CASE
EXAMPLE:
Analysis of all cause mortality rates in Suffolk
County, Massachusetts, 19891991, by CT poverty strata.
(click
here for
a pdf version of this page)











We've
created this case example as an opportunity for you to try out
our methods. The example draws on all cause mortality data from
Suffolk County, Massachusetts, between 1989 and 1991. You'll have
a chance to analyze these data by census tract poverty to see
the socioeconomic gradient in mortality in this county. We've
divided the case example into clearly defined tasks to highlight
the process of moving from raw data to summary measures of the
socioeconomic disparity. Clicking on "Methods
& SAS" will take you to a step by step comparison
of each task, the relevant analytic methods, and sample SAS code.

•
Step 1: Aggregate
the numerator data. 
The
file rawcase.csv (click to download)
is a comma delimited file containing all deaths occurring in Suffolk
County, Massachusetts, between 1989 and 1991. Each person who
died is represented by one line in the data file. The variable
“AGE” gives the age at death. The variable “AREAKEY”
is the geocode to the census tract level. Read these data into
a SAS dataset, and then aggregate deaths within each census tract
into the following age categories:


Age
Category 
AGE
(years) 

1 
014 
2 
1524 
3 
2544 
4 
4564 
5 
65+ 

•
Step
2: Aggregate the denominator data. 
The
file rawdenom.csv (click to download)
is a commadelimited file containing the estimated population
count in 31 age categories [see Analytic
Methods Aggregating Denominator Data section] for
the 189 census tracts in Suffolk County, from the 1990 U.S. Census.
Each census tract is represented by one line in the data file,
with the 31 age categories arrayed horizontally.
a. Aggregate the population counts into the five broad age categories
listed above.
b. Transpose the structure of the data, so that there is one
record for each age stratum within a census tract, with a corresponding
categorical age variable and population count. You should end
up with 5 records for each census tract, with each record represented
by one line of your output dataset.
c. Multiply the population count by 3, to yield a persontime
denominator for three years worth of death data.

View
Methods & SAS Programming for Step 2


•
Step 3: Merge the numerators and denominators by AGECAT and AREAKEY. 
For
age cells in census tracts where no cases were reported, set the
numerator to zero.


•
Step
4: Now merge the combined numerator and denominator data
from Step 3 with the ABSM data, by AREAKEY. 
The
file rawabsm.csv (click to download)
is a commadelimited file containing the 189 census tracts in
Suffolk County, and the % of persons living below poverty for
each tract, categorized into 4 categories (1=04.9%, 2=59.9%,
3=1019.9%, 4=20100%).


•
Step
5: For each category of CT poverty,
calculate the agestandardized incidence rate, using the year 2000
standard million. 
In
order to do this:
a.
Aggregate the numerator and denominator within each age X CT
poverty stratum, across all census tracts.
b. Exclude cases and denominator where CT poverty is missing.
c. Merge with the year 2000 standard million in five age categories
(see table below):


Age
in 11 categories 
Year
2000 standard million 
Age
in 5 categories 
Year
2000 standard million 

<1 
13,818 
<15 
214,700 
14 
55,317

514 
145,565

1524 
138,646 
1524 
138,646

2534 
135,573 
2544 
298,186

3544 
162,613

4554 
134,834

4564 
222,081 
5564 
87,247

6574 
66,037

65+ 
126,387 
7584 
44,842 
85+ 
15,508 
d.
Calculate the agestandardized incidence rate [see
Analytic Methods section 1], standardized
to the year 2000 standard million, and the corresponding “gamma”
confidence intervals [see Analytic Methods
section 2] for the direct standardized rates, to fill
in the following table:


CT
poverty 
IRst
(age standardized rate per 100,000) 
95%
confidence intervals ("gamma" intervals)


04.9% 


. 
59.9% 


. 
1019.9% 


. 
20100% 


. 


•
Step
6: Estimate the agestandardized incidence rate difference
and the agestandardized incidence rate ratio [see Analytic
Methods section 4]
comparing the age standardized rates in each poverty stratum
to the rate in the least impoverished poverty stratum (04.9%).

Calculate
the 95% confidence limits on the incidence rate difference and
incidence rate ratio. Fill out the table below:


CT
poverty 
IRDrst
(age standardized incidence rate difference) 
95%
confidence intervals 
IRrst
(age standardized incidence rate ratio) 
95%
confidence intervals


04.9% 
0 
(reference) 
1 
(reference)

59.9% 



. 
. 
. 
1019.9% 



. 
. 
. 
20100% 



. 
. 
. 

•
Step
7: Estimate the relative index of inequality (RII) [see Analytic
Methods section 5]
for CT level poverty in relation to all cause mortality. 
a.
Estimate the approximate cumulative distribution function for
CT poverty, based on the population denominator for each poverty
stratum (summed up over age).
b. Calculate the expected cases in each CT poverty stratum,
based on the agestandardized incidence rate.
c. Fit a Poisson log linear model, modeling the expected number
of cases as a function of the approximate cumulative distribution
of CT poverty, using the population denominator as an offset.
d. Exponentiate the beta term from this model to get the relative
index of inequality.



RII
(relative index of inequality) 
95%
confidence intervals 

Estimate 





•
Step 8: Calculate the population attributable fraction [see Analytic
Methods section 6]
of all cause mortality due to CT poverty. 
a.
Starting with the data from Step 4, sum up over AREAKEY into
strata defined by AGECAT and CT poverty.
b. Calculate (i) the total cases in each age stratum, over poverty;
and (ii) the rate in the reference group of CT poverty.
c. Calculate stratum specific rates, rate ratios, and case fractions.
d. Calculate the agestratumspecific population attributable
fractions.
e. Calculate the grand total of all cases to use in calculating
weights for all age strata.
f. Finally, calculate the aggregated population attributable
fraction, using the age specific weights based on proportion
of cases in each age stratum.



Aggregated
population attributable fraction 

Estimate 


To
see the completed tables
click here:
Answers 
