DIRECT STANDARDIZATION

For a detailed discussion of statistical issues related to DIRECT STANDARDIZATION refer to the textbook by Breslow and Day (1987) "Statistical Methods in Cancer Research, Vol II" ( BDII). Chapter 2 "Rates and Rate Standardization" considers direct standardization.

Let y(j) denote the number of "events" in age group j for some facility in a certain time period, and n(j) the corresponding person-years ( in 1000s ). The first two decisions that need to be made before age-adjusted rates can be computed are:

1) What age intervals to use , and 2) what weights to use.

The data could appear as follows ( see  Example ):

  j  Age-Interval       y(j)   n(j) r(j)   W(j)
  ---------------------------------------------
  1  18-29              .       .    .      .
  2  30-39              .       .    .      .
  3  40-49              .       .    .      .
  4  50-59              .       .    .      .
  5  60-69              .       .    .      .
  ---------------------------------------------

  where   W(j) = 1000*N(j)/NT, and  NT = sum(j=1 to 5) N(j).

N(j) is the standard population in age interval j,
 and the rate for the jth interval is

                 r(j)= y(j)/n(j).


The directly standardized rate ( BDII eq 2.1 ) is then

	R =   Sum [  w(j)* r(j) ]
	       j
UNITS FOR THE ADJUSTED RATE ARE EVENTS PER 1000 PERSONS/YEAR.

Age-adjusted rates are used to compare occupational groups within a facility in the ES annual reports ( see Rocky Flats Example). If the objective is to compare results from different facilities over time the same age intervals and weights must be used in each report. If results in another report are based on the same standard population ( US 1970 ) but different age intervals and weights then results will not be comparable.