OBJ - Overview of Study Design Flashcards
Ways to express risk
Absolute risk Risk difference Relative risk Population-attributable risk Odds ratio
Randomized clinical trials
Gold standard with control & variables
Experimental Study that investigator assigns exposure
Randomized? => RCT
Not randomized = Non-RCT
Comparison Groups • No intervention • Observation – Hawthorne effect: phenomenon whereby individuals improve or modify an aspect of their behavior in response to their awareness of being observed. • Placebo • Usual care
Blinding/Masking
Cohort
Observational with comparison group => Analytical Study
Exposure -> Outcome
First select exposed/unexposed
=> Then see if disease develops or not
RELATIVE RISK & INCIDENCE
Def: Group of people with something in common when they are first assembled
Classified as exposed/unexposed (do they have the risk factor?)
Draw conclusion for exposure/risk factor based on incidence in cohort
Criteria:
Must not have the disease/outcome when assembled
Observed over a meaningful period of time
All members must be observed over the full follow up period or methods to account for dropouts
Ex: Age, DOB, exposure, disease, therapeutic intervention, preventative invention
Sub Types:
Prospective (collected & watched), Retrospective/historical (ID’d from past records and followed up in the present
Pros:
• Best available substitute for non-experiments
• Can evaluate multiple outcomes
• Provides an actual measure of risk of the
outcome of interest
• Can extract incidence and relative risk
Cons:
• Subject to many biases
• Compare incidence of outcome between
exposed and unexposed groups
• Potential for loss to follow-up
• Requires large number of subjects
• Not intended for the study of rare outcomes
• Takes a long time – not efficient for outcomes
that take a long time to develop
• Expensive! Resources, staff, space, incentives, overhead costs, etc.
Case-control Studies
Observational with comparison group => Analytical Study
Know the outcome then look back time to exposure
Exposure <- Outcome
ODDS RATIO
Pros:
• Study of rare diseases
• Study of diseases with long latency
• Requires relatively few subjects
• Requires less time than cohort study
• Allows for the evaluation of multiple exposures as potential causes of disease
• Well-suited for outbreak investigations
Cons:
• Relies on recall of subjects for information on
past exposure
• Problems of recall bias among cases
• Selection of appropriate control group may be difficult
• Yields the odds ratio which is only an estimate of relative risk
• Cannot calculate incidence rates
Cross-sectional
Observational with comparison group => Analytical Study
Exposure & Outcome at the same time
Pros:
– Provides good measure of disease prevalence
– Can provide information on what to expect in a clinical setting
– Used in evaluating screening and diagnostic tests
– Can help plan health services
– Can be quick, easy, and inexpensive
Cons:
– Measurement of disease (outcome) and exposure status occurs at the same time
– Temporal relationships between exposure and disease difficult to assess -> Cannot determine causality
– Limited to study of prevalence
• cannot determine disease incidence
• cannot tease apart incidence and duration
Ecological studies
aka: Population/Ecologic/Correlational
Studies
Observational & No comparison group
• Compares disease rates between populations
• Exposure is only known for groups, not the individuals in the groups
• Data are based on measurements of risk and
disease averaged over populations
• Relationship between exposure and outcome is an “ecological correlation” or “aggregate risk”
• Frequently generate important hypotheses
that need further, analytic research
• Make comparisons between large groups that
otherwise would be impossible
• Can be done inexpensively, sometimes with
publicly available information
Typically Step 1 in research
Pros:
• Provide clues to etiological hypotheses
• Help in setting research priorities
• Study large populations at low cost
• Address questions of environmental health that might be difficult to study using other epidemiologic approaches
Cons:
• No data on individual study subjects
• “Ecologic fallacy”
– Inappropriate inference from ecologic data
– When a true relationship from group data does not mean a true relationship at the individual level
Case series
Observational & No comparison group
• Observational, detailed description of clinical
presentation of a patient(s)
– Case series = more than a few patients
• Analysis
– Descriptive statistics only
– Person, place, time
– No comparative analysis
Case report
Observational & No comparison group
Essential link between clinical medicine and
public health
• Morning rounds or publication of an “odd” case
– Known disease in unusual population
– Previously unrecognized syndrome
– More severe disease or different characteristics
– Transmitted in mode not generally seen
• Observational, detailed description of clinical
presentation of a patient(s)
– Case report = one or a few patients
• Analysis
– Descriptive statistics only
– Person, place, time
– No comparative analysis
Absolute risk
Define, calculate and interpret
Risk Difference
• Difference between the risk of disease in
exposed and unexposed groups
• Risk of disease following exposure which is
either greater or less than the risk experienced by someone not exposed
AR = A/(A+B) ― C/(C+D)
= Incidence of exposed ― Incidence of unexposed
Exposed risk minus Background risk of un-exposed
Relative risk
DEF: Probability of a disease occurring/developing in exposed vs non-exposed
Use with Cohort studies
CALCULATE:
Going ACROSS 2x2 table:
= A/(A+B) / C/(C+D)
INTERPRET:
Range is from 0 to infinity
RR < 1 Risk for exposed groups < unexposed
(protective effect)
RR = 1 Equal risk for exposed & unexposed
RR > 1 Risk for exposed group > unexposed
(actual “RISK” factor)
ex: 1.64 times greater for men than women or 64% greater for men than women
Population-attributable risk
Define, calculate and interpret
- AR x prevalence of exposure
- Excess incidence in a community that is due to the risk factor
Odds ratio
• Odds Ratio (OR) is useful when we CANNOT
calculate INCIDENCE
• OR gives a similar measure of risk as the Risk
Ratio (RR), if incidence of disease is low (< 1%) - why you can’t calculate a good incidence
CALCULATE:
Odds ratio = AD / BC (CRISS CROSS)
INTERPRETATION: Ranges from 0 to infinity If < 1 indicates decreased risk If = 1 indicates equal risk If > 1 indicates increased risk
Risk Difference/Attributable Risk
DEF: Difference between the risk of disease in exposed and unexposed groups
• Risk of disease following exposure which is
either greater or less than the risk experienced by someone not exposed
CALCULATE:
AR
= A/(A+B) ― C/(C+D)
=Incidence exposed ― Incidence unexposed
INTERPRETATION:
• AR = 32 per 1,000 births in the exposed group can be attributed to _____.
Ranking of Trials based on
Level of Evidence
BEST = Randomized Controlled Trial - FDA Cohort study - NIH Case-control study - CDC Cross-sectional study Ecologic study Case series