Flashcards in Part 1: Epidemiology, GATE, Errors Deck (128):
What is epidemiology?
Epidemiology is the study/science of dis-ease occurrence in a population/group over a period of time or at a point in time including the factors which determine the differences of the occurrence of dis-ease between groups.
What is occurrence?
Occurrence is the transition between being a non dis-eased state to being a dis-eased state.
Describe the hourglass analogy.
- Sand, being individuals in a population
- The middle of the hourglass being the (OCCURRENCE) transition from being non-dis-eased to the being in a state of dis-ease.
- The top glass is the non dis-eased state
- The bottom is the dis-eased state
Dis-ease occurrences is referring to the idea of....
Quantifying an event/or the outcome.
Why is important to measure the occurrence of an outcome?
It will help identify the causes of dis-eases and therefore help with fixing or preventing future case.
In what ways can measuring the outcome help? (3)
1. Can inform health planning and promotion
2. Prevention of dis-eases
3. Treatment decisions (like which treatments actually work)
What does the term population mean?
It is a group of people who share similar/common characteristic(s) or factor(s)
What's the things you say which helps you to find the common factors which could be included in a population?
"Men and women aged 18 and above who live on the North Shore who are non-smokers test a new drug to see if lung cancer can be reduced in 2018"
So what are the factors which can may be common in a population?
4. Same treatment/drug
5. Time period
6. Combo of the above
What is the difference of the focus target of clinical medicine and population health?
- TREAT the INDIVIDUAL
(How can I fix the individual? What is wrong with them?)
THE AMBULANCE AT THE BOTTOM OF CLIFF
- Look as population as a whole
(Looking at trends through epidemiological studies)
THE FENCE AT THE TOP OF THE CLIFF
What is the difference of the education background of clinical medicine and population health? What models do they use?
- Cure rather than prevention $$$$$$$$$$$$
- Biomedical model
(Individual focused, may be victim blaming, what is the individual doing to make them injured like obese?)
- Prevention before occurrence
- Social model
(If the individual is obese/injured, what is his living environment like? Are there lots of McD's around his home? Is it safe to walk around at night?)
What is the difference of the rights of clinical medicine and population health?
- Individual rights (Autonomy, patient's rights)
- Human rights
- Social and environmental justice
What are the two ways of data collection?
What is numerical data collection?
Data collection when the data you are collecting are numerical values.
ONE POINT IN TIME collected but can be displayed in GATE frame as CATEGORICAL too.
What are some examples of numerical data collection?
BMI, BP, BGL, No, of hospital visits. No. of births
NOTE: Numerical measures can be classed as Categorical sometimes.
When taking pulses, you have have fast pulse (>70 bpm) or slow pulse (<70 bpm)
What is categorical collection?
It is for easier observable onsets, can be characterised by having it or not.
YES or NO
Can be prevalence or incidence
What are some examples of categorical data collection?
Male or Female
Smoker or Non-smoker
Lung cancer or No Lung cancer
Dead or Alive
What is the basic equation of epidemiology?
E = N / D / T
Describe the GATE frame.
Triangle - population
Circle - exposure and Comparison group
Square - outcome
Arrow - time
Describe the triangle in GATE frame w subtypes and add examples:
Population - Triangle gets smaller (you filter them out to the eventual participants,) three sub types:
- Study setting
The New Zealand Population
- Secondary High School Students (Maori and Pakeha)
- Actual participants
9500 Yr 11, 12, and 13 Maori and Pakeha students.
What can be in the GATE frame sometimes for extra marks?
Allocation rectangle between the triangle and the circle
What can allocation be? (2)
- Measurement (inclu. surveys and questionnaires
Describe the circle in GATE frame w subtypes and add examples:
1/2 circle: Exposure - EG, these are the people exposed of the exposure, like smoking. So these would be the smokers.
THIS IS D for EXPOSURE GROUP.
1/2 circle: Comparison - CG, these are the comparison group. Who are not exposed to the exposure, this would be the non-smokers.
THIS IS D for COMPARISON GROUP.
Describe the square in GATE frame w subtypes and add examples:
Outcome - these are the numerators in respect of the EG and CG directly above it.
Outcomes can be numerical or categorical
How many compare groups can you have in a GATE frame?
There is ALWAYS ONE CG!!! If you have more than one, you wrong…
How do you do the maths in numerical outcomes?
You calculate the averages of the results and chuck em' in the square
Give an example of numerical outcomes.
EG is walking
CG is not walking
Because you allocate through measurement, (BMI) you calculate the averages of the BMI of each group and put in the average in the square outcome.
What is the equation for averaging for EGO and CGO?
EGO = Σa/EG
CGO = Σb/CG.
Describe the arrow in GATE frame w subtypes and add examples:
Time - this is the time when or during the outcomes are measured.
“CUT, NOW!” like a prevalence measure.
“Down the line” like an incidence measure.
What are the epidemiological measures of dis-ease occurrence?
What is incidence?
Over a period of time
Counting transition from non-dis-eased to dis-eased, easily observable characteristics
How do you remember incidence?
