Measures of Association

1
 Learning objectives
At the end of this module, participants should be
able to
Describe the ways of expressing chance of an
event
Describe how to express and interpret the relative
chance of an event when comparing groups
Calculate and Interpret the meaning of relative
risk, attributable risk and odds ratio from
epidemiological data

2
 Introduction
To establish causal relationships and to identify
effective interventions, the occurrence of disease in a
group of people exposed to a risk factor is compared to
that observed in an unexposed group.
In this way we can quantify the association between a
risk (or protective) factor and a disease (or other
outcome).
An important application of epidemiology is to
estimate how much disease is caused by a certain
modifiable risk factor.

3
 Measures of exposure effect and impact
To investigate a possible association between a risk
factor and a particular disease, the incidence of disease
in the people exposed to the risk factor is compared
with the incidence in a group of people who were not
exposed.
This comparison can be calculated by various methods
oRelative measures
• Risk ratio
• Rate ratio
• Odds ratio
o Absolute measures
4
• Attributable risk
•
 Relative measures
Relative measures estimate the size of an association
between exposure and disease
It also indicate how much more likely people in an
exposed group are to develop the disease than those in
an unexposed group.
There are also three relative measures that can be used
to calculate association between disease and exposure:
o Risk ratio
o Rate ratio and
o Odds ratio.

5
 Risk ratio
The risk ratio, also commonly referred to as relative
risk
It is calculated as the ratio between the cumulative
incidence in the exposed group and the cumulative
incidence in the unexposed group.
Consider the below 2x2 table Disease

Yes No Total

Risk factors Exposed a b a+b

Unexposed c d c+d

Total a+c b+d a+b+c+d

6
 Risk ratio
The incidence of disease in the entire population is (a +
c)/(a + b + c + d) per year.
We define the risk ratio as:

Since the incidence of disease in the exposed group is

a/(a + b) and in the unexposed group is c/(c + d)
The risk ratio can be calculated as:

7
 Risk ratio
The risk ratio is used as a measure of a etiological
strength.
The value of risk ration can be =1, > 1 or < 1
 A value of 1.0 will be obtained if the incidence of
disease in the exposed and unexposed groups is
identical
This indicates that there is no observed association
between the exposure and the disease.
 A value >1.0 indicates a positive association or an
increased risk among those exposed to the factor.
A value < 1.0 means that there is an inverse
8 association or a decreased risk among those exposed
 Rate ratio
The rate ratio is calculated in the same way as the risk
ratio, except that the incidence rates in the exposed and
unexposed groups are used:

The rate ratio takes into account the amount of time

that each person contributes to the study
It is preferred in analytical studies in which the
outcome is common, large numbers of people are
entering and leaving the study population or have
changing levels of exposure.
9
 Example
Among 2390 women aged 16 to 49 years who
were free from bacteriuria, 482 were OC users at
the initial survey in 1973, while 1908 were not. At
a second survey in 1976, 27 of the OC users had
developed bacteriuria, as had 77 of the non users.
Calculate the measure of association and interpret
it.

10
 Example:
Bacteruria
Yes No Total
O
C Yes 27 455 482

U No 77 1831 1908
S
E Total 104 2286 2390

11
 Example
 Calculate RR

RR = Ie = 27
lo 482 = 1.4
77
1908

 Interpretation: women who used oral contraceptive

had 1.4 times higher risk of developing bacteruria
when compared to non-users.

12
 Example 2
A study on postmenopausal hormone use and coronary
heart disease among postmenopausal female nurses
showed that after a total of 54,308.7 person-years of
follow-up, 30 women who reported that they had used
hormones developed CHD. For the “never users”, 60
developed CHD among 51,477.5 person-years of
follow-up.
Draw a two-by-two table and calculate the measure of
association.

