The six analysis files described in [2,5,3] contain demographic, vital status, and annual exposure history data for each cohort member. Regression analyses are based on an analytic data structure (ADS) that consists of a table of observed deaths, ``expected'' deaths (based on U. S. rates), and person-years at risk for each combination of levels of three or more risk factors for each cause of death of interest that are derived from these files.
In the simplest analysis the factors are age at risk (five year intervals), birth cohort, and facility as defined above. The next level of analysis focuses on the comparison of cause specific mortality at the Oak Ridge facilities that include length of employment, SES, and calendar year. The resulting ADS is a four dimensional table of observed and expected deaths for each cause of death. Results are presented in tabular form for race gender groups.
The most complex ADS is required for the dose-response analyses of the X-10/Y-12 subcohort. Each cause of death requires a seven dimensional table of observed deaths, expected deaths, and person-years at risk. The factors that define the table are socioeconomic status (denoted by S with 2 levels), length of employment (denoted by L with2 levels), birth cohort (denoted by B with 5 levels), age at risk (denoted by AG with 15 levels), facility (denoted by F with 3 levels), internal exposure (denoted by IG with 3 levels), and external exposure (denoted by XG with 10 levels). The ten external dose categories were defined using cut points at 5, 10, 20, 40, 80, 160, 320, and 640 mSv. In the case of the last four (time dependent) covariates, workers contribute person time to all levels that they attain. An entry criterion of at least 365 days of follow-up was used for all analyses. Each observation in the ADS consists of the index value of each of the stratifying factors, the observed deaths, the expected deaths, the person-years, and the lagged average cumulative dose (see the Appendix).