TABLE OF CONTENTS
Executive Summary
Introduction
Adjustments to ORNL External Doses
Background
Selection of Unexposed Employment-Years
Hardcopy Monitoring Records
Use of Pocket Meter Data for Adjustment of 1944-1956
Doses
Replacement of 1943 Unmonitored Employment-Years
Adjustments to Y-12 External Doses
Background
Estimating Doses for Unmonitored Employment-Years,
1947-60
Dose Adjustments for 1948 and 1949
Replacement of Missing Doses by the Nearby Procedure
Results of Dose Adjustment on Cumulative Doses
Discussion
References
LIST OF TABLES AND FIGURES
Tables
Table 1 Definition of Abbreviations Used in Hardcopy
Record
Table 2 Number of Annual Recorded External Doses
and Unmonitored Years Adjusted for ORNL
Table 3 Number of Annual Recorded External Doses
and Unmonitored Years Adjusted for Y-12
Figures
Figure 1 Sample Hardcopy External Exposure Record
from ORNL
Figure 2 Adjusting ORNL External Doses - 1943 to
1956
Figure 3 ORNL Cumulative External Doses: 1943 -1956
Figure 4 Y-12 Cumulative External Doses: 1947 -1960
EXECUTIVE SUMMARY
An earlier report, Data Collection, Validation, and Description for the
Oak Ridge Nuclear Facilities Mortality Study (Watkins et al., 1993), described
demographic and radiation exposure data characteristics, and the effect
of monitoring policies and practices on annual recorded external radiation
doses and internal radiation exposure indicators for workers employed in
nuclear facilities in Oak Ridge, Tennessee, through 1984. Investigations
of changing monitoring policies and practices from the Oak Ridge National
Laboratory (ORNL) and the Y-12 site indicated that certain annual recorded
doses were likely to be biased downward. To evaluate the effect of this
bias on dose-response coefficients, the report concluded that adjusted annual
external doses should be calculated as replacements.
As a supplement to the earlier report, this report details specific procedures
used for adjusting doses from ORNL and the Y-12. Comparisons of unadjusted
and adjusted cumulative external doses for both facilities are also presented.
For ORNL, the only major change in dose distribution was percent of workers
with a cumulative dose of zero through 1956; this percent decreased from
31.9% to 11.7%. The mean cumulative dose increased from 1.08 to 1.63 cSv,
and the median increased from 0.13 to 0.46 cSv. For Y-12, the decrease in
percent of workers with a cumulative dose of zero was much more significant
because few workers were monitored before 1961. As a result of the adjustment
process, this percentage decreased from 74.3% to 10.1%. However, most Y-12
workers still had cumulative doses through 1960 of less than 1 cSv, with
50% having less than 0.2 cSv. The results of the dose-response analysis
using unadjusted and adjusted doses will appear in a separate publication.
Back to TOC
INTRODUCTION
A previous ORISE technical report (Watkins et al., 1993), to which this
report is a supplement, described how doses were obtained for the Oak Ridge
nuclear facilities mortality study (Frome et al., 1994). The annual recorded
external dose was the sum of film badge readings for the year, and any unmonitored
year was considered to have a zero occupational dose. The magnitude of a
recorded dose reflected not only the amount of occupational radiation exposure,
but also monitoring and recording policies and practices at the time. These
policies and practices are discussed and summarized in Watkins et al. (1993).
(See "Monitoring Policies at the Facilities" and "Monitoring
Data Characteristics of the Study Cohort," and Figures 3, 4, 6, B1,
and B2.)
As a result of investigating monitoring and recording policies in effect
at each facility over the 42 year study period, it was determined that,
despite different approaches in facility monitoring programs, certain annual
recorded doses likely underestimated actual doses from both ORNL and Y-12.
Therefore, for the dose-response analysis subcohort (white male employees
of ORNL or Y 12), upward adjustments were made to appropriate annual recorded
external doses for ORNL employment-years prior to 1957, and estimated doses
were derived to replace Y-12 unmonitored employment-years prior to 1961.
For both facilities doses were also estimated for the small percentage of
annual doses that were missing during years when policy required monitoring
all workers.
Back to TOC
ADJUSTMENTS TO ORNL EXTERNAL DOSES
Background
From mid-1944 through mid-1956, film badges for ORNL workers were read weekly,
allowing only a short time to accumulate exposure. It was general practice
to record as zero those readings below the detection limit of the badge.
