledge or theories
Experiment/reversibility
-Risk of disease decreases after exposure removed
Specificity
-Disease associated with only one exposure
Analogy
-outcome has been seen in the past with similar situation
What is the hierarchy of evidence for cause and effect studies? What study types exist outside the hierarchy of evidence?
From lowest to best evidence
Case series Case-control Cohort RCT SR
Descriptive studies
DTA
Qualitative studies
What are measures of disease frequency?
Incidence= total number of new cases during a specified period within a defined population
Prevalence= total number of individuals with the disease at particular time (old and new cases)
Risk ratio/relative risk (RR)= probability of outcome in exposed group compared to unexposed group
Odds ratio (OR)= approximates the risk ratio i.e. used when the outcome is rare (used in case-controls because risk ratio cannot be calculated)
Attributable risk= excess incidence of outcome attributed to exposure i.e. risk exposed- risk unexposed
What is the process of developing a cohort study?
Decide whether prospective or retrospective more appropriate
Choose exposure cohort
-need to use occupationally/therapeutically defined populations if the outcome is rare
Choose unexposed cohort
-ensure that have similar baseline characteristics i.e. matching to try and reduce risk of confounding
Sources of measurement of exposure
-determine if exposure dichotomous or categorical or continuous
Out ascertainment
- well-defined end point
- record loss to follow up and length of follow up
What is the process of developing case control study?
Define what is meant by case
-measured with prevalence or incidence
Select controls
-selected from same underlying cohort which cases where selected from
Match cases to controls
-match potential confounding factors as part of adjust analysis
Determine bias
- sampling
- response
- recall
Which study designs are appropriate for the following research questions?
- patient or HCP experience
- Frequency or rate of disease
- Aetiology + risk factors= correlation + hypogenerating
- Aetiology + risk factors= hypo testing
- Diagnosis
- Effectiveness
- Qualitative
- Cross-sectional or cohort
- Ecological or cross-sectional
- Case-control or cohort study
- Diagnostic test accuracy (special cross-sectional)
- Randomised controlled trial
