All sections for Final Flashcards
Why use indirect adjustment?
- When the age (or other characteristic) specific mortality rates are not available (Such as in countries with poor vital records keeping (e.g. ages are not well recorded))
- Very small populations that could create greater error
- Also used for occupational exposure (For example: Do people who work in a certain industry, such as mining or construction, have a higher mortality than people of the same age in the general population?)
With direct adjustment we…
apply mortality rates of out populations of interest to a 3rd hypothetical population
in indirect adjustment we do the opposite of what?
Adding a 3rd population with direct adjustment. In indirect adjustment, we apply mortatlity rates from a real population (country, state, national) that our populations of interest (pop 1 and 2) reside within to the age structure of our populations of interest
Steps of indirect adjustment
- Apply 3rd person population age-specific rates to the age distribution to our two comparison populations and calculate the expected number of deaths if they had the same risk for death as the standard pop. Then sum expected number of deaths
- Calculate SMR (Standardized Mortality Ration) for population 1 (SMR = Observed deaths divided by expected deaths)
- Calculate the indirect adjusted rate by multiplying the SMR of pop 1 by observed deaths
- do the same for pop 2
- Compare the adjusted populations
Complete the Indirect Adjustment Example
on the ipad (1)
What is study design?
Research study design:
Set of methods and procedures used to collect and analyze data on variables in a specific research problem
Many types with own advantages and disadvantages
How is the study design type detemined
Nature of research question
Goal of research
Availability of resources
Study Design Observational
Researchers observe individuals and record information about variable of interest.
No manipulation of events
The purpose is to describe some group or situation
Study Design Experimental
Researchers intentionally impose treatments on individuals and measure their responses
Determine ‘cause and effect’ relationship
Determine whether treatment leads to positive effect
Types of observational studies
Individual based
- cohort (longitudinal)
- case-control
- cross-sectional
Population based
- ecologucal
Type of experimental study
randomized controlled trial (RCT)
Evidence strength highest to lowest
- Systematic review/meta-analysis
- RCT (randomized controlled trail)
- cohort
- case control
- cross sectional
- case reports and series
- Ideas, opinions, editorials
- animal research
- invitro “test tube” research
What is prevalence data collection
Available data sources
- Birth records, death records, etc.
- Through studies
What is a cross sectional study design?
- prevalence or transverse study
- Helps us to investigate causes of disease
- Analyzes sample population at one point in time: (Collect data, but do NOT intervene in any way. Exposure and outcomes measured simultaneously)
- Gives us a “snapshot” of the population (
—–we CAN identify prevalent cases: we know they EXISTED at the time of the study
—–we CANNOT identify incident cases: we do NOT know WHEN the condition started
What are exposures
Broadly applied to any factor that is the primary explanatory variable of interest.
- Independent variable
- The cause in the cause and effect relationship
Also applied to other variables such as confounders which may also have to be addressed.
What is an outcome
Disease, state of health, or health related event or death that we think the exposure impacts in some way
- Dependent variable
- Its value or presence depends on another variable
- The effect in the cause-and-effect relationship
Example of exposure and outcome: Does exposure increase the risk of lung cancer?
Exposure: smoking
Outcome: lung cancer
Example of exposure and outcome: Is caffeine consumption during pregnancy associated with an increased risk of low brith weight?
Exposure: caffeine consumption
Outcome: low birth weight
How to determine is there is an association in a cross-sectional study
- Calculate the prevalence of the disease/outcome
- Calculate the prevalence of exposure/risk factor
How to compare prevalence of disease
- who has been exposed with who does and does not have outcome
- who has NOT been exposed with who does and does not have outcome
A/(A+B) vs. C/(C+D)
EX: Smokers 72.5% and Non Smokers 35.8%. SUBTRACT these to get “Smokers have a 36.7% higher prevalence of lung cancer than do non smokers”
Should prevalence be a decimal, whole number, or percentage
Percentage
How to compare prevalence of EXPOSURE
- disease outcome with who has and has not been exposed
- no disease outcome with who has and has not been exposed
A/(A+B) vs. C/(C+D)
EX: Lung cancer 68.8% and no lung cancer 31.8%. SUBTRACT these to get “Among people who have lung cancer, there was a 37% higher prevalence of smoking than those without lung cancer”
What are some issues with cross-sectional studies?
