Measures of Association

and

Public Health Impact

 Learning Objectives
• List common measures of association and
measures of public health impact

• Calculate and interpret risk ratio and odds ratio

and describe their use

• Calculate and interpret attributable risk percent

and vaccine efficacy and describe their use

n
 Association

Exposure Outcome

Is there a relationship between the

exposure and outcome of interest?

What kind
 Definition—
Association is statistical relationship
between two or more variables

• Association requires comparing two groups:

– Exposed Vs Unexposed
– With outcome Vs Without Outcome
n
 Measure of Association

the strength of the

Exp Outc relationship between an
"exposure" and
Quantifies or “outcome” of interest
expresses
the difference in
Exp+ occurrence of outcome
Outc between two groups who
Exp-
differ on "exposure“
n
 “Exposure”
• Exposure in usual sense
– e.g., ingestion of contaminated food
– e.g., droplets from someone with active pulmonary TB

• Behaviors
– e.g., sharing needles, drinking alcohol etc
• Treatment
– e.g., intervention - education program: yes or no
– drug 1 versus drug 2 in a clinical trial
• Inherent trait or characterizes
– e.g., genotype n
 “Outcome”
• Disease
– e.g., infectious disease: malaria, TB
– e.g., non-infectious disease diabetes, cancer

• Event
– e.g., injury from land mine, car accident

• Condition
– e.g., blindness, other disabilities

• Death

• Other n
 Measure of Association

Exposure Outcome

Is there a relationship between the

exposure and outcome of interest?
 Types of Measure

one can compare two numbers (X & Y) either by

Dividing one by other or subtracting one from the other

X/Y (relative) X – Y (absolute)

Ratio Difference
Ratio Measures Difference Measures

n
Ratio Measures Difference Measures

Relative difference =
Relative Risk Absolute difference =
Excess risk
Tend to identify causal Risk difference
relationships
Attributable risk
Rate ratio, measures the potential or
Risk ratio, public health impact (i.e.
Prevalence ratio, public health implication
Odds ratio (Incidence OR; of association
Prevalence OR)
n
 Description of Relationship
• Variables can be Related or Unrelated to one another

• If the two variables are Related, they can be related

either:
– positively or negatively
– strongly or weakly (one variable can have large or
small effect on the other)
– significantly or not significantly

n
Relationships between variables —
Related or unrelated?
10

Dependent variable
4

0.00 20.00 40.00 60.00

Independent Variable

Related unrelated
 Relationships between variables —
Positive & Negative association

10 10

Y Y

0 0

0 X 1 0 X 1

Positive Negative
n
 Relationships between variables —
Large or Small effect
10 10

Y Y

0 0

0 X 1 0 X 1

Strong/Large Weak/Small
n
 Significant or non-significant
• Statistically significant
 the observed association is unlikely to be due to
chance alone

– Determined by statistical tests that assess how

consistent the observed association is with chance

– It is influenced or dependent on the strength of the

association and sample size

n
 Statistical significant (SS)
•SS does not necessarily mean that the finding is important
from a public health or clinical point of view.
–A very large study can find that a very small difference (ex:1
mmHg in BP) b/t two groups is SS, but it is clinically irrelevant.
–A very small study on drug with 5 patients may have a
tantalizing finding that may turn out to be important, but the study
is too small to reach SS

•SS does not mean that the relationship is causal.

–Having matches in your pocket & lung cancer may be SS, but
the matches do not cause lung cancer. They are merely a
marker of people who smoke. n
 Sample Cohort Study
• Women of childbearing age, in country X followed for
almost 3 years

- Enrolled TB Incidence

– HIV-positive 249 22

– HIV-negative 310 1

– Total 559 23
n
 Follow up Study
• Original enrollment 1971 - 1975
• Follow-up 1982 – 1984

• Complete follow-up on:

- Enrolled Alive (Follow
Up)
– Male diabetics 189 89
– Male non-diabetics 3151 2340
– Total 3340 2429

who were ages 40-77 years at enrollment

 Risk
# new cases during a specified period
population at risk during the specified period

= "Attack rate“ i.e. how may are attacked?