From 1948 through 1951 the detection limit was taken to be 0.03 cSv. During
other periods it was dependent on the sensitivity of specific batches of
film, always being 0.05 cSv or less and generally falling between 0.03 and
0.01 cSv (Wing et al., 1994). The combination of weekly readings and the
zero recording practice made it likely that annual recorded doses for many
employment-years underestimated actual doses. For a worker in 1949 receiving
0.03 cSv per week, random errors would result in badge readings below 0.03
cSv approximately 50% of the time that typically would be recorded as zero.
For remaining weeks the dose would be recorded as 0.03 cSv or slightly higher.
Based on these recording practices, nearly 50% of the actual dose may have
been unrecorded for that year. Even workers in areas with higher exposure
potential, whose weekly readings were generally greater than 0.03 cSv, could
have weeks during the year when they received some dose but had recorded
values of zero. After 1956 when film badges were read quarterly rather than
weekly, the magnitude of missed dose was likely much smaller because film
badges had 13 times as long to exceed the threshold value and record positive
readings for employees who had small weekly exposures.
Readings of film badges placed at specific fixed locations throughout the
X-10 site support the premise that doses were underreported at ORNL prior
to 1957. These data indicate that exposure potential at the selected locations
remained fairly constant for years just before and after the badge exchange
frequency changed from weekly to quarterly, although the population average
recorded dose increased significantly when quarterly readings were begun
(Morgan, 1963; Morgan, 1968).
The problem of missing dose from annual recorded doses for ORNL workers
has recently been examined by Kerr (1994). He recommends that doses for
epidemiologic studies be reassessed using the vast amount of hardcopy data
including personnel dose records, daily pocket meter readings, and monitoring
results from building surveys and fixed stations. A significant expenditure
of time and effort would be necessary to computerize these data before such
a dose assessment could begin. Therefore, as a preliminary attempt to compensate
for likely missed dose in annual recorded doses, the dose adjustment procedures
described below have been implemented. A small sample of daily pocket meter
readings were computerized for this purpose.
Back to TOC
Selection of Unexposed Employment-Years
Not all members of the dose-response analysis subcohort at ORNL had employment-years
that were candidates for upward dose adjustment. Certain classifications
(e.g., accountants) were likely to have no occupational radiation exposure
associated with them. Therefore, it was necessary to separate employment-years
that were likely to be unexposed work years from those with exposure potential.
After identifying unexposed employment-years, a reasonable dose adjustment
process could be applied to remaining annual recorded doses.
Less than 7% (2,096) of employment-years for the subcohort from 1944 through
1956 had missing annual recorded doses. Doses for these years were treated
as zeros for adjustment purposes since they had been changed to zeros when
calculating cumulative doses for dose response analysis based on unadjusted
doses. Another 45% (13,851) of 31,048 employment years during this period
had zero as the annual recorded dose. The following criteria were used to
select the subgroup of zero doses that would not be adjusted:
- If all annual recorded doses for a worker employed five or
more years were zero, then they remained zero; and
- If the department of employment for the majority of the working days in
a given year (most prominent department) had 75% or more annual recorded
doses of zero, then the dose remained zero.
Implementing these criteria resulted in approximately 20% of the zero doses
for 1944 to 1956 remaining unadjusted. Remaining employment-years were adjusted
upward by an algorithm developed after examining a sample of hardcopy records
from ORNL.
Back to TOC
Hardcopy Monitoring Records
Hardcopy radiation exposure records were maintained for all employees. Included
were all daily pocket ionization chamber (pocket meter) readings and weekly
film badge dosimetry readings, along with relevant information related to
interpretation or problems with data. The data were recorded in millirems
(mrem), where 1,000 mrem = 1 cSv. Copies of hardcopy dosimetry records for
211 employment-years from 1947 through 1955 obtained from ORNL were entered
into a computer file so that the data would be available for the adjustment
process. Eighteen other employment-years were rejected because the computerized
annual recorded dose differed by 10% or more from the sum of the weekly
film badge readings on the hardcopy record. Ninety of the 211 retained records
were from a random sample of the population of all employment-years through
1955. The remaining 121 were selected randomly from specific dose ranges,
since a random sample was likely to contain a large percentage of employment-years
with a zero dose. Two-thirds of these 121 were obtained from four specific
dose ranges up to 1.6 cSv per year. The remainder represented five dose
ranges with the lowest beginning at 1.5 cSv per year.