The prevalent cases in the study may not be generalizable - They don’t represent all people with that disease/illness
Identifying prevalent cases could exclude people who already died from the disease - Because both exposure and outcome are determined at the same time it is hard to determine the temporal relationship
what are some benefits of cross-sectional studies
Can be very suggestive of a possible exposures or risk factors for disease - After cross-sectional the next step would be to determine if there is a temporal causal step (This requires a case-control study design)
Cross-sectional studies and causality
Cross-sectional studies are not enough to determine causality.
Only mathematical associations
Case-Control designs are the next step in order to determine causality. - They themselves do not conclusively determine causality.
Cross-sectional studies and causality
Cross-sectional studies are not enough to determine causality.
Only mathematical associations
Case-Control designs are the next step in order to determine causality. - They themselves do not conclusively determine causality.
Advantages of cross-sectional studies
Relatively feasible and not too time-consuming
Relatively inexpensive to conduct
Helps to estimate burden of diseases within a population
- Useful for health planning
We can study multiple outcomes/exposures at once
- Demographics of the people they occurs in/does not occur in
- Conditions in which the health outcomes/exposures occurs
- What exposures are near the outcome at the time that the snapshot was taken.
Prompts further research
Disadvantages of cross-sectional studies
Can establish an association, but NOT causation
- Temporal relationship between exposure and outcome may be difficult to establish
VERY susceptible to bias
- Impossible to ensure that confounders are distributed equally among the groups
- Often either exposure or outcome or both depend upon recall, which is fallible
- Groups could have different sample sizes, resulting in loss of statistical efficiency
Weaknesses of using aggregated data.
- Ecological Fallacy: Ecological inference fallacy or Population fallacy
- EX: falsely assumes that every individual’s IQ is high, since the class average IQ is high
Occurs when inferences about the nature of individuals are deduced from inferences about the group to which those individuals belong.
Is a cross-sectional study design an observational or experimental study type?
Observational – there is no manipulation of events/variables.
What are the advantages of cross-sectional study design?
- Can study multiple outcomes/exposures
- Inexpensive, feasibility, not too time-consuming
- Helps to estimate burden of disease
- Prompts further research
What are the disadvantages of cross-sectional study design?
- Cannot determine causality
- May not be generalizable
- Susceptible to other bias (confounders, recall, sample size)
What does a case control study examine?
the possible relation of an exposure to a certain disease
- We identify a group of individuals with that disease or outcome of interest called (cases)
- For comparison, a group of people without that disease or outcome of interest (controls)
Why do we collect data from case-control groups
to determine the proportion of exposure in each group
if there is an association between the exposure and disease, which groups prevalence should be higher
the prevalence of exposure should be higher in cases (disease) than in controls (no disease).
what are the steps in a case-control study?
- definition of cases
- selection of cases
- selection of controls
- ascertainment of exposure status
- analysis
- conclusion and interpretation
defining cases (case-control): How will you define who is included
Be specific
- what type of ling cancer?
- level of severity
- how long have they been diagnoses
- clinical characteristics
cases should be representative of all disabled
Correctly defining cases (outcome) helps us to select the cases and limit bias.
selecting cases (case-control)
Selecting our cases is important because it can introduce bias.
What is bias?
“systematic error introduced in to sampling or testing by selecting or encouraging one outcome or answer over another.”