Imagine constantly working (overtime like 10yrs, you tired af!!!)
and imagine people popping up (you count one by one) over time and you mark a tally when they pop up,
you count FORWARDS (vertical arrow, down the line)
What is prevalence?
ONE POINT IN TIME, STATE
not easily observable characteristics like obesity when was the person obese?
Hard to pinpoint, GRADUAL.
What are the sub-types of prevalence?
How do you remember period prevalence?
“Tell me the answer now, but do you remember having lung cancer in the last 10 years?” Recalling the past. It is a mix of incidence and prevalence.
IT GIVES A TIME FRAME OF WHEN TO REMEMBER
How do you remember point prevalence?
“Tell me the answer now! But do you smoke? Do you have lung cancer?” GATE frame fill out is quick and easy.
So what is point prevalence?
Counting people in one point in time, measure OUTCOME and EXPOSURE same time.
So what is period prevalence?
Ask now, but asking them to recall back over the period.
Describe the population cloud analogy.
What is population cloud?
Population cloud is all the people (the population) which you will be trying to investigate.
What is incidence rain?
Calculated by the number of onsets (events) occurring in a period of time.
Raindrops falling illustrates the population of people which are changing state, falling into the prevalence pool.
This is when it is EASILY OBSERVABLE: CATEGORICAL
What is incidence drizzle?
NOT EASILY observable events which means that it is more difficult to count and therefore NUMERICAL and is better to count the prevalence pool rather than the incidence drizzle.
What is prevalence pool?
SNAPSHOT IN TIME
How many people are diseased? is the alternative of counting dis-ease occurrence to incidence and is counting the number of people who are now dis-eased,
AT ONE POINT IN TIME (Numerator) and then dividing by the number of people in the study at that point in time.
What are prevalence equations?
Prevalence in EG = a / EG at one point in time
Prevalence in CG = b / CG at one point in time
Population with high incidence of dis-ease could have a low prevalence if the death rate or cure rate is also high.
True or False?
A population with low prevalence if dis-ease could have a high prevalence of dis-ease, if almost no one dies of the the dis-ease in a population at a point in time.
True or False?
What did the two statements mean then?
Prevalence is a DIRTYYYY measure
Why is prevalence a dirty measure? (3)
Affected by cure cloud, death drips, and incidence rain.
This it is less useful than incidence because a high incidence of dis-ease could result in either a high or low prevalence depending on the death rate and the cure rate.
ALSO prevalence does not include the time. It is excluded so it takes away a factor.
What is epidemiology all about?
To compare EGO and CGO, comparing dis-ease occurrences in different populations
Why is measuring CG and EG important?
We need CG because if we only measure EG we are taking away all the other factors which could have affected the exposure.
CG will not always be dis-ease free. If we take the example of drive deaths non smokers (CG) other factors could have affected deaths not j the non smoking.
What is a factor of misinterpretation?
What are the two ways of comparing dis-ease occurrence?
What is the equation for relative risk?
What is the no effect line for RR?
What is the equation for RD?
EGO - CGO
What is the no effect line in RD?
If the no effect line is 0, what can the presentation of data in CI be?
EGO, CGO, RD
If the no effect line is 1, what can the presentation of data in CI be?
What does RR>1 mean?
Relative risk increase, usually a %
What is the equation for RRI?
RRI = RR - 1 x 100%
What does RR<1 mean?
Relative risk reduction =, it is below the no effect line is a %
What is the equation for RR<1?
RRR = 1 - RR x 100%
What is the RD equation?
EGO - CGO
Why is the no effect line for RD = 0?
Because if EGO and CGO were the same, it will equal to 0, there will be no effect
Why is the no effect line for RR=1?
Because if EGO and CGO were the same, the value will be 1/1 = 1
Are there units in RR?
No, there are no units/ (there is no time)
What type of number can RD be?
-infinity till +infinity
What is ARR?
Absolute risk reduction
What does ARR mean?
Risk is lower in exposure group
What is ARI?
Absolute risk increase
What does ARI mean?
Risk higher than exposure group
Which dis-ease occurrence is better?
Why is RD better than RR?
We are not eliminating a factor (the time) so we can gage time as a factor which had affected the study we are undertaking.
TIME IS NO ELIMINATED
What happens if the outcome is in mean in RD?
You take a mean difference
Incidence equation? for EG and CG are?
EG, EGO (yes outcome) : [a / EG] or = [a / (a + c)] during time T
CG, CGO (yes outcome) : [b / CG] or = [b / (b + d)] during time T
Prevalence equation for EG and CG?
EG, EGO (yes outcome) = a / EG at one point in time
CG, CGO (yes outcome) = b / CG at one point in time
What are the two types of error in studies?
1. Non random error
2. Random error
What do you associate RAMBOM with?
Non random error
What do you associate CI with?
What is non random error?
Errors which is because of poor study design, were the studies went wrong.
Poor implementation of study.
What are other names for non-random error?
What acronym can be used to remember non random error?
What does RAMBOM stand for?
Blind and Objective Measurement
What are random errors?
Errors caused by chance, they will happen.
What can limit random error?
1. Increase the sample size
2. Take multiple measurements
What is recruitment error in non random error?
