Occup. Med. Vol. 50, No. 2, pp.

141-145, 2000
Copyright © 2000 Lippincott Williams & Wilkins for SOM
Printed in Great Britain. All rights reserved
0962-7480/00

OCCUPATIONAL Occupational epidemiology

MEDICINE IN THE
21ST CENTURY T. L. Guidotti
University ofAlberta Faculty of Medicine, Edmonton, Alberta, Canada

The epidemiological literature for assessing risk in many, if not most, modern
occupations has now become sufficiently obsolete that it can no longer be depended
upon to guide either prevention or adjudication of compensation. This obsolescence
must be dealt with by developing new sources of information pertinent to occupational
hazards and the risks associated with most occupations. Ideally, a comprehensive

Downloaded from https://academic.oup.com/occmed/article/50/2/141/1396507 by guest on 03 November 2021

surveillance mechanism that would be automatically updated for the changing risk in a
changing economy would be ideal and may be attainable with further developments in
health information technology. The characteristics of such a system are described.
However, there are many obstacles to such a system which appear insurmountable in
the short term. A more eclectic plan for cooperation and data-sharing would help in the
short term and would establish a pattern of collaboration that could both place
adjudication on a more solid foundation and avoid allegations of collusion in business.
The general outline for a practical programme of collaboration along these lines is
presented.

Key words: evidence-based medical dispute resolution; occupational epidemiology;

occupational health and safety; occupational medicine; workers' compensation.

Occup. Med. Vol. 50, 141-145, 2000

INTRODUCTION • identifying possible aetiological mechanisms,

Occupational medicine depends on a vast and sophisti-
cated database of epidemiological information to inform
decisions and conclusions based on causation.1 How-
ever, the epidemiological basis for assessing risk in many,
if not most, modern occupations has now become
sufficiently obsolete that it can no longer be depended
upon to guide either prevention or adjudication of
compensation. This obsolescence must be dealt with by
developing new sources of information pertinent to
occupational hazards and the risks associated with most
occupations.
This problem is more serious that is generally realized.
Occupational epidemiology is critical to the practice of
occupational medicine in many ways:
• supporting prevention-related interventions,
• identifying occupational hazards.
If this vast database is obsolete, the practice of
occupational medicine will eventually be misled. In
particular, the critical functions for the resolution of
evidence-based medical disputes, which now act to
protect the interests of both worker and employer, will
become disconnected from reality. Scheduled diseases
and systems based on presumption may find their
assumptions to be out of date. Claims for disability or
mortality related to previously unrecognized risks may be
unfairly rejected. Decisions on individual cases which are
based on an interpretation of the literature will become
increasingly difficult to defend.

