What is the main limitation of ecological or correlationnal (aka descriptive study at a population level) descriptive study? How is this kind of study done ?

Ecologic fallacy : interpretation of statistical data where inferences about the nature of individuals are deduced from inference for the group to which those individuals belong It's basically a superposition of 2 descriptive studies : one of disease and one of exposure to make an hypothesis about those 2 things at the population level

Epidemiology (du coup) Flashcards by Marie Picard

What are the 2 main types of descriptives studies ?

Individual level
Populational level (ecological)

How well did you know this?

Not at all

Perfectly

What is the goal of descriptive studies? What information is crucial for that kind of studies ?

Goals
- Hypothesis generating
- Demonstrating trends
- Resource allocation
- Need for education
Question
- Who : age, sex, race
- What : diseases usually, programs sometimes (before/after implementing it)
- When : calendar type, season
- Whre : country, povince, city

How well did you know this?

Not at all

Perfectly

What is the main limitation of ecological or correlationnal (aka descriptive study at a population level) descriptive study?
How is this kind of study done ?

Ecologic fallacy : interpretation of statistical data where inferences about the nature of individuals are deduced from inference for the group to which those individuals belong
It’s basically a superposition of 2 descriptive studies : one of disease and one of exposure to make an hypothesis about those 2 things at the population level

How well did you know this?

Not at all

Perfectly

What are the advantages of analytical: observational studies?

Representative populations
Represents routine clinical practice
Usually more time and cost-efficient than clinical trials

How well did you know this?

Not at all

Perfectly

What are the limitations of analytical: observational studies?

More susceptible to bias

How well did you know this?

Not at all

Perfectly

What are the 3 types of observational studies?

Cohorte
Cross-sectional
Case-control (includes nested studies)

How well did you know this?

Not at all

Perfectly

What is a cohort study and its advantages?

Designated group of individuals, who are followed or traced over a period of time. We compare those exposed and those NOT exposed
- we also classify by exposure and compare outcome/event rates
Advantages
- Can study multiple diseases/outcomes
- Exposure more clearly precedes disease
- May be only way to study rare exposures

How well did you know this?

Not at all

Perfectly

What are the limitations of a cohort study?

Major difficulty is loss to follow up
Not efficient for rare diseases, long latency

How well did you know this?

Not at all

Perfectly

What is a cross-sectional study and its advantages/disadvantages?

Gather data on exposure and disease simultaneously
Limitation: What came first? Exposure or disease?

How well did you know this?

Not at all

Perfectly

What is a case-control study? What’s its main advantages ?

Subjects are selected based upon the OUTCOME: presence (case) or absence (control) of a particular disease
Advantages
- Only rare diseases
- Mimics the clinical paradigm
- Simultaneous assessment of multiple exposures
- Should include all incident cases and come from and represent the same population

How well did you know this?

Not at all

Perfectly

What are the limitations of a case-control study?

Not good for rare exposures
Often a challenge in field study to have the same population of origin

How well did you know this?

Not at all

Perfectly

What are the goal and the 2 types of analytical: experimental studies?

Goal: Try to determine causality, studies of prognosis

Types:

RCT
Systematic review

How well did you know this?

Not at all

Perfectly

What is Rate Ratio?

Incidence in exposed

___________________

Incidence in unexposed

How well did you know this?

Not at all

Perfectly

What is hazard ratio?

Time to event in exposed

____________________

Time to event in unexposed

How well did you know this?

Not at all

Perfectly

What is risk ratio?

Cumulative incidence in exposed

__________________________

Cumulative incidence in unexposed

How well did you know this?

Not at all

Perfectly

What is Odds ratio? How do you calulate it ? How do you interpret it ?

Odds of exposure in cases / odds of exposure in controls
- = ad / bc
people who have an injury (cases) have (insert odds ratio number) times more chance of having the exposure

How well did you know this?

Not at all

Perfectly

What is attributable risk?

incidence risk of exposed population – incidence risk of un-exposed population

gives you a number of cases we could have prevented without the exposure / time on which the incidence rate is calculated

How well did you know this?

Not at all

Perfectly

What is Attributable fraction?

Incidence in exposed - incidence in unexposed

______________________________________

Incidence in exposed

*the numerator is actually the attributable risk

gives you a sense of the impact of removing the exposure (X% numbers of cases are caused by the exposure)

What is Pop attributable risk (PAR)?

Incidence in all - incidence in unexposed

What is Pop attributable fraction (Pop attributable risk percent)?

PAR / incidence in exposed

What is a bias?

A bias is a systematic error(s) in the design and/or conduct of studies. It is unaffected by study size and results in deviation from the truth (may create or mask associations).
- Only precision in affected by size.
We have to consider the direction and magnitude of the bias.

Define selection bias.