Figure 1 shows a hardcopy record in one of the formats used during these
years, while Table 1 defines the abbreviations used in the record.
Back to TOC
Figure 1
Table 1:
Definitions of Abbreviations
Used in Hardcopy Record
Weekly Totals Section
I. Pocket Meter (PM) Subsection
TSR - total significant reading
E - number of days with pocket meters
IRR - pocket meter irregularities
II. Film Badge (FILM) Subsection:
OW - open window reading
S - shielded reading
PME - probable maximum exposure for the film
IRR - film irregularities
NP - number of processes for film
Back to TOC
Pocket meters were assigned to workers in pairs because technical characteristics
of these devices made them susceptible to discharging and registering a
reading for reasons other than radiation exposure. The weekly pocket meter
reading, reported as TSR, was then obtained by adding the minimum of each
pair of daily readings. Not all of the hardcopy records had pocket meter
readings or even film badge readings. Eleven of them were blank, and nearly
all the others had weeks for which there were no film badge or pocket meter
readings. These blank periods reflected an early ORNL policy to make monitoring
available to everyone but only require monitoring devices to be worn by
workers entering areas with exposure potential (Watkins et al., 1993).
Other irregularities were noted, such as pocket meter readings off scale
or film badges damaged in processing. Some records had weekly readings of
30- (30 minus), indicating that the value was below the usual threshold
of 30 mrem (0.03 cSv). Such values had been taken as zeros in calculating
annual recorded doses.
Back to TOC
Use of Pocket Meter Data for Adjustment of 1944-1956 Doses
An annual film badge-pocket meter (FBPM) estimate of external dose from
hardcopy records was obtained by summing weekly shielded film badge readings
plus weekly pocket meter results, if available, for weeks with missing or
zero shielded film badge readings. When weekly pocket meter readings were
substituted for missing or zero film badge results, an upper limit of 0.06
cSv was imposed on the substituted value. The rationale was that the film
badge should have responded to an actual dose above that value. This limit
was an additional safeguard against elevated pocket meter results not attributable
to radiation exposure.
An annual recorded dose of zero was associated with 55 of the 211 hardcopy
record employment-years. Fifteen of these fifty-five annual doses were identified
as unexposed employment-years using the selection criteria listed above
in the section "Selection of Unexposed Employment-Years." Data
from these 15 employment-years were not used in the dose adjustment process
since they did not belong to the population to be adjusted. All fifteen
of these employment-years had a FBPM dose of zero, confirming that the selection
criteria were able to determine employment-years that were not associated
with radiation exposure.
The FBPM doses contained dose estimates based on readings from both personal
dosimeters, while annual recorded doses were based only on weekly film badge
readings. The amount of dose missed by weekly readings and recording practices
could be evaluated by subtraction. Therefore, the Estimate of Missed Dose
(EMD) was calculated as the difference between FBPM dose and annual recorded
dose.
Graphical inspection and smoothing techniques were used in selecting an
appropriate method of incorporating pocket meter data into the adjustment
procedure. Figure 2 shows the relationship between EMD and annual recorded
dose for the 196 potentially exposed employment years from the sample employment-years
with hardcopy records. Although there is a cluster near zero, most annual
recorded doses of zero had associated personal monitoring data indicating
some external radiation dose during the year. In addition, most annual recorded
doses above zero have monitoring data that demonstrate missing dose. A variable
span smoother (Friedman, 1984) was applied to the data using the S-PLUS
function, supersmu (Statistical Sciences, 1993) with the minimal smoothing
option. The resulting fit appears in dashed lines. It revealed that a reasonable
algorithm for adjusting annual recorded doses could contain two distinct
parts - a quadratic curve up to approximately 1.6 cSv, followed by a horizontal
line from 1.6 to 5 cSv.
A second degree polynomial was fit to doses below 1.6 cSv in the sample
using weighted least squares regression with EMD as the y-variate and annual
recorded dose as the x-variate. The weight for each observation was the
ratio of number of days worked in the year to 365. The regression yielded
the following formula for the estimate of missed dose in the sample:
where X is the unadjusted annual recorded dose in cSv.