Lots of different types
Bias is not always purposeful
What is selection bias?
selection of individuals, groups, or data for analysis in such a way that the sample is not representative of the population intended to be analyzed.
sources
- hospital patients, physicians practice patients, clinic patients, disease registries
what are the risks of selecting patients from hospitals?
- some tertiary and quarternary care facilities selectively admit severely ill patients
- Risk factors only in individuals with severe forms of the disease
- Not generalizable to all patients with the disease
if used would need to select from several hospitals in the community
primary, secondary, tertiary, quaternary
- Quarternary (experimental medicine and uncommon specializations)
- Tertiary (hospitalization and higher specialization like neurosurgery)
- Secondary (specialists like cardiologists, oncologists, ect.)
- Primary (essentials of screening, illness, injury, referral
what is the issue with using incident cases
- Must often wait for new cases to be diagnosed
- Prevalent cases have already been diagnosed and represent a larger number of cases available for the study
Tub example
- prevalence is in the tub
- mortality (death) is the drain
- recovery is evaporation
- incidence is the faucet
what is the issue with using prevelant cases
Risk factors that we identify in a study using prevalent cases may be related more to survival with the disease rather than to the development of the disease
Such a study is more likely to include longer-term survivors.
Non-representative group of cases
selecting controls (case-control)
One of the most difficult problems in epidemiology.
If we find more exposure in the cases than in the controls
- Would like to conclude that there is an association between the exposure and the disease.
The way controls are chosen is a major determinant of whether this conclusion is valid.
Fundamental conceptual issue:
- Should controls be similar to the cases in all respects other than having the disease in question?
- OR, should they be representative of all persons without the disease in the population from which the cases are selected.
selection of controls: non-hospital
School rosters
Birth records
Selective service lists
Insurance company lists
Residents of a defined areas (neighborhoods)
Random digit dialing
- Why is this a problem?
what is a best friend control?
A case is asked to provide the name for a friend who may be more likely to participate in a study
- Friends tend to be similar in age and sociodemographic variables
- Spouses and siblings
May be too similar, including the variables under investigation.
Hospital selection of controls
Hospital inpatients are a ‘captive population’ and are clearly identified.
Sick people go to hospitals
- Many of the risk factors that got them there may also be a risk factor for whatever disease is being studied.
- For example, cigarette smoking
Selecting from hospitals may mean that controls do not represent the general population.
If we use hospital controls, which do we use?
- Use a sample of all other admitted patients?
Or select patients with a specific ‘other’ diagnosis?
- Unlikely to be representative of the reference population.
- Perhaps have a lower-than-expected prevalence of the exposure of interest.
Hospital selection of controls
Hospital inpatients are a ‘captive population’ and are clearly identified.
Sick people go to hospitals
- Many of the risk factors that got them there may also be a risk factor for whatever disease is being studied.
- For example, cigarette smoking
Selecting from hospitals may mean that controls do not represent the general population.
If we use hospital controls, which do we use?
- Use a sample of all other admitted patients?
Or select patients with a specific ‘other’ diagnosis?
- Unlikely to be representative of the reference population.
- Perhaps have a lower-than-expected prevalence of the exposure of interest.
What is matching in case control?
Major concern is that cases and controls may differ in characteristics or exposures other than the one that has been targeted for study.
- Is an observed association due to factors other than the exposure being studied? Confounding
One approach is to match the cases and controls so that they are similar to the cases in certain characteristics.
Matching may be of two types:
- Group matching
- Individual matching
Match only on variables that we are convinced are risk factors for the disease and are not interested in investigating in this study
Matching variables other than these is called overmatching.
What is group or frequency matching?
Selecting the controls so that they have the proportion of a characteristic identical to that in the case group.
Requires that all the cases are selected first, so that proportions can be calculated before selecting controls.
what is individual matching or matching pairs
For each individual case selected for the study:
- A control is selected who is similar to the case in terms of the specific variable(s) of concern.
Often used for hospital controls.
What are some problems with matching
If we want to match cases on too many characteristics, it may be too difficult to find appropriate controls.
