Topic 9 - Study Designs Flashcards
A study design plans for selecting study _____ and obtaining _____ about them
A study design plans for selecting study subjects and obtaining data about them
A descriptive study is a study undertaken without a specific ________.
A descriptive study is a study undertaken without a specific hypothesis
An Observational studies (AKA Analytical studies)
* Test a __________: Concerned with an ________ causing a certain ______ (disease)
* _____ is an important characteristic of these studies: One of _____ criteria for causation
- Test a hypothesis: Concerned with an exposure causing a certain outcome (disease)
- Time is an important characteristics of these studies: One of Hill’s criteria for causation
Experimental studies
- Test the effect of an _______ (e.g. therapy, challenge)
- ______ or _____ setting
- Non-_______ and ________
- Test the effect of an intervention (e.g. therapy, challenge)
- Laboratory or Field setting
- Non-randomized and Randomized
What is a research synthesis?
- Research Synthesis
- Non-systematic reviews, Systematic reviews, and Meta-analyses
Descriptive studies describe?
Attributes
Explanatory studies explains?
Associations
Repeated associations are a summary of?
Summary of what consistently changes disease status ?
List the study designs in order from least to greatest strength of evidence?
Cohort = best amount of info (almost a control trial - due to different arms and groups being monitored over time)
Case reports are used for:
______ (or very _____) cases
- _____ or _____ condition
- _______/____ manifestation or _______ of a common condition
- Could help to build _________
Single (or very few) cases
- Rare or new condition
- Unusual/new manifestation or management of a common condition
- Could help to build hypothesis
Case series describe the _______ course of a condition shared by a ______ of subjects
- Document the ____/_____/______
- ________ estimation when representative of cases in the population
Case series describe the clinical course of a condition shared by a group of subjects
- Document the Who/When/Where
- Prognosis estimation when representative of cases in the population
Surveys describe the _______ & ________ of a condition in a population at a point in time
- Focus on _____ feature at the time
- Requires proper _______ protocol and collection of information (___________)
- ________ valid when representative sampling
Describe the frequency & distribution of a condition in a population at a point in time
- Focus on one single feature at the time
- Requires proper sampling protocol and collection of information (questionnaire)
- Extrapolation valid when representative sampling
not looking at risk factors, only looking at presence, or lack thereof, of disease
Case-control
* Investigating ____ factors for _____ diseases
* …you can also think about _______
situations
Case-control
* Investigating risk factors for rare
diseases
* …you can also think about outbreak
situations
This case below is an example of what type of study design?
Perfect dataset with medical records
for every dog in a country and details
about exposure in first year of life
* Risk of bladder cancer
* Benzidine + : 32 per 100,000
* Benzidine - : 20 per 100,000
* Risk Ratio = 1.6
* Odds Ratio = 1.55
* (60 x 278,943) / (57 x 188,940)
Case-control
Contd.
- Instead, enroll ALL dogs with bladder
cancer and a sample of dogs who did
not - Can’t calculate risk of bladder in E+
or E- because we don’t know the
size of the population at risk - Risk in E+ is definitely not 24%
of the population (60 / 249) - OR = 1.55
- (60 x 279) / (57x189)
Once you collect first few cases, go back in time and ask them questions related to past: where did you eat, go, etc. anything you think led to them testing positive.
What are the advantages of case-control studies?
- Efficient with rare diseases
- Quick to undertake
- Cheaper than other studies
What are the disadvantages of case-control studies?
- No information on disease ____ or ______ in the studied population
- Reliant on _________ of study participants
- Difficult to ensure _________ selection of control group
- No information on disease risk or incidence in the studied population
- Reliant on recollection of study participants
- Difficult to ensure unbiased selection of control group
What are cross sectional studies?
Sample of individuals from a population taken at a single or relatively short period of time
* “Snap-shot” in time
Often requires
* Surveys (sampling strategy)
* Questionnaires
Seeing what is occurring at a single moment in time.
Look for breed, age, sex, risk factors, exposure to anything you think might be related to disease.
What are the advantages of an observational studies?
- _______ to undertake
- Can estimate ____ (________)
- ________ than other studies
- No ________-___ period
- Quick to undertake
- Can estimate risk (prevalence)
- Cheaper than other studies
- No follow-up period
What are the disadvantages of cross-sectional studies?
- No information on disease ______ in the studied population
- Not suitable for diseases of ______ duration
- Imagine a X-sec study for milk fever!
- First 24-hrs post-parturition
- Imagine a X-sec study for milk fever!
- Difficult to investigate _______/_____ relationships
- Don’t know if E__ occurred before D__
Not good for ____ studies because if you are showing up randomly for a moment in time, you are most likely not going to get anything.
