Measurement scales, numbers, rates, ratios

Question 1

Three domains of health

Answer 1

health protection
Health improvement
Health

Question 2

Probability vs stats

Answer 2

Knowing whats in the bucket, probability of picking red ball out

Not knowing whats in the bucket but trying to work it out from whats in your hand (sample)

Question 3

Want sample to be…

Answer 3

Representative
Unbiased
Precise

Question 4

Two types of error in study

Answer 4

Chance (random error) - reduce if sample size increases

Bias (systematic error) - does not reduce as sample size increases

Question 5

What is bias?

Answer 5

Difference between true value and expected value

Question 6

What is chance/random error due to?

Answer 6

Sampling variation

Question 7

Sources of bias (selection bias)

Answer 7

Study sample (external validity) - not representative of population

Group selection (internal validity) - groups not comparable

Healthy worker effect - workers usually lower mortality rates

Question 8

Sources of bias (information)

Answer 8

Recall error - differences in recollection

Observer/interviewer error - pre-conceived expectations

Measurement error

Missclassification - put into wrong group (eg diseased when not diseased)

Question 9

Precision vs bias

Answer 9

Need no bias and high precision

Question 10

What does a large study mean in terms of bias?

Answer 10

Large (more precise) study does not necessarily mean less bias

Question 11

Prevalance

Answer 11

Proportion of people who have disease at given time (filled bath)

• old and new cases
• snapshot at given time
• represented as proportion

Question 12

Incidence

Answer 12

Number of new cases of disease within given timeframe (dripping tap into bath)

• new cases only
• incidence rates (events per person per year)

Question 13

Incidence rate ratio

Answer 13

Compares incidence rate to another

• Doesn’t tell if a disease is rare eg 2 in one million vs 1 in one million is double

Question 14

What does relative risk do?

Answer 14

Approximates relative risk if underlying disease is rare or time period is short `

Question 15

Odds ratio

Answer 15

ratio of ratios (disease/non diseased for each groupthen divided over eachother)

ad/bc (a/b / c/d)

Question 16

Relative risk and odds ratio group

Answer 16

Cohort study - consistent follow up

calculate relative risk and odds ratio

Question 17

If no difference in odds/risk number will be

Answer 17

1 (dividing one by the other)

eg if 1.5 - 1.5x more likley
if 0.5 - half as likely

Question 18

What happens if OR and RR is <1

Answer 18

Numerator has lower odds - desired if this event has bad outcome

Question 19

What happens if OR and RR >1

Answer 19

Numerator has higher odds - desired if outcome is good

Question 20

MAIN ISSUE to be aware of when comparing groups

Answer 20

CONFOUNDING factors

Question 21

What is a confounding factor?

Answer 21

Something that can allow the exposure and disease to seem linked as this is actually linked to both

Question 22

Confounding factor example

Answer 22

Gender and cancer of the mouth

Males > likely but is this because on average they drink more alcohol?

Question 23

How can we ensure confounding factors dont influence?

Answer 23

If we are aware of them can adjust for them using standardisation

(problem arises when unaware of confounding factor)