Gender
Age
Race
Education
Income
Geography
Once we have matched controls to cases according to a given characteristic, we cannot study that characteristic.
Why?
- Matching creates an artificially identical proportion of a factor in the control group.
- Because case-control studies looks at how a much exposure occurs naturally in a population given an outcome:
- We cannot look at a factor whose proportion has been artificially established.
ascertainment/measurement of exposure
Set criteria for exposure
- Understand the methods of exposure
Collect information:
- Observation
- Interviews
- Questionnaire
- Examination of records
Information bias: limitation of recall
Data is collected from subjects by interviews/questionnaires
- All humans are limited in their ability to recall information
People may also not have the information being requested
- For example, circumcision
If limitation of recall regarding exposure affects all subjects in a study to the same extent (both cases and controls).
- Causes misclassification of exposure:
- Cases or controls who were exposed will be classified as unexposed
- Leading to an underestimate of the true risk of a disease associated with an exposure
More serious than information recall
- Social desirability
- “Skin in the game”
Differential recall of important exposures between cases and controls
- Can artificially suggest a relationship between exposure and outcome that is not there or is not as strong
What are dichotomous exposures
Sometimes the exposure that you are investigating is not dichotomous
Ie. You cannot have it or not have it.
However, sometimes you have an exposure for which you legitimately cannot have a zero amount of.
i.e. Blood pressure, weight, height, blood sugar
Dichotomous:
Have blood pressure vs. have no blood pressure
Can we instead think of the exposure as having multiple levels/categories?
Using reference level and categories in non-dichotomous exposures
We choose one of the levels/categories to serve as the ‘unexposed status’
We call the group that we consider ‘unexposed’ the ‘reference’ group/category/level
We compare this group to those we consider to be ‘exposed’
The other levels/categories serve as ‘exposure status’
Each level is a different ‘exposure’
For many health issues, this would be the expected normal/healthy value:
The reference is not the reference because it has the most people in it, though this may often be the case
We can also arbitrarily pick a reference category for characteristics that have no ‘normal/healthy’
Gender
Race
Education
SES
What is statistical power
Statistical power
The probability that a statistical test will say there are no significant differences when there really are none
Power is typically gained when you have a ratio of at least 1 case to 4 controls
Investigators can determine how many controls will be used per case in a case-control study.
Multiple controls are frequently used
Why do we use multiple controls?
For rare diseases, there may be a limit to the number of possible cases.
Because the number of cases cannot be increased without extending the length of the study a different option is to increase the number of controls.
Either:
Controls of the same type
Controls of different types
Multiple Controls of the Same Type
All controls come from the same type of source
Registry OR Hospitals OR Clinics
Can be multiple registries, multiple hospitals etc.
Multiple Controls of Different Types
Suppose we’re concerned that our controls are not representative of the rate of exposure that is expected in the population of non-diseased persons.
i.e. the controls may be highly selected subset of non-diseased individuals and may have a different exposure
To address this problem, we may choose to use an additional control group
Like neighborhood controls
Advantages of case-control studies
- Can be used to study low-prevalence conditions
- Relatively quick and easy to complete
- Usually inexpensive
- Involve smaller number of subjects
- Multiple risk factors can be assessed at one time
Disadvantages of case-control studies
- Measurement of exposure may be inaccurate
- Representativeness of cases and controls may be unknown
- Only provides indirect estimates of risk
- Data quality may be a factor (information bias)
- The temporal relationship between exposure factor and outcome cannot always be ascertained.
Setting up analysis for case control study
We’ve selected all our cases and controls and we ascertained their exposure.
If exposure is dichotomous (exposure has occurred, or not occurred), break down into four groups is possible:
A cases who were exposed
B controls who were exposed
C cases who were not exposed
D controls who were not exposed
If exposure is associated with disease:
The proportion of cases who were exposed should be greater than the proportion of the controls who were exposed.
