Definitions Flashcards
A Cross Sectional study
measures a disease state in one point in time
Population/group
a group of people sharing a common feature – ethnicity, gender, disease
Ecological study
rather than studying individuals, it studies whole populations
Purposes of Epidemiology
Improve the health of the population by preventing disease.
Identifying differences in frequency – inequity
Determine the causes of the differences
Intervene to reduce difference – prevention or treatment
Disease frequency is also referred to as
disease occurrence/ risk/ rate
The definition of health/ disease by epidemiologists can be broad or
narrow
Where do epidemiologists start?
the living population
categorical data is
yes/ no
numerical data is
numbers of events
numerical data can be changed to
categorical data
P in PECOT
population is the top of the triangle, participants is the bottom = Participant population
Prevalence =>
is measuring the number of people in a disease state at one point in time
Incidence =>
is measuring the occurrences of disease over a specific time frame
Is prevalence or incidence given without a unit of time?
prevalence except the time it was measured it mentioned
When is prevalence expressed as a percentage?
when numerator is categorical
When is prevalence expressed as a mean?
when numerator is numerical
E =
exposure group
C =
comparison group
O=
outcomes
T=
time
examples of longitudinal study
cohort, RCT
cohort study aka
observational follow up study
incidence
measures occurrence of disease over time
when is it best to measure prevalence
when you want to find out IF something is happening or the thing you’re studying doesn’t have an easily identifiable starting point.
what is prevalence
the occurrence of disease in one point in time so T= 1
What type of study can you measure both prevalence and incidence
longitudinal
what type of study can you not measure incidence
cross sectional
Relative risk is
EGO/CGO
Risk difference is
EGO- CGO
which method of risk comparison has no context
relative risk
which risk comparison has units
RD
RD is sometimes called
absolute risk difference
RR is a ratio so it has no
units
Relative risk reduction is calculated when RR is less than
1
If RR is 0.67, RRR is
33%
Beware when u see a large RR but a
small RD
the true benefit of treatment can only be ascertained when you measure the
risk when you don’t treat
only good thing about RCT is that it
reduces confounding by making the same amount of confounding factors in both groups.
What is the main goal of epidemiology (IIDI) ?
Improve the health of populations by preventing disease
Identify differences in frequency- inequities
Determine the causes of differences
Intervene to reduce difference
What is it epidemiologists job to measure ?
the health and disease frequency of populations
Where can the numerator only come from
the denominator
What is age standardisation?
Age standardisation is converting the age structures of different populations being studied to fit a standard model.
Why age standardise?
It reduces confounding by age by comparing CG and EG groups of the same age and predicting age specific death rates
How do you age standardise?
First allocate people by age. find the age specific death rate in different age divisions of both populations being studied. Then convert the age structures to a standard age structure and predict the number of dead people/ population/ time.
crude death rate
total number of deaths / total population (without age standardising)
How to do overall age standardisation?
all expected age standardised death rates/ total standard population
what is the prevalence of disease in a population affected by?
the rate of incidence and the rate of people dying/ being cured.
if the numerator in prevalence is numerical it is expressed as a
mean
Study type when you observe a group for a period of time
cohort study/ observational follow up
what are EGO and CGO
measures of occurrence
incidence or prevalence involving almost always counting categorical disease events
incidence
incidence or prevalence involving measuring numerical disease states as well as categorical disease events
prevalence
what direction is the arrow for cohort study
down
the numerator of prevalence could be measured when?
right now and if it has occurred in a prior period
period prevalence is signified by arrow in what direction
backwards
What is the square grid representing
different disease outcomes
horizontal arrow represents
prevalence
when comparing disease rates, it is important to measure both CG and EG because
there may be a normal level of disease in both groups already before the exposure is affecting it
Relative risk is essentially a
ratio of occurences
Difference between occurrences is called
risk difference
risk difference is also called
absolute risk difference
base all decisions on
risk difference
If EGO=CGO relative risk is
1, meaning no effect
When relative risk is less than 1 you can calculate
Relative risk reduction
When relative risk is more than 1 you can calculate
Relative risk increase
all measures of risk or occurrence (incidence/prevalence) are
absolute
‘no effect’ value for risk difference is…. This is when CGO=EGO
0
The RD is an absolute risk reduction if the risk is
lower in the exposure group
The RD is an absolute risk increase if the risk is
higher in the exposure group
If the outcomes are calculated as a mean then the difference between two means is called
mean difference
Key design features of RCT
participants are allocated randomly to CG and EG group
Reasons you can’t do an RCT
Unethical due to randomising people to something that could cause them harm or randomising them to something that is known to save lives with the others not getting it.
