The use of automatic wire coding to evaluate control selection bias in the Savitz et al. study. Final report
- Radian International LLC, Denver, CO (United States)
In the Savitz et al. case-control study performed in Denver in the mid-1980s, wire codes were found to be associated with childhood cancer, but magnetic field measurements taken at the same time were not. One explanation is that using select random digit dialing to select controls may have resulted in a deficit in controls living in VHCC homes. To evaluate this possibility, the authors wire coded several hundred thousand homes across the Denver area and then determined which homes housed ``potentially eligible`` children who Savitz could have selected as controls. Automated wire coding was done using computerized data sets of power line wiring and tax assessor records. ``Potentially eligible`` controls were identified using 1980 census data. Because the authors did not know precisely which children lived in which homes across Denver in 1985, they used two assignment processes to determine the wire code distributions of the ``potentially eligible`` controls. First census block data were used to weight the number of homes in each wire code category by the fraction of homes occupied by potentially eligible children, or by the number of children per block. Second, a Monte Carlo assignment process was used to assign children to homes within each block. The authors observed limited evidence of control selection bias especially for VHCC and Buried wire code when they compared the distributions of various versions of these inputed controls to the wire code distribution of the Savitz controls. These results were partially supported when the authors selected 100 sets of controls from each of the Monte Carlo populations to determine if the apparent disagreement of Savitz` controls with the inputed population controls could be due to chance. Given the fundamental limitation of not knowing which homes in 1985 had ``potentially eligible`` children, suggestions of control selection bias must be considered tentative and suggestive only.
- Research Organization:
- Electric Power Research Inst. (EPRI), Palo Alto, CA (United States); Radian International LLC, Denver, CO (United States)
- Sponsoring Organization:
- Electric Power Research Inst., Palo Alto, CA (United States)
- OSTI ID:
- 355074
- Report Number(s):
- EPRI-TR-108044; TRN: AHC29924%%280
- Resource Relation:
- Other Information: PBD: Aug 1998
- Country of Publication:
- United States
- Language:
- English
Similar Records
AEP researcher casts doubt on earlier EMF study
Childhood cancer in relation to a modified residential wire code