= Probability of getting disease

= Risk of disease

= Cumulative incidence

= Incidence proportion
 Risk Calculation
Example

1. Deaths in diabetic men

100 deaths
189 men at start of follow-up period
Risk = 100/189 = 0.529 = 52.9%

2. Deaths in non-diabetic men

811 deaths
3151 men at start of follow-up period
Risk = 811/3151 = 0.257 = 25.7%
 Measures of Association
Risk ratio =
cumulative incidence in exposed group
cumulative incidence in unexposed gp

Rate ratio =
incidence rate in exposed group
incidence rate in unexposed group

Prevalence ratio =
prevalence in exposed group
prevalence in unexposed group
 Relative Risk or Risk Ratio
• It compares the risk of a health-related events
(disease) in two groups

• It directly compares new occurrence (incidence) of

disease among exposed and non-exposed

• It estimates the likelihood of developing disease in

exposed relative to non-exposed
 Relative Risk or Risk Ratio

incidence of a disease among exposed (a/(a+b)) Re

incidence of a disease among non-exposed (c/(c+d)) Rè

Risk in exposed
Indicates the likelihood of
developing the disease in the
Risk in unexposed exposed group relative
to those who are not exposed.
 Relative Risk or Risk Ratio

RR = Re / Rè
Disease
Yes (+) No (-) K
. a .
Exposure n
RR = a+b .
Yes (+) a b
a+b
o
c
. .

No (-) c d c+d
w
c +d
n

• It is a direct measurement of risk

• It is usually used in Cohort and Experimental study
design
 Relative Risk or Risk Ratio
EXAMPLE 1:

Relative risk of death among diabetic men vs.

non-diabetic men

risk of death among the diabetics

risk of death among the non-diabetics

RR = 100/189 = 0.529 = 52.9% = 2.1

811/3151 0.257 25.7%

“Diabetics were 2.1 times as likely to die during the follow-up

period as were non-diabetics.”
 Characteristics of Risk Ratio
• RR=1= there is no risk associated with exposure to
outcome of interest [i.e. there is no difference in
ratio of risk between the two groups]

• The further away from 1, the stronger the

association between exposure & disease [RR of 5
is stronger than RR of 2, etc]
 Characteristics of Risk Ratio
• Since a risk ratio is a ratio of risks (incidence, attack
rates) we can only calculate a Risk Ratio if we can
calculate risks.

• So we need the right data to calculate ‘risk’, i.e.

• the number of cases and
• the size of the population from which the cases
came,
• Hence we can calculate RR from cohort study.
• But we can’t from case control since we cant’ calculate
27
risks n
 Probability
 Probability is a quantification of the likelihood, or relative frequency, of
occurrence of a given event or set of events, over a sequence of
unlimited random trials done under similar conditions.

• Probability (P) refers to the likelihood that an event will or will not
occur, based upon other factors.
– Probability compares the number of desirable events (A) to the total number
of trials during which the event may occur (N).

• Probability determination is the actual mathematic calculation of the

likelihood of an event occurring.
 Odds ratio is ‘probability’

Odds are also a measurement of the likelihood of

occurrence of an event; however, odds are a ratio
of the probability that an event occurs to the
probability that the event does not occur:

Odd = Probability of occurrence

Probability of non-occurrence
 Odds ratio is ‘probability’
The odds of rolling a dice 1 are the probability of the desirable
event (i.e., rolling a 1) divided by the probability of undesirable
events (i.e., rolling anything besides 1, i.e. 2,3,4,5 or 6). In this
case,

Odds of rolling 1 = Probability of rolling 1

Probability of not-rolling 1
= (1/6) / (5/6); =1/5; = 0.2
 Odds ratio is ‘probability’
If the probability that a given event will occur is known, the
odds can be calculated as;

Odds = Probability/ (1 – Probability)

if the odds that a given event will occur are known,

probability can be calculated as;

Probability = Odds/ (1 + Odds)

 Odds ratio
 Indicates the likelihood of having been exposed among
cases relative to controls

Cases and controls are predetermined and we are calculating

to determine whether cases or controls are more exposed to
a postulated risk factor

 It is an indirect measure of a risk in a disease of rare

occurrence.