Odds ratio/relative odds in CC studies
In a case-control study:
Do not know the incidence in the exposed population or the incidence in the non-exposed population
Because we start with the diseased people (cases) and non-diseased people (controls).
Because of this, we cannot say that someone has a risk of developing a disease given an exposure.
However, we can calculate another measure called the odds ratio.
Under certain conditions, the odds ratio can serve as an estimate of risk.
What is an odds ratio
Compares the odds of exposure in the cases to the odds of exposure in the controls
How are odds ratios versatile measures
Calculate them from data from
- Cross-sectional studies
- Cohort studies
- Case-control studies
How to calculate odds ratio in cc study
- We calculate the odds of exposure in cases
# of cases with exposure / # of cases without exposure
A/C - We then calculate the odds of exposure in the controls
# of controls with the exposure / # of controls without the exposure
B/D - Calculate the odds
Odds of exposure in the cases/ odds of exposure in the controls
(a/c) / (b/d)
How to interpret odds ratio
OR = 1
- No difference in the odds of the exposure, given outcome status
- Those with the outcome have the same amount of exposure than those without
- Those with (insert the case outcome) had the same odds of (insert the exposure) than those without (insert the outcome).
- There was no relationship between (insert the outcome) and (insert the exposure).
OR > 1
- There are higher odds of exposure, given outcome status
- Those with the outcome have more of the exposure than those without
- Exposure is a risk factor for the disease
- Those with (insert the case outcome) had (insert OR) higher odds of (insert the exposure) than those without (insert the outcome).
OR < 1
- There is lower odds of exposure, given outcome status
- Those with the outcome have less of the exposure than those without
- Exposure is a protective factor for the disease
- Subtract OR from 1 to indicate how much less the odds are.
- Those with (insert the case condition) had (1 – OR) less/lower odds of (insert the exposure) than those without (insert the case condition).
What can odds ratios do for us?
- Are used to determine whether an exposure is related to a disease
- Used to compare the magnitude of different risk factors
What is a cohort study?
- The investigator selects a group of exposed individuals and a group of non-exposed individuals
- Cannot have the outcome of interest at the start of the study
- Follows them over time to see who in each of the groups develops the disease.
- Compares the two groups based on exposure.
How to design a cohort study
Typically involves two groups (exposed and unexposed) but can have multiple exposure groups of varying levels.
- i.e. blood pressure categories, types of diets, amount of smoking.
- Need to pick a ‘reference’ group to serve as the ‘unexposed’ group.
Because we are identifying NEW (incident cases) of the disease we can determine whether a temporal relationship exists between the exposure an disease.
- Absolutely necessary for determining causality in a disease.
What does a cohort study look for
incidence of disease
example of cohort study
We want to look at whether a heart murmur (exposure) is a causal factor in atrial fibrillation (outcome).
We find a group of individuals (4500) and determine which of them have a significant heart murmur and which of them do not.
We follow them for 30 years.
Of those with a heart murmur (exposed), 102 developed atrial fibrillation.
Of those without the heart murmur (unexposed), 76 developed atrial fibrillation.
What is the incidence of atrial fibrillation among those with heart murmur (exposed)?
a/a+b
102/ 900
0.11333 x 10n
Incidence of atrial fibrillation in exposed persons is 113.3 per 1000 people
What is the incidence of atrial fibrillation among those without the heart murmur (unexposed)?
c / c+d
76 / 3,600
0.021111 x 10n
Incidence of atrial fibrillation in unexposed persons is 21.1 per 1000 people
What is the incidence of atrial fibrillation among the entire population?
a+c/ a+b+c+d
178 / 4500
0.039555 x 10n
Incidence of atrial fibrillation in entire population is 39.6 per 1000 people
Selection of Study Populations: Option 1
Option 1: We create a study population by selecting groups for inclusion in the study on the basis of whether or not they were exposed.