• evidence-based medical dispute resolution (such as

adjudication in workers' compensation, third-party
litigation and insurance settlements),2
• setting priorities in occupational health and safety
practice,
• designing periodic health surveillance protocols,
• supporting worker education,
Correspondence to: T. L. Guidotti, Department of Environmental and
Occupational Health, School of Public Health and Health Services, The
George Washington University Medical Center, 2300 K St., NW, Ste. 2ol,
Washington DC 20037, USA. Tel: +1 202 994 1765; Fax: +1 202 994
0011; e-mail: eohtlg@gwumc.edu
THE OBSOLESCENCE OF THE BODY OF
KNOWLEDGE
In the early years of the 20th century, when chronic
disease epidemiology was still evolving and most
occupational health studies were anecdotal reports or
case series, exposure levels were relatively high. The
occupations documented to be at greatest risk tended to
be those that were historically recognized as hazardous,
such as mining or forest harvesting. The introduction of
modern methods of epidemiology allowed the detection
of increasingly subtle risks and their documentation for
142 Occup. Med. Vol. 50, 2000
purposes of control and compensation. This led to the
documentation of contemporary hazards, such as asbes-
tos and vinyl chloride, and permitted a body of literature
to develop which largely reflected the experience of
workers who sustained their exposures in mid-century.
Following the pioneering studies of occupational
mortality and morbidity in the 1960s, the 1970s were a
sort of 'golden age', especially for studies of cancer and
for the application of both cohort and case-referent
study designs. Cross-sectional studies largely passed
from the scene because they cannot infer causation and
are prone to bias. However, cohort studies, especially, are
expensive and the current methods of chronic disease
epidemiology required persona] identifiers, which made
them difficult to reconcile with privacy legislation in
Europe. Funding declined in the 1980s and occupational
epidemiology became targeted more narrowly to answer
specific questions. The era of the large-scale 'fishing
expedition', which generated numerous hypotheses but
was prone to report associations arising by chance alone,
was largely over.
Hypothesis-testing studies, to link exposure with
outcome and to suggest mechanism, have been con-
ducted on occupations with high visibility or economic
importance, such as firefighters and machinists. How-
ever, these studies have not begun to explore all the
possible associations hinted at in the surveillance
literature. Associations suggested but not proved by the
extant literature mostly have been investigated piece-
meal, not in a systematic fashion with due attention to
changing risk profiles. In addition, they have seldom
examined exposures that may lead to future evaluation of
risks that may be anticipated as a result of changing work
practices.
In the normal course of events, when occupations
would show stability in employment and would change
only slowly in their exposure patterns, this approach
would eventually have sketched in a fair assessment of
the risk associated with most occupations. However,
these occupations have not stood still. The profile of risks
has changed markedly in some occupations; for example,
machine oils that could well have caused an increased
risk of cancer among machinists have been replaced by
other formulations for many years.
The obsolescence of the general body of knowledge
on occupational risks complicates the problem of
assessing work-relationship and compensation. Under
the Industrial Injuries Scheme or the various other extant
workers' compensation or work injury insurance systems
there is an absolute need for empirical information
germane to current hazards at sufficient levels of
resolution to apply to individual cases. A key aspect of
the adjudication is to determine whether a common
condition, such as lung cancer, is more likely than not the
result of an occupational exposure.1 (When an associa-
tion is such that more than half the cases among workers
in an occupation can be attributed to that occupation,
this is the basis for presumption and scheduling
conditions as work-related.) The identification of a
possible association, therefore, is not sufficient; some
indication of magnitude compared to the frequency of
the condition in the general population is required. The
existing information base derived from occupational
epidemiology is rapidly becoming obsolete for this
purpose.
CONDITIONS IN THE DEVELOPED WORLD
Working conditions in the developed world, including
the European Community, North America, Japan,
Australia and isolated enclaves elsewhere, are not ideal.
There is considerable room for improvement and there
are enough cases of ignorance, cupidity and neglect to
ensure a need for vigorous enforcement of occupational
Downloaded from https://academic.oup.com/occmed/article/50/2/141/1396507 by guest on 03 November 2021
health and safety standards for generations to come.
However, the workplace of 2000 in the developed world
generally presents hazards of a different magnitude from
the workplace of 1950.
Notwithstanding remaining problems, the workplace
in developed countries has changed drastically. New
processes have profoundly altered the risk profile of these
historical occupations. New occupations now exist in
industries that, in some cases such as biotechnology,
have never been comprehensively studied or that, as in
the case of semiconductor manufacture or the formula-
tion of composite materials, changes rapidly and for
reasons mostly unrelated to health. Even traditionally
hazardous industries for which control options are
limited, such as sawmills and firefighting, have been
made safer by personal protection, improved equipment
and better management practices. Indeed, for industry as
a whole the provision of a safe workplace has been
accomplished to the extent that egregious lapses are
unusual and all the more culpable.
Accompanying this profound change in the workplace
environment has been a change in the working popula-
tion. Figure 1 presents the standardized mortality ratio
for firefighters in two western Canadian cities by decade
from the 1930s to the 1980s, together with a plot of
observed deaths (among firefighters) compared to
expected deaths (based on the general population),
taken from a recent cohort mortality study.3 It is clear
that in the 1940s the firefighters lost most of their
comparative mortality advantage and so the healthy-
worker effect was much reduced. This reflects a
convergence in health status between an employed and
necessarily fit population and the general population.
The general population shows less difference in fitness
compared to employed persons than in the past.
The worker in a developed country today is much
healthier, as reflected in a majority of health indices, than
his or her generational counterpart of the past, due to
improved living conditions and healthcare. Many of the
preplacement screening examinations were once per-
formed on workers to detect common chronic diseases
such as tuberculosis; for example, the chest examination
and the routine chest film. These ceased to be cost-
effective long ago. In time they will be replaced by
straightforward tests of fitness to work and to accomplish
job-related tasks. This has occurred already in certain
industries, but the testing technology is still primitive. As

