Nellie Wong

SleepyHead.nsey
Station 1: EPI INFO (Static Station Play with that PC)
Sample question:
In a study, it was noted that out of 68 cases of stomach cancer, 40 of these cases has had history of
eating food X before developing the disease. The control group which consisted of 82 non-stomach
cancers in which only 5 of them has had history of consumption of food X.
Step 1:
Read the scenario give. Determine whether its cross-sectional, case-control (OR), cohort study or
experimental study (RR). If you dont know this, go read Station 2 guide. That might help.

Above scenario is a case-control. Therefore, measure of association (MoA, not mechanism of action,
k?) used is OR.

Step 2:
Construct that 2x2 table.

This is the outcome/disease (dependant variable)

Positive +ve Negative ve

This is the risk Positive +ve a b

factors/exposure
(independent
variable) Negative -ve c d

From the given scenario, a = 40, b = 5, c = 28 (68-40), d = 77 (82-5)

Step 3:
Epi Info time! First go to StatCalc.

Next, go to Tables (2 x 2 x n).

You will see this;

Just fill it in. Be sure to not mess up the placement of outcome and exposure!
Ill be using the data from sample question.

After you fill it up, the clever software will automatically generate some numbers for you below the
2x2 table.
For this sample question, were looking at OR. Therefore, its 22!

OR = 22; Therefore,
Those with history of eating food X (exposed) is 22 more likely than those without history of eating
food X (unexposed) to develop stomach cancer (outcome).
If OR = 1, exposed is as likely as unexposed to develop outcome.
If OR < 1, exposed is less likely (do not need to mention number) than unexposed to develop
outcome.
As Dr Htoo Htoo always said, this value is only for the study population. To know whether the results
would be significant in the big population, we need to look at the Confidence Interval (CI). So let us
have a look at the CI now!

CI = 7.8903 to 61.3408
To make it simple, if ONE/1/SATU/ is not within the range, then the findings is significant!

Yay done!

Station 2: Study Design (Active Interview Station)

While there are also descriptive studies such as descriptive cross-sectional and case studies, I doubt
these will come out so I am going to focus on only 4 Cross-sectional, Case Control, Cohort and
Experimental.

If there is intervention done, it will be experimental.

If you are selecting controls aka matching cases to be compared with your study cases, its
case-control.
If you are following up a group of population, its prospective cohort study.
If you take a group of population and trace backwards for exposure, its retrospective
cohort study. (you do NOT select the controls here)
If youre taking a random population at one particular time (prevalence!) to study
relationship of exposure and outcome, its cross-sectional.
1. Cross Sectional
Pros;
1. Short duration.
2. Quick results.
3. Relatively inexpensive.
4. Useful initial overview of the problem
5. Good pre-study for cohort.
6. Can study association of several risk factors with the disease in question.
7. Stronger likelihood of participation as the respondents participate one time .
8. Subject dropout is not a problem.
Cons;
1. Not possible to establish temporal relationship between exposure and onset.
2. Can not reflect causal relationship
3. No incidence rate
4. Increased bias potential (regarding selection of population)
5. Not suitable for diseases of short duration
6. Cases of diseases of long duration over-represented (Prevalence = Incidence X Duration)
7. Inefficient for rare disease need large sample size
MoA OR
Important notes measures prevalence; cheap and fast!
To run the study Choose study population, perform sampling (read the sampling lecture notes for
various random sampling methods, too lazy to summarise), data collection and analyse! :D

2. Case Control
Pros;
1. Easy.
2. Rapid.
3. Inexpensive.
4. Few subjects.
5. Optimum for rare disease.
6. No risk to subjects.
7. Able to study several exposure
8. Identify risk factors prevention and control programmes
9. No attrition.
10. Less ethical problems
Cons;
1. No incidence or prevalence.
2. Not suitable for rare exposure.
3. Temporal relationship difficult to establish.
4. Prone to bias:
- Due to confounding
- Recall bias
- Selection bias
- Berkesonian bias
- Interviewers bias
5. Confounder.
6. Not suitable for evaluation of therapy or prophylaxis of disease
MoA OR
Important notes this is a retrospective study! Both outcome and exposure has already occurred
and you are selecting the control group to compare.
To run the study Select a disease (outcome) -> select cases and controls, set your definition
including inclusion and exclusion criteria -> Matching -> measure the exposure -> analyse!

3. Cohort Study
Pros;
1. Incidence rate.
2. RR, AR, PAR.
3. More than one outcome (disease) related to single exposure can be studied.
4. Dose-response ratios can be calculated.
5. Temporal relationship.
6. Rare exposure.
7. No recall bias.
8. Most definitive information of the three analytical observational studies.
Cons;
1. Large sample.
2. Not suitable for rare diseases.
3. Attrition.
4. Time consuming.
5. Expensive.
6. Needs more man power.
7. Ethical issues.
8. Concentrate to limited number of exposure possibly related to outcome.
9. Selection of comparison groups is difficult.
10. Study itself educates people and alter behavior.
11. Bias:
i. Information (misclassification) bias.
ii. In assessment of outcome (masking the investigator).
iii. Due to attrition / non-response.
iv. Analytic bias
MoA RR
Important notes
- This is a forward study! Yes, even in retrospective cohort. The outcome has YET to happen.
- Studies incidence
- Expensive and time consuming
- Risk of attrition is high
- Used for rare exposure
To run the study Selection of study cohort -> Obtaining data on exposure -> Selection of
comparison group -> Follow up -> analyse!

4. Experimental
Pros;
1. Confirms relationship between independent variable and the dependent variable.
2. Definite time-sequence between cause and effect.
3. No recall bias.
4. Incidence rate.
Cons;
1. Expensive.
2. Difficult to conduct.
3. Impractical for harmful risk factors.
4. Attrition.
5. Problems of cross-over.
6. Involves a lot of ethical issues.
7. Contamination control participants incidentally receive experimental intervention.
8. Co-intervention.
MoA RR
Important notes measures incidence; there is intervention!
To run the study Select a disease (outcome) -> select a population -> randomization (again, read
up on the different type of randomization. Im a lazy lazy sleepyhead) -> intervention/manipulation ->
follow-up -> analyse!

Station 3: Calculation! (Passive, calculator given)

This is so vague. Not sure where to start Just read posting notes. Ill summarise some formulas;

Vaccine Efficacy
F F
Vaccine ef icacy =
F
Fu disease frequency in unvaccinated
Fv disease frequency in vaccinated

Chi square
(O E )
=
E
Oi observed value
Ei expected value

Mean

= or =
x sample mean
population mean

Median
( )
Median =

Coefficient Variant
SD
Coef icient Variant = x 100%

( )
SD =
n1