13
 Measures of Association

Hormone CHD Person-Years

use
Yes No

Yes 30 - 54,308.7
No 60 - 51,477.5
Total 90 105,786.2

14
 Measures of Association
Ie/Io = IDe/IDo = a/PY1
c/PYo
= 30/54,308.7
60/51,477.5
= 0.5
Interpretation:-The risk of developing CHD in those who
use post menopausal hormone is half of those of non
users.

Kifle Woldemichael (MD, MPH)

15
 Odds ratio
Odds ratio is calculated as the odds in the exposed
group are compared with the odds in the unexposed
group:

Based on the above 2x2 table, the odds of disease in

the population is calculated as (a + c)/(b + d).
The odds of disease in the exposed group are a/b and
In the unexposed group are c/d
Therefore the odds ratio is:
 =
16
 Odds ratio
Odds ratios are usually used in studies where:
The incidence of the disease of interest is not
known or
If the study participants are selected on the basis of
their disease status rather than because of their
exposure status.
In this case, rather than calculating the odds of disease
in the exposed and unexposed groups, the odds of
exposure are calculated in those with and without
disease.
Odds ratio is most commonly calculated in cross

17
sectional and case control study designs
 Example 3
Of 156 women with Myocardial Infarction (MI),
23 were current OC users at the time of their
hospital admission. Of the 3120 control women
without MI, 304 were current OC users. Calculate
the measure of association and interpret it.

18
 Example 3

Myocardial infarction
yes No Total

Yes 23 304 327

Current
OC use No 133 2816 2949

total 156 3120 3276

19
 Example 3

Calculate OR
OR = ad = (23) (2816) = 1.6
bc (304) (133)

Interpretation: Women who were current OC users had 1.6 times

higher risk of developing myocardial infarction

when compared to non-users of OC

RR can be estimated by OR if the following conditions are fulfilled:

The controls are representative of the general population

The selected cases are representative of all cases
The disease is rare

20
 Absolute measures
Relative measures are useful when we want to know
how strongly an exposure is associated with a
particular disease
But, they do not give us any indication of the impact
of that exposure on the incidence of disease in that
population.
This has important implications for any public health
prevention measures we may want to take.
Absolute measures are therefore used to indicate
exactly what impact a particular disease or condition
will have on a population, in terms of the numbers or
21
percentage of that population affected by their being
exposed.
 Absolute measures…
For example: The risk of cancer in people exposed to
a certain carcinogen may be 20 times that observed in
people not exposed.
However, this tells us nothing about how common that
type of cancer might be
However, the absolute or attributable risk will tell us
exactly how many more people are affected in the
exposed group than in the unexposed group.

22
 Attributable (absolute) risk
The attributable or absolute risk can give information on
how much greater the frequency of a disease is in the
exposed group than in the unexposed group assuming
the association between the exposure and disease is
causal.
Attributable risk measures the difference in frequency of
a disease between two groups, not the magnitude of
association
It is the risk of disease in the exposed group that is
attributable to the risk factor, after taking into account
the underlying level of disease in the population (from
other causes).
23
 Attributable (absolute) risk…
Attributable risk is also known as risk difference or
excess risk.
Attributable risk=Incidence in exposed group −
Incidence in non-exposed group

For example: If we know that the risk of disease in

those exposed is 5 cases per 1000 people per year and
in those unexposed is 3 cases per 1000 people per year
The risk that is attributable to the exposure (i.e. the
attributable risk) is 2 cases per 1000 people per year.

24
 Attributable (absolute) risk…
This can be expressed as the proportion of disease in
the exposed group attributable to the exposure.
 This is known as the attributable risk percent of the
aetiologic fraction.

25
 Attributable (absolute) risk…
In the above example, the attributable fraction will be
the attributable risk we have calculated divided by the
risk in the exposed group
This gives an attributable fraction of 0.4.
If we multiply this by 100, we get an attributable risk
of 40%.
This means that 40% of disease cases could have been
prevented if the exposed group had not been exposed to
the risk factor.

