How do study designs differ?
1. Number of Observations Made
2. Directionality of Exposure
3. Data collection methods
4. Timing of data collection
5. Unit of observation
6. Availability of subjects
Describe the difference between the experimental and observational approach
1. Is the study factor (i.e. exposure) artifically manipulated by the investigator or by others?
2. If the exposure is artifically manipulated by the investigator or by others, is the exposure randomly allocated to all subjects?
1. Question to ask: Is the study factor exposure artifically manipulated by the investigator or others?
Yes?--then ask, is there a random process used to determine who has the exposure and who doesn't?
If there is a random process used to determine who has the exposure and who doesnt then its a experimental study.
If there is NOT a random process used to determine who has the exposure and who doesn't then it is a QUASI-EXPERIMENTAL study
If it is observational design (the study factor (exposure) is not artifically manipulated by the investigator or others), then what is the next question you must ask
Is there hypothesis testing?
If yes: Analytical study
If no: Descriptive study
If descriptive study then ask... is the unit of the analysis the group?
if yes... then ecological study
If the exposure and outcome status are determined at the same time, its a cross-sectional study.
Why do we study descriptive study designs
Remember, no hypothesis testing.
We use descriptive study designs to study the frequency and distribution of health related states.
It is hypothesis generating, not hypothesis testing
Examples include Cross sectional (which can also be an analytic hypothesis testing study), ecological study, case reports, and case series.
What are case reports and case series?
2 types of descriptive study designs
case reports: Describing a few cases. Interesting findings
Case series: Summarizing the characteristics of cases.
Both will ONLY include cases.
Analytic Study design include
Cohort, Cross sectional, And case control
Hypothesis testing method of investigation of an association between a study factor and an outcome.
Types of Ecological Studies
These are descriptive studies. Hypothesis GENERATING.
Ecological Comparison study
Ecological Trend Study
What is Ecological Comparison Study?
1/2 types of the ecological studies which are a type of descriptive study which does not require hypothesis testing. Rather, it is hypothesis generating.
Exposure rates and disease rates and disease rates are collected at te same time over the same time period.
Aggregates are defined by geographic area
What is ecological trend study
Changes in exposure and changes in disease within the same geographic unit
Aggregates are defined by time.
Ecological comparison study
Exposure and disease rates are measured over the same time peruod with aggregates defined by geographic area
Ex/ Lung cancer due to smoking in france vs lung cancer due to smoking in spain.
The unit of analysis is the population, NOT the indivdual.
Ecological trend study is
Changes in exposure and changes in disease within the SAME geographic unit with aggregates being defined by time.
For ex/ Death due to smoking leading to lung cancer in 1999 in France vs Death due to smoking leading to lung cancer in 2009 in france.
An example of an ecological comparison study
Exposure rates and disease rates are measured int he same time period but in DIFFERENT geographic areas.
Ex/ Obesity=Outcome vs. Percapita Supply of calories=exposure in France, germany, USA in 2009
Give an example of ecological trend study
What is ecological trend study?
Changes in exposure and changes in disease within the same geographic Unit
Place: Same geographic area
Aggregates defined by?: time.
like lung cancer and smoking in france in 1999 vs lung cancer and smoking in france in 2009
Ecological studies. The unit of analysis is
the group NOT the individual. The correlations are obtained between exposure rates and disease rates among different groups or popu;ations
Measures available in ecological studies
Summary measure of the frequency of exposure
Summary measure of the frequency of disease
Joint distribution of exposure and disease AT individual level is NOT known.
So you do a 2*2 table for each area and ten plot outcome vs exposure.
What is the ecological fallacy
It is what makes the ecological study, a descriptive study which means it is HYP testing not HYP producing, and also means that it has either ecolgoical comparison or ecological trend study, LESS STRONG than a study like a cohort study!
In the indvidual populations, the outcome may decrease with exposure. BUT in the mean population, the outcome will increase with exposure.
This means that observations made at the group level may not reflect THE REAL exposure disease relationship at the individual level. Incorrect references about individual level association. We might think that the individuals outcome status is going to increase with exposur estatus, but in relatiy, that could be wrong! It coudl be resverse! In short, the conclusions obtained from an ecological study may be the reverse of thsoe from a study that collets data on individual subjects
Advantages of ecological study
A good first approach in determining whether an association exists between an exposure and an outcome
Allows you to generate a hypotehsis
Suggests promising avenues of research
Group data not individual data.
Does not demonstrate a causal relationship.
Information about the confounders is not available generally
The quality of the data might be shittty
You might be susceptible to ecological fallacy
CROSS SECTIONAL STUDIES
1. This is a type of which study?
2. Exposure and disease measures are obtained how?
2a. How many periods of observation are there?
3. Exposure and disese measures are obtained when?
4. Also known as what?
5. What kind of sampling does it use?
6. Define this kind of sampling
7. Design of a cross sectional study
8. How do we measure exposure and disease in this kind of study
9. What do we use cross sectional studies for?
10. Advantages of cross sectional studies
1. Considered BOTH analytical, AND descriptive
2. at the indivdual level
3. at the same time
4. Prevalence study
5. Probability and non-probability sampling
6. Probability sampling has the characteristic that every individual in a population has a nonzero chance of being included in the sample. Examples include simple random samples, systematic samples, stratified samples.
Nonprobability Sample: One based on a sampling plan that doesn't have the above feature. Like quota sampling and judgemental samples.
7. Start with a population-->Gather data on exposure and disease outcome. Then you get 4 groups possible: exposed hve disease, nonexp have diseased. exp don't have disease, non exposed, do not have disease
8. measure with questionaries, records, physical exam measurements, lab tests, defined diagnositc criteria, (clear, objective, standardized criteria) (may divide cases into definite, probable, and possible disease categories)
9. To determine the health status of a population. Hypothesis generating ( descriptive study design) & Hypothesis testing (because it also falls into the category of analytical study design) , to design intervetion planning, and to estimate the magnitude and distribution of a health problem
10. a) it is a fast study (short time needed)
b) If the results come from a general population, the results are generalizable
c) it's cheap
a. its not useful for rare diseases (because not a lot of people in population will have it, so itll be really insignificant) (low prevalence diseases)
b. does not gives us any incidence data (case control studies do not directly provide incidence data)
c. It can't determine the temporality of exposure and disease (remember, its just a snapshot of the population at a certain time!)