The solid line shows this weighted least squares fit with the connecting
horizontal line. To obtain adjusted annual doses this formula was applied
to all employment-years with annual recorded doses less than 1.6 cSv that
were selected to remain zero. The placement of the horizontal line indicated
that 0.3 cSv should to be added to all annual recorded doses in the range
of 1.6 to 5 cSv. For years of partial employment, the EMD was prorated to
reflect the portion of the year worked. For approximately 5% of the doses,
the number of days worked during that year was unknown, so no dose was added.
Back to TOC
Figure 2
Back to TOC
Replacement of 1943 Unmonitored Employment-Years
The earliest assigned film badges at ORNL were for 1944. Annual recorded
doses for 1943 were based entirely on pocket meter readings. Because 1943
doses had been derived in a different manner, they could not be included
for adjustment with the doses from 1944 to 1956. However, workers who were
not monitored during 1943 were eligible to receive an estimated dose derived
from the algorithm described below in the section entitled "Replacement
of Missing Doses by the Nearby Procedure." In the case of the 1943
doses, no departmental medians were available because of sparse monitoring
data during this first year of plant operation. Therefore, the departmental
median section of the algorithm was not applied. All estimated doses were
prorated to reflect the portion of the year employed.
Table 2 gives the distribution for ORNL of the numbers of annual doses that
were adjusted or estimated and the methods by which they were obtained.
Back to TOC
Table 2
Number of Annual Recorded External Doses
and Unmonitored Years Adjusted for ORNL
1943 (682 possible annual doses)
Nearby Procedure 210
Plant median 97
1944-1956 (31,048 possible annual doses)
Set to zero by criterion 1(a) 631
Set to zero by criterion 2(b) 5,976
Adjusted by formulac,(d) 23,149
Adjusted by constant value(e) 1,147
1957-1984 (114,499 possible annual doses)
Nearby procedure 410
Annual department medians 96
Annual plant median 73
(a) Employed for five or more years with all annual recorded doses of zero.
(b) Department of employment for majority of annual working days had 75%
or more zero doses.
(c) Adjustment amount = 0.128 + 0.595X - 0.306X2, where X is annual recorded
dose in cSv.
(d) 9,322 of the doses adjusted by the formula were former zero doses.
(e) 0.3 cSv.
Back to TOC
ADJUSTMENTS TO Y-12 EXTERNAL DOSES
Background
Emphasis at Y-12 was placed on monitoring internal radiation exposure because
of the nature of production processes performed at this facility (Watkins
et al., 1993). Before 1961 the policy at Y-12 was to issue film badges only
to those employees who, based on their exposure potential, could exceed
10% of the occupational external radiation protection standard, which was
15 cSv per year at that time. This practice resulted in 80 to 95% (depending
upon the year) of employment years being unmonitored for external dose prior
to 1961. Because workers monitored with film badges at Y-12 before 1961
were selected according to their potential to receive at least moderate
external doses, 7,358 (14%) annual recorded doses that were based on these
film badge readings were not adjusted. It was reasoned that for these workers
the annual recorded dose would be the best estimate of dose available, particularly
since workers with higher exposure potential would be less likely to have
film badge readings below the dosimeter threshold. An exception to this
was the revision of 104 doses from 1948 and 1949, which were based on a
very few film badge readings during the years when the external monitoring
program was just beginning at Y-12. The revision of these doses is discussed
in the section entitled "Dose Adjustment for 1948 and 1949."
Back to TOC
Estimating Doses for Unmonitored Employment-Years, 1947-1960
An algorithm was developed to estimate doses for unmonitored employment-years
during this time period. It was applied to Y-12 members of the subcohort
used in dose-response analysis. This algorithm was based on work history
information for the subcohort, and it derived doses for unmonitored employment-years
from logically related monitoring data in an immediately following time
period. Because some workers likely received no occupational exposure, specific
employment-years were identified for which unmonitored values should be
replaced by an estimated annual dose of zero. These employment-years were
determined by the following criterion: If the most prominent department
for the year had fewer than 10% monitoring during that year, then the dose
estimate was zero. A zero dose estimate was applied to nearly 40% of unmonitored
employment-years in this period. This percentage was judged to be credible
because the nature of the work being conducted resulted in internally deposited
uranium being the primary source of radiation exposure.