- No information on disease incidence
in the studied population - Not suitable for diseases of short
duration- Imagine a X-sec study for milk fever!
- First 24-hrs post-parturition
- Imagine a X-sec study for milk fever!
- Difficult to investigate cause/effect
relationships - Don’t know if E+ occurred before D+
Not good for rare studies because if you are showing up randomly for a moment in time, you are most likely not going to get anything.
What are cohort studies?
- Comparing disease ______ over _____ among groups (cohorts) with different ______
- ________ or _________
- Comparing disease incidence over time among groups (cohorts) with different exposures
- Prospective or Retrospective
Observational studies
More powerful studies go forward in time
Less powerful studies go back in time b/c they get history charts
What are the advantages of cohort studies?
- Absolute incidence of disease in E__and E__
- Exposure is recorded _______ disease
- Well-suited for _____ exposures (but must be a fairly _______ disease; if doctor is doing a rare treatment for a common disease)
- No need to reflect E__ or E___ prevalence from population
- Absolute incidence of disease in E+
and E- - Exposure is recorded BEFORE disease
- Well-suited for rare exposures (but must be a fairly common disease; if doctor is doing a rare treatment for a common disease)
- No need to reflect E+ or E- prevalence from population
What are the disadvantages of cohort studies?
- ______ follow-up period
- _________
- Losses to follow-up are problems
- Introduce potential _____
- Rare diseases require ______ groups
- Long follow-up period
- Expensive
- Losses to follow-up are problems
- Introduce potential bias
- Rare diseases require LARGE groups
What are some examples of hybrid studies?
Nested case-control
Case-crossover
Panel
Repeated survey
What is an ecological study?
Unit of analysis is a _____ (e.g. cities, state, countries, etc.)
* Summary measures of ______ and summary measures of ______ are compared and inference is made at the individual level
*Never take these results seriously?
Unit of analysis is a group (e.g. cities, state, countries, etc.)
* Summary measures of exposure and summary measures of outcomes are compared and
inference is made at the individual level
*Never take these results seriously?
What are the advantages of ecological studies?
- Extremely _____ and _______ to perform
- Provide clues about associations between?
- Extremely quick and inexpensive to perform
- Provide clues about associations between E+ and D+
What are the disadvantages of ecological studies?
- Extremely prone to ______!!!!
- “Ecological ________”
- Assumption that an observed relationship in ________ data will hold at the ________ level!
- Extremely prone to bias!!!!
- “Ecological Fallacy”
- Assumption that an observed relationship in aggregated data will hold at the individual level!
Questions should be of individuals, not of the entire country.
Pointless!
Homogenous: make all subjects the same to control environment as much as possible
Randomization is crucial
What are controlled trials?
Population has to have a condition and then see if the condition remains or goes away over time.
Want the doctors and patients to be blind to treatment
The higher on the y axis, the better you get.
Just b/c in study they get better –> hawthorne effect
thinking when they take something they will feel better –> placebo
Looking for lameness scores in dogs with secondary arthritis. Asking owners if lameness was getting better over time vs. pressure plate in animal and if it was making condition better or worse? owners saying animal getting better even though they were not
What are other terms used to refer to controlled trials?
Intervention
control group as comparison group is important
looking for improvement of disease condition
B/c treatment, you expect patients in treatment group to have a lower risk of disease compared to the placebo group.
How do you calculate a risk ratio?
What can happen in controlled trials?
- Loss to follow-up (a type of selection bias)
- Losses should be minimal (<20%)
- Were they equal in both groups?
- If losses were related to the exposure and also to the outcome = BIAS (untestable)
Make sure you don’t lose anyone along the way.
Worry when we have reached 20% loss in a group. Need to make sure you are losing the same number in each group, if not matching then you need to worry about bias
EXAMPLE
If patients who didn’t respond to treatment died (or dropped out) before the end of the study, the treatment effect could look much better than it actually is!
b/c it would only effect people who did not respond
What are the steps of a systematic review?
Qualitative review is still considered a systematic review because you went through all of the steps
How does meta-analysis pool weight across studies?
- _________ average treatment effects
- Study weighted by their _______
- More _____ = more _____!
- Example: Effect of bovine somatotropin (rBST) on risk ratio
for clinical mastitis - 1.27 (95%CI: 1.13-1.43)
- Weighted average treatment effects
- Study weighted by their precision
- More precise = more weight!
- Example: Effect of bovine somatotropin (rBST) on risk ratio
for clinical mastitis - 1.27 (95%CI: 1.13-1.43)
What is an extra-advantages of a meta-analysis
Use a ________ plot to see when you gained sufficient evidence over time to make decision!
Use a cumulative plot to see when you gained sufficient evidence over time to make decision!
Meta-analysis are applicable to many study designs.