Example: find a group of people with a specific occupational exposure to a chemical we think leads to CVD and equal numbers of people with an occupational exposure that did not use that specific chemical.
Selection of Study Populations: Option 2
Option 2: We can select a defined population before any of its members become exposed or before their exposures are identified.
- We select a population on the basis of some factor not related to exposure (such as community of residence)
- Take histories of or perform tests on the entire population.
- Using the results of the histories or the tests, one can separate the population into exposed and nonexposed groups.
How to select populations for cohort studies
Cohort studies often require a long follow-up period, lasting until enough events (outcomes) have occurred.
- Length of follow-up in the second option for study population selection is even greater.
What is a problem with cohort study design?
The population often must be followed-up for a long period to determine whether the outcome of interest has developed.
The study takes a very long time to complete.
Funding from grants typically only lasts between 2-5 years
What is a problem with cohort study design?
The population often must be followed-up for a long period to determine whether the outcome of interest has developed.
The study takes a very long time to complete.
Funding from grants typically only lasts between 2-5 years
What is a prospective cohort study?
Also called ‘concurrent cohort’ or ‘longitudinal study’
Investigator identifies the original population at the beginning of the study.
Exposure and non-exposure are ascertained during the study
Groups are then followed (concurrently/prospectively) into the future, and incidence is measured.
Prospective means the data is collected as it happens
What is a retrospective cohort study?
Also called ‘historical cohort’ or ‘nonconcurrent’ study
- Cohorts are defined from a previous point in time and are not followed into the future.
- Exposure status is determined from information or data that was collected in the past (i.e. past clinical records)
- The outcome of interest is determined by the researchers in the present
Designs for both prospective cohort study and retrospective or historical cohort study are identical:
We are comparing exposed and non-exposed populations.
Only difference is ‘calendar time’.
Can have a combination:
Exposure is ascertained from objective records in the past, and follow-up and measurement of outcome continue into the future.
Student records about smoking and alcohol from the past
Follow them until they graduate school
Which ever study design, the participant cannot have the outcome of interest at the time that exposure status is determined.
What is the difference between prospective and retrospective cohort studies?
There are two main types of cohort studies, prospective and retrospective.
Combination of the two
The main difference between these types:
When researchers are collecting information about exposure/outcome in relationship to the present time period.
Framingham Study
Prospective
One of the most important and best-known prospective cohort studies is the Framingham Study of cardiovascular disease:
- Begun in 1948
- Framingham, MA
- 30,000 lived there in 1948
- Planned to follow for 20 years
- Residents were eligible if they were 30-62 years of age.
Took a sample of 5,127 men and women, who did not have heart disease at the time the study began.
Exposures: smoking, obesity, high blood pressure, elevated cholesterol, low levels of physical activity, and others.
New coronary events were identified by:
Examining the study population every two years
Daily surveillance of hospitalizations at the only hospital in Framingham.
This study used the second option of selecting a cohort population:
A defined population was selected on the basis of location of residence or other factors not related to the exposures in question.
The population was then observed over time to determine which individuals developed or already had the ‘exposures’ of interest.
Doing so provided an important advantage:
Permitted the investigators to study multiple exposures at one time
Complex interactions among those exposures by using advanced statistical procedures.
Hypothesis of Framingham study
Tested the following hypotheses:
Incidence of CHD increases with age
It occurs earlier and more frequently in males
Persons with hypertension develop CHD at a greater rate than those who are ‘normotensive’
Elevated blood cholesterol is associated with increased risk of CHD
Tobacco smoking and habitual use of alcohol are associated with increased risk of CHD
Increased physical activity is associated with decrease in development of CHD
An increase in body weight predisposes a person to the development of CHD
An increased rate of development of CHD occurs in patients with diabetes
All of the hypotheses seem like common knowledge, right?
They are only common knowledge BECAUSE of the Framingham study
The study is the origin of the word risk factor.