When is it ethical to do an RCT
When you’re comparing the effectiveness of interventions that is already known not to cause harm. When you don’t know if it increases the risk of death but you know it decreases the risk of disease.
What are the four main causes of non random error in epidemiological studies?
poor study design, processes, measurement, and analyses
Non random errors are often called
systemic error / bias
Tell me what RAMBOMAN is and what is stands for
RAMBOMAN is used to identify where non random errors can occur. It stands for: Recruitment, Allocation, Maintenance, Blind or Objective Measurement and Analyses.
What does study validity depend on
the amount of NR and R error
Recruitment error is sometimes called…
external validity error
In what order from top to bottom is the P triangle layered
Setting, Eligibles, Participants
Two types of recruitment error
non response bias and non representative error
non representative error is what?
non representative error is a recruitment error where the participants recruited are non representative of the eligibles.
Selection bias makes the most impact on results when the aim of the study is more observational prevalence rather than testing interventions because
they want the results to be able to be applied to the populations or group they’re studying
define non response bias - when does it happen and what is the risks of it
Non response error could happen when only a small part of the eligible population ‘responds’ to participate in the study. If there are differences between the responders and non responders, this could mean that the results may not be representative of the eligible population.
Recruitment error can be combated by having: (3 levels)
the setting well described, participants representative of the eligible population and risk profile reported?
Main question of Recruitment error:
Can these results be applied to a particular group/ population
Main question of Allocation:
Were the study participants correctly and successfully allocated to the Exposure Group (EG) and the Comparison Group (CG)?
What are the two ways people can be allocated into EG and CG groups?
Randomly or by measurement
how participants are allocated to EG and CG has a big effect on the presence of ….- what is this?
confounders- they are other factors that, in addition to the exposure, can effect outcomes, therefore if there is an association between them
To be a confounder what is the conditions of this
they have to have a relationship, but also independently cause deaths so you can’t tell how many deaths were caused by the exposure.
In what study are participants randomly allocated to EG and CG groups
Random Controlled Trials
In what study are participants allocated to EG and CG by measurement
Observational study
What are allocation measurement errors
This is when the measurement of the exposure isn’t accurate so people end up being allocated in the wrong EG or CG group
How do you reduce allocation measurement errors?
well designed questionnaires, relying on biological measures rather than the memory/honesty of the person
What is baseline comparison and when should it be done?
baseline comparison is checking for differences in the EG and CG groups. This is done for especially small RCT studies as well as observational studies in case there is still differences in each group due to chance
How do you make sure allocation is done properly in large RCTS. Describe?
Make sure there is ‘concealment of allocation’. This is stopping doctors tampering with the random allocation of interventions process and giving interventions to patients they think is suitable.
Why is it important to collect information about other differences in the EG and CG groups?
this can help to adjust for them
The ‘M’ question is:
Were most of the participants maintained throughout the study in the groups (EG & CG) to which they were initially allocated?
When does maintenance error occur
when participants change their exposure status - ie by not maintaining their exposure status/ being exposed to other exposures that could skew the results
or
When participants are lost to follow up
BOM stands for … and is about
Blind or Objective Measurement is about the accuracy of the measurements of disease outcomes
The BOM question is
were the people doing the measurements unaware of the participants exposure status (blind) and were the measurements made objectively?
Making patients and the doctors blind (double blind) helps to reduce what error and why
maintenance error because people don’t know that they’re taking something useless
If a study isn’t blind what might happen
participants might have a bias on their outcome if their trying to test the effectiveness of something
Objective measurements help to reduce measurement errors by
making sure that patients responses are aligned to a single standard
give an example of an objective measure and a subjective measure
death, the amount of glucose in blood vs measures of pain or how you feel
define AN- analyses
Analyses is where known confounders were adjusted for. This can be stratified analysis
describe stratified analysis
this is like conducting sub studies by splitting the triangle into different circles and comparing EG CG groups from the same strata and then standardising the structures
age standardisation is a form of
Stratified analysis
why does comparing stratas help to reduce confounding
people with the same confounders will be compared to each other
There will always be confounding that will cause bias to happen but we need to think which
direction the bias will lie