26
 Attributable (absolute) risk…
The attributable fraction can also be calculated from
the relative risk (risk ratio or rate ratio) by using the
following formula:

If we knew only that the risk ratio was 1.67, and did
not know the risks in the exposed and unexposed
groups, we could use the above formula to give an
attributable fraction of 0.4

27
 Attributable (absolute) risk…
The attributable fraction is usually used if there is a
positive association between the exposure and the
outcome.
However, if the exposure prevents the outcome, the
attributable fraction will be negative, which is difficult
to interpret.
In such a case, we would look instead at the risk or
rate that is attributed to not being exposed by replacing
the denominator by the risk in the unexposed group.
This measure is known as the preventable fraction in
the non-exposed group:
28
 Attributable (absolute) risk…

This formula can also be expressed in terms of the risk

or rate ratio, as follows:
Preventable fraction=1−Risk (or rate) ratio.

29
 Population Attributable (absolute) risk
The same concept can be applied to the population as a
whole.
It is of benefit to public health if we can estimate the
excess disease present in a population that is due to a
particular risk factor, or estimate the relative
importance of different risk factors.

.
Population attributable risk=Incidence in the
whole population−Incidence in non-exposed
population

30
 Population Attributable (absolute) risk…
The population attributable risk fraction is the
proportion of disease observed in the whole population
that is attributable to exposure to the risk factor.
It estimates the proportion of disease that might be
prevented if the risk factor were removed.

31
 Population Attributable (absolute) risk…
The population attributable risk fraction can be used to
estimate the benefit of a proposed intervention
Example: It estimate number of lung cancer deaths
that could be prevented by introducing a smoking
reduction program in a large population.
Measures of attributable risk rely on a number of
assumptions:
First, the association between the risk factor and
the outcome is causal
Second, that there is no confounding or bias in the
measurement of incidence (e.g. selection bias).
32
 Population Attributable (absolute) risk…
If we do not know the risk of disease in the population,
we can still calculate the population attributable risk, as
long as we know the proportion of the population in the
exposed group (p):
Population attributable risk = p×Attributable risk

The population attributable risk will therefore always

be less than the attributable risk in the exposed group,
since p should always be less than 1

33
 Population Attributable (absolute) risk…
For example:
Suppose we know that the rate of liver cancer to be
602 per 100,000 person-years in a group of people with
chronic hepatitis B infection and 4 per 100,000 person-
years in those without chronic hepatitis B infection.
If the rate of liver cancer in the entire population is 94
per 100,000 person-years
The population attributable risk is (94 −4) = 90 per
100,000 person-years.

34
 Population attributable fraction
We can also indicate what proportion of the total risk
of the disease in the population is associated with the
exposure.
This is expressed as the population attributable
fraction (or the population attributable risk per
cent, if multiplied by 100)
It indicates the proportion of the disease in the
population that could be prevented if exposure to the
risk factor could be eliminated, and the entire
population was unexposed.

35
 Population attributable fraction…..
This is calculated in a similar way to the attributable
fraction:

36
 Population attributable fraction…..
In our previous example the rate of liver cancer in the
population was 94 per 100,000 person-years, and
The population attributable risk was 90 per 100,000
person-years.
Therefore we calculate the population attributable
fraction (given by 90 divided by 94), which gives a
value of 0.96.
This means that, if we could prevent chronic
hepatitis B infection in this population, we estimate
that 96% of cases of liver cancer could be prevented.

37
 Practical example
The following is a hypothetical example to illustrate
the use of attributable risk.
A causal association has been established between
standing and backache.
Imagine we wanted to establish how important
standing was as a risk factor for backache in relation to
all cases of backache.
The risk of backache per 100 female workers over a 2-
year period in standing female workers is 12.3 and for
other female workers it is 7.7. The risk of backache in
all female workers is 8.3.
38
 Practical example….
a. Calculate Attributable risk
b. Calculate attributable risk fraction
c. Calculate population attributable risk and
d. Calculate population attributable risk fraction

39
 THANK YOU

40