Week 2 Revision Flashcards
Pros and Cons of observational studies
Pros
• Often cheaper and less resource-intensive
• Long term (e.g. 20yr) follow up possible; may
not need to wait for follow up to occur (e.g. case
control studies)
• Can be very large; more representative
• Can include vulnerable/high risk individuals
Cons
• Exposed & unexposed differ in many ways
• Possibility of selection bias
• Not as good for causal inference (temporality,
bias, confounding)
What are descriptive studies?
Designed only to describe the distribution of an exposure/
outcome
e.g., in terms of person, place and time: who, where, and
when
Pros and Cons of Descriptive Studies
Pros- Can describe patterns of disease or exposure
• Can help generate hypotheses
• Can help determine extent of problem/public
health importance
• Can be quickand relatively cheap to conduct
Cons-Generally, cross-sectional
• Can test limited range of hypotheses (e.g., not
‘does X cause Y?’)
What are analytic studies?
Can test hypotheses
Investigate potential determinants (i.e., risk and protective factors)
of a disease/outcome
What are the Big four study designs?
Ecological
▪ Cross-sectional
▪ Cohort
▪ Case-control
What are the common characteristics of ecological studies?
Unit of study = groups, not individuals
Aggregate/summary data on exposure and outcome
for whole population
Can compare groups from different geographic
regions, or at different points in time
Pros and Cons of Ecological Studies
Pros-Often quick and cheap, using routine data
• Generate hypotheses for further investigation of
exposure-outcome relationships
• Useful first step if individual-level data not
available
• Reasonably stable estimates, due to the large
numbers involved
Cons- Outcome status of exposed vs. non-exposed
individuals within the groups unknown
• Often no info on potential confounders
• Unclear if ‘exposure’ caused the ‘outcome’, or
vice versa
• ‘Ecological fallacy‘: drawing (incorrect)
conclusions about individuals based on grouplevel data
What are cross-sectional studies?
Information collected at one point in time
(a ‘snapshot’)
Main outcome measure is prevalence
Can be descriptive or analytic (compare prevalence in
exposed vs. unexposed)
Pros and Cons of Cross-Sectional Studies
Pros-Quick and cheap
• Useful in identifying health needs of a
population at certain time
• Can examine (cross-sectional) associations
between exposures and outcomes
• Generate hypotheses
Cons- Temporality • Weak evidence for causality Considerations: • Generalisability; selection bias; potential confounders
What is the difference between cohort and case-control studies?
Cohort studies:
• select based on exposure
• look forwards from exposure to outcome
Case-control studies:
• select based on outcome status
• look backwards from outcome to exposure
Which cohort study is this?
Sample is selected on basis of exposure (i.e., exposed vs. non-exposed), then followed up
to estimate risk of outcome
e.g. UK Gulf War Study: compared later mental health outcomes between three exposure
groups: those deployed in Persian Gulf War, those deployed in Bosnia, and those serving at
same point in time but not deployed (Unwin et al., 1999)
+ve: temporality (exposure measured before onset of disease); good for rare exposures.
-ve: costly and time-consuming (especially for diseases with long latency period); poor
choice for rare diseases; restricted to single exposure
Prospective (Classical) Cohort Study
Which cohort study is this ?
Exposure(s) of interest were previously measured (for other purposes), can then trace
these individuals with complete data to see if developed outcome of interest
e.g. using data linkage to see whether victims of childhood sexual abuse (information
collected at the time of incident) were associated with later contact with mental health
services for psychosis (Cutajar et al., 2010)
+ve: quicker than prospective cohort study because follow up period already elapsed
-ve: reliant on completeness & accuracy of routine records
Retrospective (Historical) Cohort Study
Which cohort study is this?
Sample selected to be representative of target population. Multiple exposures are
ascertained at baseline and related to multiple outcomes of interest at follow-up.
e.g. the Avon Longitudinal Study of Parents and Children (ALSPAC), investigating the impact
of adverse childhood experiences on later mental health (Fisher et al., 2012)
Population (Birth) Cohort Study
What is accelerated cohort study?
An accelerated cohort design is a structured population cohort design that takes multiple
single cohorts, each one starting at a different age
e.g. 1) Great Smoky Mountains Study (Costello et al., 1996) – first example of accelerated
cohort used in psychiatric epidemiology; recruited all school children aged 9, 11, 13 living
in defined catchment area and followed up at nine time points; 2) REACH study
(www.reachstudy.com)
+ve: ability to span the age range of interest in a shorted period of time than possible
when using a single cohort; should be less affected by drop-out
-ve: possible cohort effects
How to select the study population in cohort studies?
Balance of feasibility and generalisability.
Unexposed group should be as similar as possible to
exposed group on all factors except exposure
What are the sources of data for cohort studies?
Primary data collection
Electronic Healthcare Data
Health Insurance Data
Registries
Name a few specific considerations for cohort studies
Loss to follow up & misclassification bias
• “Disease free” at baseline
• Threshold of “caseness”
• Prodrome (symptoms prior to ‘disorder’)
• Relapsing and remitting conditions
• Lack of routine outcome data
Pros and Cons of Cohort Studies
Pros-Can study rare exposures
• Can look at many outcomes
• Establish temporal sequence
• Can look at reciprocal relationships
Cons- Loss to follow-up • Can be expensive • Not good for rare outcomes • Can be time consuming • Constrained by the exposures you measure at baseline
Pros and Cons of Case-Control Studies
Pros-Good for rare outcomes
• Quicker and cheaper vs. cohort
• Good for chronic diseases and diseases with
long latency periods – don’t have to wait for
disease to occur
• Examine multiple exposures
Cons- Reverse causality
• Biases
• Selection bias – are cases + controls from
same population?
• Recall bias – do cases remember their
exposures differently?
• Observer bias – if not blind to case/control
status, may influence exposure measurement
• Not good for rare exposures
• Confounding – matching
