Steps in Measuring Association

1
Steps in Measuring Association

• Measure the frequency
• exposure or outcome
• Compare the frequencies
• Quantify the comparison
• (measure of association)
• Quantify the impact
• (characteristic on the condition)

2
Measures of Association

• How strong is the relationship between two
  factors
• Association how much does one factor vary
  according to the value of another factor
• Impact how much does one factor account for the
  value of the second factor

3
Measures of Association

• Absolute
• AR attributable risk, risk difference
• PAR Population attributable risk
• Relative
• relative risk and odds ratio
• Impact
• AR Etiologic fraction, Attributable risk
  percent
• PAR Population etiologic fraction, Population
  attributable risk percent

4
Examples

• If you are 5 and your brother is 10. How are your
  ages associated?
• ABSOULTE 5 year difference
• 10-5 5
• RELATIVE 2 times older
• 10/5 (age 1 relative to age 2)
• IMPACT
• (10-5)/ 10 50
• (difference relative to age 1)

5
Examples

• If you are 25 and your brother is 30. How are
  your ages associated?
• ABSOULTE still 5 year difference
• 30-25 5 year difference
• RELATIVE rather than twice
• 30/25 1.2 times older
• IMPACT
• (30-25) / 30 17

6
Measures of Association
7
Measures of Association
Overall proportion of HIV 88/475 0.19
Overall proportion of HIV among IVDU 61/136
0.45
Overall proportion of HIV among IVDU- 27/339
0.08
8
Attributable Risk(AR)

• Also known as Risk difference
• AR Ie Io
• AR 61/136 27/339 36.9
• 36.9 cases of HIV per 100 women IVDUers entering
  NY prisons can be attributed to the IVDU
• or
• Among the exposed, 36.9 cases of HIV per 100
  women were due to IVDU

9
Risk Difference (Cumulative Incidence)
30/1900 - 38/1906 - 4.1/1000 4.1 cardiac
deaths per 1000 cholestyramine users were
prevented by the use of that drug Placebo users
had 4.1/1000 excess deaths compared to
cholestyramine users. JAMA 251351-374, 1984.
10
Risk Difference (Incidence Density)

• 30/54,308.7 60/51,477.5
• 61.3 events per 100,000 persons
• Postmenopausal hormone use prevented 61.3 CHD
  events per 100,000 persons using PHRT compared to
  women not using postmenopausal hormones

11
Etiologic Fraction (AR)
The association between cardiac deaths and
treatment with cholestyramine. JAMA 251351-374,
1984.
Ie 30/1900 Io 38/1906 (cohort
study)
12
Etiologic Fraction (AR)

• AR (Ie Io) / Ie
• AR (30/1900 38/1906) 30/1900
• 26.3
• Among the exposed, 26.3 of all cardiac deaths
  were due to untreated high cholesterol levels.

13
Etiologic Fraction (AR)

• In a case control study you cannot estimate
  incidence of disease among exposed and
  non-exposed
• So, AR (OR 1) / OR x 100
• AR (9.3 1) / 9.3 x100 89.2

(Case Control)
14
Etiologic Fraction (AR)

• AR
• (Ie -Io) / Ie x 100

AR 659/1643 - 25/373 x 100
659/1643 85
(Cohort)
15
Etiologic Fraction (AR)

• The AR calculated from cohort study was 85
• The AR calculated from case control was 89.2
• Why are they different?
• BECAUSE we when we use the OR we are working
  with estimates of the RR

16
PAR/PAR

• The PAR is helpful in determining which exposures
  have the most relevance to the health of a
  community
• The population etiologic fraction(PAR) provides
  an indication of the effect of removing a
  particular exposure on the burden of disease in
  the population

17
Population Attributable Risk
IT Incidence of disease in the total population
IO Incidence of disease among the
non-exposed Pe Prevalence of the exposure in
the population AR Attributable Risk
18
Population Attributable Risk (PAR)
19
Population Attributable Risk (PAR)
20
Population Etiologic Fraction (PAR)
21
Population Etiologic Fraction(PAR)
22
Population Etiologic Fraction (PAR)
(Case Control)
23
Relative Risk

• Disease Occurrence Among Exposure compared to
  Non-Exposure

24
Relative Risk

25
Relative Risk

• IVDU are 5.6 times more likely to develop HIV
  than non-IVDU.

26
Relative Risk (Incidence Density)
RR 30/54308.7 / 60/51477.5 0.47 Women who
take postmenopausal hormones are at almost ½ the
risk of developing CHD as women who do not take
postmenopausal hormones Exposed are .47 times
less likely to develop CHD than Unexposed.
27
Odds Ratio

• Odds of Exposed vs Non-Exposed Among Disease and
  Non-Disease Cases

28
Odds Ratio
Tobacco smoking as a possible etiologic factor in
bronchogenic carcinoma a study of 684 proved
cases. JAMA 143329-336, 1950
29
Odds Ratio

• Individuals with bronchogenic carcinoma were 9.33
  times more likely to have been smokers than
  individuals without bronchogenic carcinoma.

30
In Summary
RR 61/136 / 27/339 5.6
AR 61/136 27/339 36.9/100
AR (61/136 27/339) / (61/136) 82.2
PAR 88/476 27/339 10.6/100
PAR (88/475 27/339) / 88/475 57.0
31
Interpretations

• RR IV drug users were 5.6 times more likely to
  develop HIV than non-users
• AR 36.9/100 cases of HIV among IV drug users
  can be attributed to IV drug use
• AR - 82.2 of all HIV cases among IV drug users
  were due to IV drug use
• PAR 10.6/100 cases of HIV in the total
  population can be attributed to IV drug use
• PAR - 57 of all HIV cases in the total
  population were due to IV drug use