Study Designs Flashcards
Ecological Studies Disadvantages -3
1) May not have data on confounders
- limited for assessing independent associations and causal interpretations
2) Ecologic Fallacy
3) Limited in assessing causal interpretations
Ecologic Fallacy
An association at the aggregate level does not guarantee an association at the individual level
-without individual-level data, do not know joint distribution of exposure and outcome
Ecological Study Design Pros - 4
1) relatively quick/cheap, often able to use existing data
2) Simple and intuitive study design
3) Useful for generating leads in more definitive studies
4) Often appropriate for studying broad social and cultural processes (i.e. laws regarding environment/personal safety)
Correlation Coefficient
Used for Ecological Study design, -1 to +1 0 = no correlation strength of linear relationship between two variables positive/negative correlation
Cross-Sectional Disadvantages - 3
1) Difficult to establish a temporal relationship
2) Not as useful for rare diseases or those with SHORT duration
3) Only includes prevalent cases!
-gives lesser weight to those cases who have already overcome/survived disease
-fails to include those who have died
=>association may reflect SURVIVAL/DURATION and NOT INCIDENCE/RISK - depends on question
*careful to calculate Prevalence Ratios (and only cumulative ratios when NEW cases are counted)
Cross-Sectional Study Advantages - 3
1) often quick/ inexpensive
2) useful for describing characteristics of target population and estimating the burden of the disease
- can investigate many potential risk factors
- can provide estimates of exposure/association
3) Associations may be useful for hypothesis generation (“suggestive”)
Cohort Studies Advantages - 6
1) Useful for investigating a multiple exposures and multiple outcomes
2) Evaluate absolute impact of exposure/disease on the population
3) Know temporal association for exposure/disease
4) less vulnerable to certain types of bias
5) Special cohorts = rare exposures
6) Can estimate CI, IR
Prospective Cohort Study
2 pros, 2 cons
1) Starts with exposure and follow through to disease
Cons
2) expensive as it takes more time to follow everyone
3) inefficient for diseases with long latency/induction
Pros
4) Better exposure data
5) Better Follow-up data
Retrospective Cohort Study
2 pros, 2 cons
-exposure and incidence have already happened
Pros
1) Faster and less costly
2) Good for disease with long induction/ latent period
Cons
3) Often encounter missing/ incomplete medical data
4) Recall Bias affects exposure data
General Cohort
w/ one pro
Randomly select individuals without regard to exposure status
-ex. professional groups, educational groups, geographic regions, etc.
Comparison group = INTERNAL
-cohort separated by exposure categories
Pro - often the similarity between exposure groups is high
Special (exposed) Cohort
w/ one pro
Focus on population with shared exposure
-ex. proximity to waste sites/ factory toxicants
Comparison = GENERAL population/ External “COMPARISON COHORT”
-beware of healthy worker effect
PRO - good for rare exposures
“Healthy Worker Effect”
Type of Bias
-workers tend to be healthier than the general population, therefore, when comparing disease in a worker population to a general population, the burden of the disease is often underestimated.
Open Cohort (dynamic)
membership is transient, and is defined by a changeable characteristic
i.e. smokers/ residents of an area
Fixed Cohort
membership=permanent
i.e. people present at one location at a specific time
OR people employed somewhere on a certain date
Closed Cohort
Membership=permanent
- no loss to follow-up!!!