For approximately 8% of employment-years before 1961, an unmonitored year
had an actual recorded dose or doses within two years. In this situation
the Nearby Procedure, as described in the section below, was implemented
to obtain dose estimates. Remaining unmonitored employment-years were assigned
doses of 0.045, 0.070 or 0.130 cSv, according to whether the most prominent
department for the year was classified as having lower, moderate, or higher
exposure potential, respectively. These values were selected as dose estimates
based on recorded doses from the 1961 to 1965 period, when essentially all
personnel were monitored. The 140 departments from 1961 to 1965 were ranked
from lowest to highest potential exposure by departmental median doses and
subdivided into three groups of approximately equal size. Medians of these
three groups were taken as dose estimates for the 1947 to 1960 period. All
dose estimates were prorated depending on the number of days worked during
the year.
Back to TOC
Dose Adjustment for 1948 and 1949
During 1948 and 1949 at Y-12, pocket meters were more commonly used than
film badges as dosimeters for monitoring external radiation exposure. Annual
recorded doses were adjusted because many weeks had no recorded results
and many others had assigned results of 30 mrem (0.03 cSv), which was considered
to be the minimum detectable limit for film badges at that time. Adjustments
to dose estimates were made by first adjusting weekly results and then summing
these to obtain the adjusted annual doses. Less than detectable, noncredible,
or absent weekly results were replaced by 0.009 cSv., which was derived
by probit analysis (Strom, 1983) and later verified to also be the weighted
mean of a subset of representative pocket meter readings and another subset
of credible film badge readings from 1948 and 1949.
All estimated doses were prorated to reflect the portion of the year employed.
Table 3 shows the numbers of Y-12 annual doses that were adjusted or estimated
and the methods by which they were obtained.
Back to TOC
Table 3
Number of Annual Recorded External Doses
and Unmonitored Years Adjusted for Y-12
1947-1960 (51,854 possible annual doses)(a)
Nearby 4,242
Set to zero(b) 20,671
Higher exposure departments(c) 5,677
Moderate exposure departments(d) 5,154
Low exposure departments(e) 8,648
Revised (1948 and 1949) 104
1961-1984 (121,069 possible annual doses)
Nearby 2,104
Annual department medians 52
Annual plant median 128
(a) 7,358 not adjusted because annual recorded doses were based on film
badge records that had been selected for monitoring because of higher exposure
potential.
(b) Most prominent department for the year had less than 10% monitored workers.
(c) Dose estimate was 0.130 cSv.
(d) Dose estimate was 0.070 cSv.
(e) Dose estimate was 0.045 cSv.
Back to TOC
REPLACEMENT OF MISSING DOSES BY THE NEARBY PROCEDURE
After 1956 at ORNL and 1960 at Y-12, fewer than 3% annual recorded doses
were missing for all years combined. To replace these missing values with
best estimates of doses, the Nearby Procedure was developed to obtain doses
based on an individual's doses for up to two years on either side of the
missing employment-year (Watson et al., 1994). This technique averaged available
doses if they appeared both before and after the missing year or copied
a dose when it appeared either only before or only after the missing year.
When no doses were within two years of the missing value, the dose estimate
was the median dose of all monitored workers in the most prominent department
for the year, provided at least 25 monitored doses were available for calculating
the median. When a departmental median could not be determined or the worker's
department was not known, the plant median for the year was used.
Back to TOC
RESULTS OF DOSE ADJUSTMENT ON CUMULATIVE DOSES
Because fewer than 3% of ORNL doses after 1956 and Y-12 doses after 1960
were adjusted, the impact of the dose adjustment procedures on cumulative
external doses used in dose response analysis can be appraised by examining
plots of cumulative dose distributions in earlier years. Figure 3 shows
unadjusted and adjusted ORNL doses accumulated through 1956, while Figure
4 displays unadjusted and adjusted Y-12 cumulative doses through 1960.