Selection IN and OUT
- Systematic differences between groups being compared arising from
  - method of selection or
  - reasons for participation
- = over - or under - representation between comparison groups
Lost to follow-up
- error due to systematic differences in the retention of subjects between the groups being compared
- one of the most important threats to study validity

How do we minimize selection bias?

Clear inclusion and exclusion criteria.
Include all new (incident) cases.
Assure high participation and low dropout rates.
For selection bias to occur, selection of subjects or loss to follow up has to be associated with both the exposure and the outcome

Define information bias.

Misclassification or error of exposure or outcome, can be due to the source of data or an investigator error

Two types :

Non-differential : same error between the test and control groups
Differential : error affects one of the groups being compared more than the other

How do we minimize information bias?

1. Use identical measures in groups being compared (case/control, exposed or not) 2. Blinded assessment of exposure and outcome.

Define cofounding bias?

* Results from a third factor that distorts the association between exposure and outcome. Error due to systematic differences in baseline characteristics of the groups being compared “Comparing apples & oranges”. * can completely hide the association * common cofounding factors : age, sex, SES

How do we minimize cofounding bias?

Occurs in observational studies of drug benefits and risks At the design stage: * Randomization * Restriction * Matching At the analysis stage: * Stratified analysis * Multivariate analysis

What are the 2 main type of study designs in epidemiology ? How are they different ?

* **Descriptive** * ****hypothesis *generating* * demonstration of trends * **Analytical** * hypothesis *testing* * provides evidence to establish causality

Can you conclude a link or causal relation from an ecological study ?

Nop. **Hypothesis generation only**

What are the 2 kinds of study designs in analytical studies ?

* **experimental/interventional** * systematic reviews/meta-analysis * randomized controlled trials * **observational** * cross-sectional studies * cohort studies * case-control studies * other designs

What are the 2 kinds of cohort studies and what's their difference ?

* **Prospective** and **retrospective** * both prospective and retrospective cohort studies asses exposure first and outcome second, however this is done in different time frames

what should be your odds of exposure in both the control group and the cases groups in a case-control study ?

The same (you should have the same proportion of exposed people on unexposed people)

what is a nested case-control study ?

An efficient way to select your cohort population so that controls come from and represent the same population than your diseased persons

what are the different measure of association in analytical studies ?

* all are sometimes termed **relative risks : RR** * **rate ratio** : incidence rate in exposed/incidence rate in the unexposed * **hazard ratio** : time to event in exposed / time to event in non-exposed * **risk ratio** : cumulative incidence in exposed/cumulative incidence in unexposed * **odds ratio :** odds of exposure in the cases/ odds of exposure in the controls

in well designed studies, what measures of association are equivalent ?

* all the relative risks (RR) * rate ratio * risk ratio * odds ratio

when should the odds ratios be equivalent to the relative risk ?

* if the disease is **rare** (\<10 % pop.) * if controls are sampled at **equivalent times** to when cases are occuring (risk set sampling, aka a subject chosen as a control at one point in time may later become a case)

What are the different possible measures of impact (it's ACTUAL NUMBERS !) in studies and their significance ?

* **attributable risk :** difference in incidence rates or in risks (_cumulative incidence)_ * *amount of disease we could prevent by removing the exposure* * **attributable fracion (aka attributation risk %) :** incidence in exposed minus in unexposed (_attributable risk)_ / incidence in exposed

what is a better measure of public health impact : relative risk or attributable risk ?

attributable risk

can impact measure be calculated in case of case-control studies ?

* **Attributable risk (risk difference) cannot be calculated** * Attributable fraction = odds ratio - 1 / odds ratio (OR-1/OR) * Population attributable fraction (PAF) = AF x prevalence of exposure in cases

How could you measure how much a disease in the whole population can be attributed to a certain exposure ? What do they mean ? What is it dependant of ?

* by the **population attributable risk** and **population attributable** **fraction** * **PAR :** incidence in all - incidence in unexposed * **PAR% :** PAR / incidence in all * Both are dependant on the **prevalence** of exposure in the population

What are the different biais seen in class for design study ?

1. selection biais 2. information biais (measurement) 3. confounding biais Not all observed associations are causal; they may result from errors in *_selection*_ of subjects, or in _*measurement*_ of exposure and outcome, or result from _*confounding_*.

What can cause information bias ?

* source/quality of data * imperfect measurements * investigator error * incorrect choice of exposure window * incorrect definition or timing of the outcome * prior knowledge of exposure or outcome status

What are the criteria for a confounder (exam question so you better learn it or you're gonna fail it)

1. Confounder is associated with **outcome** 2. confounder is associated with **exposure** 3. confounder is **not** in a **causal pathway** 1. *****aka not : cocaine causes rise in blood pressure causes stokes (rise in blood pressure wouldn't be a cofounding factor* ## Footnote **ALL 3 CRITERIA REQUIRED !**