Comparing RatesHas the cancer mortality rate changed from 1940to 1980?Are people more likely to die in Floridathan in Alaska?
•
1940
–
120.2/10
5
/yr
•
1980
–
183.8/10
5
/yr (crude)
–
132.7/10
5
/yr (adjusted)
151
152
Vital Statistics of the US, 1991
153
Age
‐
adjusted Mortality Rates for FL and AK, 1988
Calculated by Direct Adjustment
Florida = 1,996,000/(245.7 x 10
6
) = 812.4 per 100,000
Alaska= 1,879,000/(245.7 x 10
6
) = 764.8 per 100,000
154
Are people more likely to die in Florida
than in Alaska?
155
Picking the ‘Standard’
•
Actual adjusted values will vary with the
selection of the ‘standard’ population
–
Comparisons between groups will usually
remain fairly constant
•
Important: All groups in the comparison
should be adjusted to same standard
156

27
Standard Populations
157
1979
1981
1983
1985
1987
1989
1991
1993
1995
Death Rate per 100,000 population
1200
1000
800
600
400
200
1940 standard
Crude Death rate
2000 standard
158
Indirect Adjustment
•
Standard population rates x observed population
–
Apply age
‐
specific rates from standard population to
•
age
‐
specific population under study
–
Calculate ‘expected’ number of cases (if the rates from a
standard population were applied)
–
Then compare to what was observed:
SMR = observed/expected x 100%
159
Indirect Standardization
160
SMR = o/e x 100% =
90.7%
89.9%
*
*
*
*
*
*
*
Interpreting the SMR
•
< 100%
–
This population has fewer events than you would
expect based on the standard rates
•
= 100%
–
This population has the same number of events
that you would expect based on the standard rates
•
> 100%
–
This population has more events than would be
expected based on the standard rates
161
When to use indirect adjustment?
•
When category
‐
specific rates are not known
•
When populations are small (and rates not
stable)
–
Occupational settings
–
Small communities for short time periods
162

28
Direct vs Indirect Adjustment
•
Adjustment
Strata pop X Strata Rate
–
Direct Adjustment
•
Standard Population X OBSERVED RATES
–
Indirect Adjustment
•
OBSERVED POPULATION X Standard Rates
163
Summary
•
Different types of rates
–
Crude, category specific
•
How to compare two rates?
–
Strata
‐
specific always appropriate
–
Direct
‐
adjustment for summary comparisons
•
Direct: Standard populations
•
Indirect: Standard rates
164
Interpretation: Chance
Introduction
165
Learning Objectives
•
Understand the role of
chance
in epidemiologic studies
166
Sampling
•
Goal of study
–
Determine the true relation between exposure
and disease
•
Actual results may vary
–
Sample vs. whole population
–
Sampling variability
167
Statistics 101
168
Urn holds 100 marbles
red and/or blue
draw 4

29
Statistics 101
169
Red
Blue
Conclusion
0
4
All blue
1
3
75% blue
2
2
50% blue
3
1
25% blue
4
0
All red
Statistics 101
170
Red
Blue
Probability
Conclusion
0
4
6%
All blue
1
3
25%
75% blue
2
2
38%
50% blue
3
1
25%
25% blue
4
0
6%
All red
50 Blue, 50 Red
Statistics 101
171
50 Blue, 50 Red
Sample Size
Probability of sample with 1 color
(%)
4
12.0
5
5.6
6
2.7
7
1.2
8
0.6
9
0.3
10
0.1
15
0.0018
20
0.000018
Statistics 101
•
Given
–
Hypothesis
•
the chance of drawing a red marble on any one try is
50%
–
information about sample size
•
Possible to calculate probability of
–
a bad sample, or
–
observing a particular result from a set of sample
data
172
Statistics 101
sample size
sampling variability and
probability of an unrepresentative sample
173
“Bad samples happen” – S. Pettygrove

#### You've reached the end of your free preview.

Want to read all 46 pages?

- Fall '18
- Confounding, Case-control study