When the recruited participants are not repping the real population you are bringing the study back into.
What is a typical example of a recruitment error in non random errors?
In most studies you can volunteer to take part of a study and there is a possibility that the people who "volunteer as tribute' are different to the people who turn down the opportunity to take part.
What is another name for recruitment error in non random error?
Recruitment bias/selection bias
What does the A stand for in RAMBOM and what is it?
Allocating a person into the study was wrong due to wrong measurement taking or confounding.
What is confounding?
Confounding or allocation bias is when we conduct a study and there are factor which determined whether the individual was in the EG or the CG.
There are more factors which are different compared to the CG.
Give an example of confounding.
EG and CG are allocated and the people in EG (who like to exercise) are likely to be younger, eat healthier, and smoke less.
Which study has little to no confounding?
What are ways to prevent confounding? (5)
Undergo blind studies (a person who is not part of the study can allocate EG and CG)
Restrict eligibility criteria (only non-smokers can participate)
Be ethical (don’t put people into the ‘smokers’ category if they weren't before)
Do RCT if possible (with a baseline comparison to ensure EG and CG are similar)
Adjust analysis, have more subsets in age; Age Standardisation (like older people, younger people)
What is maintenance error in non random error?
The maintenance of participants in the study in each group.
Did anyone drop out?
Stayed in the original group they were allocated?
What is BOM?
Blind and objective measurement
What is BOM?
It is basically asking if the people who assess the dis-ease outcomes unaware of (bling to) the participants exposure status or were these measurements made objectively
What is an example of measurement errors in this?
Faulty scales. which show height than actual weight
Blinding participants does whaat?
“Blinding” participants can reduce assessment biases.
Example of blinding?
when trying to work out if physio or a drug is better for pain relief, those patients receiving physio are given a placebo, this means that all patients think they are being given drug. Therefore those patients receiving placebo will not skew the results if they thought that physio has no effect. The nurses administering the drugs are also unaware of those receiving placebo this is called double blinding.)
Assessment bias can be reduced by using objective measurements T/F?
be reduced using objective measurements where possible. (Eg measure the heart rate of people rather than asking them whether they think they are fit or not
How about in surveys for BOM? How do you reduce?
Using standardized questionnaires as objective measures.
CI is associated with what type of error?
Random error is NOT
What does it mean to be not statistically significant?
The CI crosses the no effect line
If it is not statistically significant, can it be clinically significant?
NO IT CANNOT
What is the p value in general?
What is the p value for NOT SIGNIFICANT?
p > 0.05
What is borderline significant?
The CI is just touching the no effect line.
What is the p value for borderline significance?
p = 0.05
What is statistically significant?
When the CI doe not overlap the no effect line at all/
What is the p value for statistically significant?
p < 0.05
What are the components of a CI?
A no effect line (either 1 - RR or 0 - RD, EGO, CGO)
A box and whisker with the
box (the value of YOUR study)
What is the 95% CI actually meaning?
How certain you are (95%) that when we put our study INTO THE POPULATION the participants were repping, if it can translate to the population.
Can a statistically significant study be clinically significant?
Yes and No
What can affect the width of the CI? (2)
1. How much random error is present
2. The % interval
How can the amount of random error affect the width?
The more random error you have, the greater the width. of the CI
How can study size affect the CI width?
Because you are increasing the participant size, you are decreasing random error therefore, you are decreasing the CI width.
How can the % of CI affect the width?
The more certain you want to be (95%) the more leeway you are going to give yourself so the greater the CI width.
You are not that certain (like 20% CI) well the width is smaller
In summary what is the relationship between the CI width and the CI %?
The greater the percentage, the greater the width of CI (leeway)
How do you remember the relationship between the CI width and the CI %?
Remember in a lecture filled with smart professors and you want to be SUPER CERTAIN so you give yourself leeway to make sure your value if taken back to the population will be within the ranges you said.
How do you interpret in words,
EGO = 9.0 (95% CI 8.0 - 10.0)
There is 95% probability that the true value of EGO in the whole population from which the participants were recruited lies between 8.0 to 10.0
CGO = 8.5 (90% CI 8.0 - 11.0)
There is 90% probability that the true value of CGO in the whole population from which the participants were recruited lies between 8.0 - 11.0
Confounding is not in RCT why?
Confounding is allocation error. The only study which has random allocation which reduces confounding is RCT so
all have confounding non random error
What are three things you write out for when you are comparing whether 2 studies (like country A and B) are truly different? (4)
1. Does the CI overlap?
2. What is the upper limit of one?
3. What is the lower limit of one?
4. So are they truly different or nah?
What does truly different mean?
That the upper limit of one CI and the lower limit of one CI are NOT overlapping
What does not truly significant mean?
The upper limit of one CI and the lower limit of one CI IS OVERLAPPING.
Country A CI -- 10 (95% CI 8.0% - 11.0%)
Country B CI -- 20 (95% CI 18% - 22%)
The upper limit of Country A is 11% while the lower limit of Country B is 18% which means these CI do not overlap. Because this doesn't overlap, these two value must be truly different.