 Thus, usually used in a case-control and cross-sectional

analytic study designs
 Odds ratio
.

Odds of exposed among cases (a/c) Oe .

OR =
Odds of exposed among controls (b/d) Oè

occurrence of exposure among cases

non-occurrence of exposure among cases
occurrence of exposure among control
non-occurrence of exposure among control

OR = Oe /Oè
 Odds ratio= cross product ratio
Disease
OR = Yes (+) No (-)

Exposure
Yes (+) a b
No (-) c d
. a/c .
OR = b/d
a+c b+d

KNOWN

OR = Oe /Oè = ad/bc
 OR

Dead Alive
Diabetic 100 89
Nondiabetic 811 2340

OR = ad / bc

OR = ad = 100 x 2340 = 3.2

bc 89 x 811
When Can the Odds Ratio be Used
to Approximate the Relative Risk?
D+ D- Total Risk

E+ a b a+b a/a+b
E- c d c+d c/c+d
a a
RR = a+b ≈ b ≈ ad = OR
c c bc
c+d d
For a rare disease, a <<< b, so a+b ≈ b
c <<< d, so c+d ≈ d
 Example of the “Rare Disease”
Assumption
D+ D- Total Risk
E+ 90 499,950 500,040 0.00018
E– 10 499,950 499,960 0.00002

RR = 90/ 500,040 = 0.00018 = 9.0

10/ 499,960 0.00002

OR = ad = (90)(499,950) = 9.0
bc (499,950)(10)
 Example of the “Rare Disease”
Assumption
D+ D- Total Risk

E+ 90 50 140 0.643
E- 10 50 60 0.167

RR = 90/140 = 0.643 = 3.9

10/60 0.167

OR = ad = (90)(50) = 9.0
bc (50)(10)
 Odds Ratio

 It is nearer to Risk Ratio when disease is

rare

 Can be calculated from case-control and

cross-sectional study

 Has good statistical properties

 It is always far from unity (1.0), compared

to RR
 How to do a 2x2 table
• We are able to make several 2X2 tables
• We should take one category as a referent
Lung CA
Yes No
Lung CA 1 1- 7 cigarettes 35 1,560
Yes No Never smoke 25 3,456
Never smoke 25 3,456 Lung CA
1- 7 cigarettes 35 1,560
2 Yes No
8+ cigarettes 58 890
8+ cigarettes 58 890
n
Never smoke 25 3,456
What is the excess risk
among
exposed individuals?

Attributable Risk
(AR)
n
 Definition
Relative Risk
A measure of the strength of association
based on prospective studies.
Also tend to identify causal relationships

Attributable Risk
The amount of disease that can be
attributed to a certain exposure.
n
 Measures of Public Health Impact
• Places exposure – disease associations from public health perspective

• Quantifies expected reduction in disease occurrence if harmful

exposure could be eliminated

• It is calculated as risk difference (RD)

– The baseline is the ‘exposed group’

Risk Difference = Risk exposed - Risk non-exposed

n
Quantify the excess risk in the exposed that can be
attributable to the exposure by removing the risk of
disease that could have occurred anyway due to other
causes.

Indicates the number of cases of the disease among

the exposed that can be attributed to the exposure
itself.

What proportion of cases is attributed to

the actual exposure among exposed
people? n
 Attributable Risk
Ris
ka
t t ri
ex buta
po
su ble t
re o
Incidence
Incidence

I exposed – I unexposed

ba
se
line Exposed Unexposed
unexposed
ris exposed
ko
for fd
all ise
as I = Incidence
e
n
 Risk Difference

• Relative Risk: (Multiplicative)

RR = Ie / Io
(strength, cause ?)

• Attributable Risk: (Additive)

AR=Ie-Io
(impact).

n
 Risk difference

Risk difference in Butajira Asthma Study

= Risk in urban residents – Risk in rural residents
= 0.031 – 0.011
= 0.02

Interpretation: 2 cases of asthma per 100 people

around Butajira are attributable to living in an urban
environment.
 Attributable Risk Percent (AR%)
• AR% is an attributable risk expressed as a percentage
of risk in exposed

AR% = Risk in exposed - Risk in unexposed

Risk in exposed
• It addresses the question, what proportion of the illness
or deaths can be attributable to the exposure?