e. g.g attendees at a company picnic
Cons Cohort - 3
1) expensive and laborious
2) Inefficient for rare diseases
3) Losses to follow-up/ recollection bias
Cons Intervention Studies - 6
1) Losses to follow-up
2) Loss of willingness to participate in intervention/ compliance with intervention
- this needs to be measured as it can affect exposure status
3) Not the same degree of control as laboratory studies
4) Generalizability of the findings
5) Uniqueness of participants/ those willing to take on an active intervention protocol
6) Sustainability/ Feasibility
Compliance = Adherence
following intervention protocol
Intervention Studies
CI or IR (depending on PT data, if everyone was followed for the same amount of time, you often see CI)
-estimate RD (risk difference)
Intent-to-Treat
- all individuals assigned to intervention/treatment are analyzed with the group regardless of compliance
- this maintains the comparability of groups and statistical power
- allows for “real world” analysis conditions
Per-Protocol
Only analyze individuals in the treatment who adhered to to the intervention/ treatment protocol
Individual vs. Community level
Treatment is randomly assigned at the individual level
Community: Treatment is assigned to people at the community level
Comparison conditions for Intervention Studies - 3
1) Placebo
2) Standard of care
-often used in drug trials for new drugs
3) Something else - often behavioral intervention trials
e.g. delayed intervention
new vaccine; give existing vaccine
Blinding
makes participants/ investigators/ data collectors unaware of treatment assignment
- sometimes not possible (when participants have to be aware of receiving treatment due to study design)
- sometimes not necessary (possibly with hard outcomes like death, where hope/expectation has little effect upon the outcome)
Equipoise
when there is general doubt regarding efficacy yet sufficient belief that it will work
Ethics of Intervention Research
-need equipoise
-need sufficient evidence from other study designs
-need intervention that is feasible to change
-sometimes mid-way through trial it is no longer ethical to continue study because it is either obvious that the treatment works, or that it is adverse
-comparison group
-sustainability/feasibility
can the intervention be continued after the study ends? logistics? funding? acceptability?
Pros Intervention Studies - 2
-scientifically likely to produce a fair comparison
use of randomization/ blinding/placebos
-participants may benefit from study
Nested Case Study
identified cases of disease from an existing COHORT study, and then taken a sample of that population (at-risk population) for comparison. Choose some number (e.g., 2 times as many as cases). Exposure histories are determined for each group.
- Done for purposes of EFFICIENCY
- ALL case-control studies are nested in a hypothetical cohort, but ones labelled as such are nested in an existing cohort
Case-Control Study Protocol:
1) Clear Case-Definition
2) Identify the source of cases (may involve sampling)
3) Incident vs. Prevalent cases
- incident allows for behavior reporting before response to onset of disease
-Prevalent can allow people time to won up to risky behavior responsible for disease onset
4) Control Selection
“WOULD criterion” - would have been caught if they had developed the disease
Controls: Case-Control Study
1) Population
i) high likelihood controls = same base population
ii) high refusal rate, recall inaccuracy, time consuming, need enumerated list
2) Hospital/Clinic - better recall, accessible, more willing, similar selection factors
i) Disease needs to be unrelated to exposure of interest
ii) Disease needs to have similar referral patterns
3) Others - friends/ spouses/ siblings/ twins
Base Population Sampling
investigators select controls from the population at risk at the BEGINNING of the case diagnosis/ accrual period
-this control base is allowed to include those who go on to develop the disease, allowed because the control group is supposed to present risk
Risk-Set Sampling
Every time a case is diagnosed one or more controls are selected from other members of the cohort who, at that time, do not have the diagnosis
- Allows for calculation of IRR, as you have PT
- this control base is allowed to include those who go on to develop the disease, allowed because the control group is supposed to present risk
Survivor Sampling
investigators select controls from the population at risk at the END of the case diagnosis/ accrual period
Pros Case-Control - 6
1) Exposure data is expensive/difficult to obtain
2) Outcome is Rare
3) Disease has a long-induciton/latent period
4) Little is known about exposure, allows for assessment of multiple exposure
5) Underlying population is dynamic
6) Outbreak investigations
Cons Case-Control - 3
1) Can make it difficult to assess temporal relationship of exposure/outcome
2) Assessing validity of control group
3) Recall bias in exposure analysis
4) can’t calculate: RiskDifference, PRD
need rare disease: RR, APe, APt,
risk-set sampling allows: IRR,