How does the framingham study work continue
- Genetic research
- Looking at how social networks affect health and prevention
Cowan breast cancer and progesterone deficiency study
In 1978, Cowan identified a population of women who had been patients at Johns Hopkins Infertility Clinic from 1945-1965
- All had a late first pregnancy
- During infertility evaluation, detailed hormonal profiles were taken.
- Patients were divided into exposure groups based on their past medical records (hormonal abnormality vs. normal hormones).
- Both groups were followed for subsequent development of breast cancer
Results showed that women who had hormonal abnormalities had 5.4 times greater risk of developing premenopausal breast cancer.
This study:
- Was a cohort because it compared exposed and unexposed persons
- And was retrospective because an investigator in 1978 looked at records from 1945-1965 to determine exposure status.
What is non-participation bias
Selection bias in cohort studies
Were the people who participated in a cohort study different from those who didn’t?
Does this difference also impact the likelihood of being exposed or unexposed?
What is non-response bias
Selection bias in cohort studies
Sometimes study individuals are unwilling to participate in all of the study procedures:
- Surveys/Questions
- Clinical evaluation
This is an issue if those who responded and those who didn’t, are different in meaningful ways that effect exposure or outcome.
- Sexual behavior questions and STI outcomes.
What is loss to follow-up (attrition)
Selection bias in cohort studies
Participants who at one time were actively participating in the study, but have become ‘lost’ at the time of follow-up.
- If those who are lost to follow up differ from those not lost to follow-up
- The incidence rates calculated in the exposed and non-exposed group will be difficult to interpret.
ex: TOP Project - students move schools or leave state
Information bias in cohort studies
Quality and extent of information obtained is different for exposed persons than for unexposed person’s.
- Particularly likely to occur in retrospective cohort studies, in which information is obtained from past records.
Quality of information is an issue….and is understandable to a degree.
- However, there should be no difference in information between exposed and unexposed individuals.
Investigator Information Bias in Cohort Studies
If the person who decides whether disease has developed in a subject also knows that the subject was exposed:
- If that person is aware of the hypothesis being tested
- That person’s judgement as to whether the disease developed may be biased by that knowledge.
The problem may be addressed by using different people to make the disease assessment and determine whether this person was exposed.
Epidemiologists and statisticians who are analyzing data my have strong preconceptions.
- They may unintentionally introduce bias into their analyses and their interpretation of the study findings.
when do we use cohort studies?
We must have some idea of associations of exposure to disease….so that we know what exposures to track.
- Indicated when good evidence suggests an association of a disease with certain exposures.
- Perhaps after a case-control studies
Because of the long follow-up period, the cohort study is a good approach if we can minimize attrition (loss to follow-up)
- Easier to do when the interval between exposure and development of disease is shorter.
what is the Purpose of Epidemiological Studies
Determine whether there is a an excess or reduced risk of a certain disease in association with a certain exposure or characteristic.
what is risk?
Risk is defined as the probability of an event (such as developing a disease) occurring.
There are several different measures of risk:
- Absolute risk
- Relative risk
- Attributable risk
- Odds ratio
What is absolute risk
Absolute risk:
- The incidence of a disease in a whole population.
- The number of people experiencing an event in relation to the population at risk
Absolute risk can indicate:
- Magnitude of the risk in a group of people with all exposures
Absolute Risk
- Because it does not take into consideration the different risks of disease in unexposed individuals vs. exposed individuals: we Cannot indicate whether the exposure is associated with an increased risk of disease.
However, comparison of risk is fundamental to epidemiology.
- We do not all have equal risk
- Making prevention ineffective without this knowledge
why is absolute risk important?
Absolute risk has important implications in both clinical and public health policy.
- Though absolute risk does not stipulate any explicit comparison, an implicit comparison is often made whenever we look at the incidence of disease.
However, to address the question of association:
- We must use approaches that involve explicit comparisons.
