Skip to Content
Discovering the causes of cancer and the means of prevention

Publications Search - Abstract View

Title: An approach to evaluating bias and uncertainty in estimates of external dose obtained from personal dosimeters.
Authors: Gilbert ES,  Fix JJ,  Baumgartner WV
Journal: Health Phys
Date: 1996 Mar
Branches: REB
PubMed ID: 8609025
PMC ID: not available
Abstract: This paper describes an approach to quantifying errors in recorded estimates of external radiation dose obtained from personal dosimeters and applies the approach to dose estimates of workers at the Hanford site. Because a major objective of this evaluation is to provide the information needed for adjusting epidemiologic dose-response analyses of worker data for errors in dose estimates, the paper addresses the extent that errors for different workers are correlated, focuses on recorded doses as estimates of organ doses, and focuses on recorded doses as estimates of organ doses, and gives consideration to both annual and cumulative doses. The evaluation emphasizes errors resulting from the fact that dosimeters are limited in their ability to respond accurately to all radiation energies to which workers are exposed or to radiation coming from all directions. For each of several sources of error, systematic bias factors are estimated for two energy ranges (100--300 keV and 300--1,000 keV), two geometries (anterior-posterior and rotational), and four calendar year periods. These are then combined using information provided by health physicists on energies and geometries in Hanford exposure environments. Except for the period before 1958, deep dose, the objective of modern dosimetry systems, was found to be fairly accurately estimated. Lung dose was found to be overestimated by about 10%, and bone marrow dose was found to be overestimated by about 50%. However, many aspects of this evaluation relied heavily on subjective judgments, and, thus, these factors are subject to considerable uncertainty. Estimates of uncertainty in the bias factors and uncertainty reflecting random error are provided.