Figure 3 depicts a smooth, gradual shift toward higher values of cumulative
dose resulting from the dose adjustment process. Except for the relatively
large drop in the percentage of workers with a cumulative dose of zero through
1956 (from 31.9% to 11.7%), no dramatic changes were observed. With dose
adjustment, the cumulative dose mean increased from 1.08 to 1.63 cSv, and
the median increased from 0.13 to 0.46 cSv. The picture seen in Figure 4
is quite different. Because few Y-12 workers were monitored before 1961,
the adjustment process resulted in a notable decrease in percentage of workers
with a cumulative dose of zero (from 74.3% to 10.1%). However, even after
dose adjustment, most Y-12 workers still had cumulative doses through 1960
of less than 1 cSv, with 50% less than 0.2 cSv.
Comparing Figure 3 to Figure 4, it is apparent that, even after adjustment,
ORNL workers had generally higher cumulative external doses than Y-12 workers
over the first 14 years of each plant operation. This fact reflects the
difference in work activities and materials present in the two plants.
Back to TOC
Figure 3
Back to TOC
Figure 4
Back to TOC
DISCUSSION
To evaluate the potential effect of missing dose on dose-response coefficients,
a preliminary dose adjustment procedure was developed to assign adjusted
doses to specific individual employment years. This adjustment procedure
made use of personal external monitoring data currently computerized; historical
knowledge of evolving monitoring policies, practices, and dosimeters; knowledge
of operations and production processes at these facilities; and a sample
of pocket meter readings currently available only on ORNL hardcopy records.
Limited data exist to further refine the Y-12 dose adjustment process. However,
detailed hardcopy exposure records for approximately thirty thousand employment-years
before 1957 at ORNL contain daily pocket meter and weekly film badge data
that have not been computerized. More sophisticated and accurate adjustment
techniques could be implemented for future studies if these personal monitoring
data are computerized (Mitchell et al., 1993).
Back to TOC
REFERENCES
Friedman, J. H., A Variable Span Smoother. Technical Report No. 5, Laboratory
for Computational Statistics, Department of Statistics, Stanford University,
Stanford, CA, 1984
Frome, E. L., Cragle, D. L., Watkins, J. P., Wing, S., Shy, C., Tankersley,
W. G.,
and West, C. M., A Mortality Study of Employees of the Nuclear Industry
in Oak Ridge, Tennessee. In preparation, 1994
Kerr, G. D., Missing Dose from Mortality Studies of Radiation Effects among
Workers at Oak Ridge National Laboratory. Health Physics, Vol. 66, No. 2,
February 1994
Mitchell, T. J., Ostrouchov, G., Frome, E. L., and Kerr, G. D., A Method
for Estimating Occupational Radiation Dose to Individuals, Using Weekly
Dosimetry Data. Technical Report ORNL-6778, Oak Ridge National Laboratory,
Oak Ridge, TN, December, 1993.
Morgan, K. Z., Snyder, W. S., and Davis, D. M., Applied Health Physics and
Safety Annual Report for 1963. Technical Report ORNL-3665, Oak Ridge National
Laboratory, Oak Ridge, TN, 1964
Morgan, K. Z., Davis, D. M., and Hart, J. C., Applied Health Physics and
Safety Annual Report for 1968. Technical Report ORNL-4423, Oak Ridge National
Laboratory, Oak Ridge, TN, 1969
Statistical Sciences, S-PLUS User's Manual, Version 3.0, Seattle: StatSci,
a division of MathSoft, Inc., 1991
Strom, D. J., A strategy for Assessing Occupational Radiation Monitoring
Data from Many Facilities for Use in Epidemiologic Studies. Doctoral Dissertation,
University of North Carolina, Chapel Hill, N. C., 1983
Watkins, J. P., Reagan, J. L., Cragle, D. L., Frome, E. L., West, C. M.,
Crawford-Brown, D. J., and Tankersley, W. G., Data Collection, Validation,
and Description for the Oak Ridge Nuclear Facilities Mortality Study. Technical
Report ORISE 93/J-42, Oak Ridge Institute for Science and Education, Oak
Ridge, TN, October, 1993
Watson, J. E., Wood, J. L., Tankersley, W. G., and West, C. M., Estimation
of Radiation Doses from Workers without Monitoring Data for Retrospective
Epidemiologic Studies. Health Physics, In Press.
Wing, S., West, C. M., Wood, J. L., and Tankersley, W. G. Recording of external
radiation exposures at Oak Ridge National Laboratory: implications for epidemiologic
studies
JEAEE, 1994
Back to TOC
Back to Title Page
Return to CER Home Page