• i.e what is the proportion of disease among the exposed

which can be attributed to the exposure?
(AR%) = Attributable proportion
= Attributable fraction =Etiologic fraction (EF)
AR% = Risk in exposed - Risk in unexposed
Risk in exposed
EXAMPLE:
 AR% for deaths among diabetic men vs. non-diabetic
men
(0.529 - 0.257)/0.529 = 0.272/0.529 = 51.4%

"About half of the deaths among diabetic men may be

attributed to their diabetes.“ & the other half to
baseline
n cause
 Attributable Risk Percent (AR%)

Incidence

Iexposed - Iunexposed RR - 1
%  x 100
Iexposed RR

Exposed Unexposed
 AR & AR% in Case-Control Studies
• No direct estimate of risk

• No risk estimates in case-control study

– Attributable risk (risk difference)
– and AR percent

• If odds ratio approximates relative risk (rare occurrence), it is

possible
OR - 1
AR% = X 100
OR
Prevented fraction
 Prevented Fraction (PF)
If relative risk is < 1: exposures is associated
negatively with outcome variable

– Hence the exposure is protective

• Protect proportion of potential cases

– which could have occurred if the exposure had
been absent or
– could have prevented by the exposure
Prevented Fraction (PF)
I unexposed – I exposed
PF =
I unexposed
= 1 - RR

Unexposed exposed
 PF: Vaccine efficacy
Pop. Cases Cases/1000 RR

Vaccinated 301,545 150 0.49 0.28

Unvaccinated 298,655 515 1.72 Ref.

Total 600,200 665 1.11

PF = I unexposed – I exposed 1.72 - 0.49

PF = = 0.72
I unexposed 1.72

= 1 – 0.28 = 0.72
 What proportion of cases is
attributed to the actual exposure
among the general population?

Population Attributable Risk

 Population Attributable Risk
• We want to study health related problem in a community

• But we usually do the study in a sample of a population

• We also conclude our result to the population where the

sample come from.

• When we discuss about population attributable risk, we

are trying to generalize what we found in our sample
 (PAR) Population attributable risk
•Risk in population - Risk in unexposed

•AR * Proportion of exposed in the population

•Measures public health impact of an exposure

•Estimate the excess rate of disease in the total study

population that is attributable to the exposure.

•Indicates the number of cases of the disease in the total

population that can be attributed to the exposure itself.
n
 (PAR) Population attributable risk

• Excess risk of disease in total population

attributable to exposure

• Reduction in risk achieved if population was

entirely unexposed

• Helps determine exposures relevant to public

health in community
 PAR

Risk

I population – I unexposed

Population Unexposed
 (PAR%) Population attributable risk Percent
Population attributable Fraction
• Proportion of cases in the population attributable to the
exposure thus could be eliminated if the exposure were
eliminated
• expressed as a percentage of total risk in population

Risk in population - Risk in unexposed

Risk in population

I population – I unexposed
PAR% = X 100
I population
Population Attributable Risk Percent
(PAR%)

P (RR - 1)
X 100
PAR% =
P (RR - 1) + 1

where P = proportion of population who are exposed

 Summary
• RR & OR • Association

• AR (AR%) • Impact on sample

• PAR (PAR%) • Impact on population

 Interpretation
RR & OR

No association

Preventive Risk

0 1
 Interpretation
RR & OR
1. RR/ OR > 1, the exposure is risk

2. RR/ OR = 1, there is no association

3. RR/ OR < 1, the exposure is Preventive

4. If confidence intervals of RR & OR touch the unity (1),

then there is no significant association
5. If the RR & OR is far from the unity (1), it is a sign of
presence of significant association
 Interpretation
AR

No association

Preventive Risk

0 0
 Interpretation
AR/ PAR
1. AR& PAR > 0, the exposure is
attributing
2. AR & PAR = 0, there is no
attribution
3. AR & PAR < 0, the exposure is not
attributing
 Summary
• Among exposed
– Attributable risk
– Attributable risk percent
– Prevented fraction

• In the population
– Population Attributable Risk
– Population Attributable Risk percent