1. EPID 503 – Class 6
Standardization

2. Last Class: Total Mortality Rate
# people who became newly dead (in some time and place)
# people in population
Same as incidence but all people at risk
*On the population-level, often prevalent disease is assumed to be so rare that total population ~ population at risk for incidence estimation
*Learned that for cervical cancer, this was not the case

3. Last Class: Total Mortality Rate
# people who became newly dead (in some time and place)
# people in population
Rate vs. proportion (ie. risk): On the population-level, we cannot tally person-time, so these “rates” on the large scale are assumed to occur in one year to make people*1 year or person years
Estimated with mid-point population to get best estimate of average population size during time period

4. Today We Discuss Age Standardization
What’s the issue?

5. What we really want to know is – Is a person more likely to die if they were a member of population A as compared to population B?

6. Even with the same age-specific rates, a population that is younger will appear to have lower overall mortality rates.

7. Unadjusted May Be Good for Funeral Director but Problematic for Public Health
While unadjusted rates may be useful if you run a funeral home, we can’t conclude that one population is healthier, at less risk, etc. than another without making them more comparable.

8. To Compare Across Populations We Need Comparable Groups
Method 1 asks “How would the rates of death compare in two populations if they had the same age distribution?” What is this method called? When is it most useful?
While unadjusted rates may be useful if you run a funeral home, we can’t conclude that one population is healthier, at less risk, etc. than another without making them more comparable.

9. Direct Standardization Explained
Age-Specific Rates
from Population A
Age-Specific Rates
from Population B
Because using age-specific death rates from populations typically only used in large groups
Applied to
The age distribution of a
standard population
(eg. US population in 2000)
Note: Standard population is
somewhat arbitrary

10. How to Implement the Direct Method
For each population:
Calculate age-specific rates
Multiply age-specific rates by the # of people in corresponding age range in standard population (generates the expected deaths in each age group of the standard population if they had the rates of your population)
Sum the expected # of deaths across all age groups (generates the total expected deaths in the standard population if they had the rates of your population)
Divide total # of expected deaths by total standard population (generates the mortality rate in standard population if they had the rates of your population )
Result: Age-adjusted mortality rate for your population that can now be compared to crude from standard or other similarly standardized rates

11. To Compare Across Populations We Need Comparable Groups
Method 2 asks “How many deaths would I have expected if this population had the same mortality rates as some standard population (e.g. the US)?” What is this method called? When do we pick this method?
While unadjusted rates may be useful if you run a funeral home, we can’t conclude that one population is healthier, at less risk, etc. than another without making them more comparable.

12. Indirect Method of Standardization
Rates from the
Standard population
Useful when I don’t have or trust the group-specific rates (i.e. population is too small)
Applied to the age distribution of the study population

13. How to Implement the Indirect Method
Acquire age-specific mortality rates for standard population
Multiply standard population’s age-specific rates by # of people in age range in study population (generates the expected number of deaths in your population if it had the mortality rate of the standard)
Sum expected # of deaths across age groups in study population (generates the total number of expected deaths in your population if it had the mortality rates of the standard)
Divide observed # of deaths by expected # of deaths in study population (observed/expected)
Result: SMR (>1 more than expected, =1 as expected, <1 less than expected)

14. Let’s Look at Populations with the Same Age-Specific Rates but with Different Ages
Young
0.002
1000
Middle
0.005
500
Old
0.010
200
Total
1700
Rate N
Young
0.002
200
Middle
0.005
500
Old
0.010
1000
Total
1700

15. With this Distribution, Let’s Estimate the Crude Rates that Would be Observed
Expected Deaths
Young
0.002
1000
1000*0.002 = 2
Middle
0.005
500
500*0.005 = 2.5
Old
0.010
200
200*0.01 = 2
Total
1700
6.5
6.5/1700=0.0038
Rate N
Expected Deaths
Young
0.002
200
200*0.002 = 0.4
Middle
0.005
500
500*0.005 = 2.5
Old
0.010
1000
1000*0.01 = 10
Total
1700
12.9
12.9/1700=0.0076

16. Mathematically It’s a Weighted Average
0.002* * *0 =
17000
0.002* (1700/1700) *(0/1700) * (0/1700) =
0.002* * *0 = 0.002
We’re basically just shifting the overall rate to more closely resemble the rates in the groups with the most number of people
Rate N
Young
0.002
1700
Middle
0.005
Old
0.010

17. Summary of Direct and Indirect Adjustment
Population Structure
Age-specific death Rates
Direct
Standard
Observed
Indirect