How do we determine whether a certain disease is associated with a certain exposure?
We must determine using data obtained in case-control and cohort studies:
- Is there excess risk of the disease in persons who have been exposed to a certain agent?
Remember foodborne disease
what can relative risk also be defined as?
probability of an event occurring in exposed compared to probability of the event in unexposed
how to interpret relative risk
= 1 : There was no relationship between the (insert the exposure) and the (insert the outcome).
Those with (insert the exposure) had the same risk of (insert the outcome) as those without (insert the exposure).
greater than 1 : Those with (insert the exposure) had (insert RR number) times the risk/increased risk of (insert the outcome) compared to those without (insert the exposure).
Less than 1: Those with (insert the exposure) had (1 – RR number) less risk/decreased risk of (insert the outcome) compared to those without (insert the exposure).
what is relative risk important?
measure of the strength of an association
establishes causal relationships
why is attributable risk important
to know how much of the disease that occurs, can be attributed to a certain exposure
what is background risk?
Incidence not due to exposure
there are multiple factors for multiple diseases
everyone shares the background risk, regardless of wether or not they have the exposure in question
if we want to kjnow how much of the total risk in exposed person is due to the exposure,
we should subtract the background risk from the total risk
what is a RCT
randomized control trial
evaluated both the effectiveness and the side effects of new forms of intervention
what is a RCT
randomized control trial
evaluated both the effectiveness and the side effects of new forms of intervention
what are the three observational study types
what is one experimental study type
observational
1. cross-sectional
2. case-control
3. cohort
experimental
1. RCT
what are the three observational study types
what is one experimental study type
observational
1. cross-sectional
2. case-control
3. cohort
experimental
1. RCT
purposes of RCT
evaluate new drug/tech
assess screening program
assess intervention program
evaluate new ways to deliver health services
history of RCT
- Ambroise Pare (1510-1590)
- unplanned trail with cauterization and topical treatment - James Link (1716-1794)
- Scottish naval physician
- planned clinical trial for scurvy
- advanced preventative med and nutrition - Austin Bradford Hill (1948)
- first published RCT
how to design an rct
- define population
- Randomize treatment and no treatment
- follow subjects
how to design an rct
- define population
- Randomize treatment and no treatment
- follow subjects
how to select subject
inclusion and exclusion criteria
must be extremely specific
why is subject selection important
should be no element of subjective decision-making on the part of the investigator in deciding who is included or not included in the study.
Test of adequacy:
If we have spelled out our criteria in writing, and someone not involved in the study walks in off the street and applies our criteria to the same population, will that person select the same subjects whom we would have selected?
why cant you match treatment and control members?
- can only match on variables that are measurable
why cant you match treatment and control members?
- can only match on variables that are measurable
stratified randomization
divide or stratifies the population into groups that differ in important ways
select random sample from within each group
basis for grouping must be known before sampling
to address the issue of randomized groups being weighed inefficiently
what is crossover RCT
Participants receive each treatment in a random order
they serve as their own controls
used when researchers think it will be difficult to find participants
what is crossover RCT
Participants receive each treatment in a random order
they serve as their own controls
used when researchers think it will be difficult to find participants
what are the advantages of a crossover study
- influence of confounders are reduced because each participant serves as own control
-statistically efficient, requiring fewer participants
what are the advantages of a crossover study
- influence of confounders are reduced because each participant serves as own control
-statistically efficient, requiring fewer participants
disadvantages of crossover study
- possible order treatment is administered may affect outcome
- carry-over effect, combated with wash-out period
- cant make a group unlearn a skill to act as a control in later phase
purpose of factorial RCT
tested aspirin for primary prevention of cardio disease
tested beta-carotene for primary prevention of cancer
- study pop split into aspirin and placebo
- aspirin was later split into B-C and placebo
- placebo was later split into B-C and placebo
result could be both, only Aspirin, only BC, or neither