Figure 1. Standardized mortality ratio, by decade, for firefighters in two Canadian cities, Edmonton and Calgary, from 1927 to 1987, showing
reduction in the healthy-worker effect and convergence of mortality experience with the general population.
10 -
1927-29 1930-34 1935-39 1940-441945-49 1950-54 1955-591960-64 1965-69 1970-74 1975-79 1980-84 1985-87
Year
predicted over a decade ago, the primary determinant of
health status among the working population appears to
relate to recruitment and retention of workers of various
ages and social classes rather than specific employment-
related factors.4
Given the levels of exposure extant in developed
countries, the classical form of periodic health surveil-
lance has become of little value. Today the predictive
value of many routine screening tests is so low that there
is little point in continuing to incur the expense. These
tests are still valuable for high-risk populations as a
backstop against ineffective prevention, but their appli-
cation across entire industries is of questionable value.
Likewise, occupational hygiene testing was extensively
cut back by industries during the last round of economic
rationalization. One may predict that in future the
increasing use of biological exposure indices, sophisti-
cated biomarkers and other indicators of personal
exposure or risk will come to replace routine environ-
mental measurements except for purposes of audit,
quality assurance and health hazard investigation.
There are grave concerns on the part of many
professionals, this author among them, that there is a
risk of hazards creeping back into the workplace through
neglect and lack of recognition. For now, however, the
situation appears to be no worse than it used to be and
may be much better.
T.L. Guidotti: Occupational epidemiology 143
Downloaded from https://academic.oup.com/occmed/article/50/2/141/1396507 by guest on 03 November 2021
AN OLD SOLUTION FOR A NEW PROBLEM
The solution to this problem may be to revive a type
of study that has long gone out of favour in
occupational epidemiology: population surveillance.
These studies have been discussed extensively else-
where.5 These studies have a venerable history,
especially in the United Kingdom where the British
Registrar-General's Decennial Reports were once the
standard resource for assessing risk of mortality by
cause associated with occupation. In North America, a
number of studies of shorter duration were conducted,
mostly at the state level, pioneered by Milham in the
state of Washington. Massachusetts, California and
other states were studied but the Washington studies
were the template for them all. Earlier efforts were
directed toward the US population as a whole,
including a massive three-way analysis by Guralnick,
which was unfortunately limited by restricted coverage
of health outcomes, and a study on disability using
Social Security data.
These surveillance data resources were immensely
useful. They suggested hypotheses for testing by specific
studies and in the absence of a targeted study they
provided at least partial guidance for the assessment of
health risks. However, they had other highly desirable
features, among them:
144 Occup. Med. Vol. 50, 2000
• they defined a general population that was compar-
able to the working population;
• they presented the data on risk in a way that could be
interpreted as an elevation above the level of the
general population;
• they were longitudinal in design and could, in theory,
demonstrate trends over time or adjust risk estimates
as they changed;
• they made data accessible to any user through
technical reports.
These studies were casualties of decreasing budgets
and declining interest. They were subject to all the usual
biases of misclassification3 made worse by inaccurate
reporting of occupation on the death certificate among
the mortality studies. In genera], these biases tended to
underestimate risks. The problem of multiple compar-
isons also made individual findings suspect unless they
were consistent with other such studies or confirmed by
targeted investigations. The emphasis on mortality made
interpretation difficult for disorders that were disabling,
but seldom fatal. Even so, the advantages of these
surveillance studies were considerable. Mortality by
cause of death was by far the most common endpoint
because, for all the uncertainties in death certification, it
was a more reliable description than morbidity and, for
serious illness, was a useful surrogate for incidence or
morbidity. They could even be adjusted for known
smoking prevalence among occupational groups, as
demonstrated in one data series.6
What is needed today is a database, much like that of
the occupational surveillance studies, on a larger
population base. Beebe made this recommendation in
1983.7 A programme in the US supported by the
National Cancer Institute, the National Institute of
Occupational Safety and Health, and the National
Center for Health Statistics was begun in 1984.8 It
covers 24 states; although there is some diversity in
coverage, it is not complete as an industrial profile of the
country, and omits some important industrial states with
transitional economies, such as California, New York,
Texas, Illinois and Maryland. It is limited to mortality.
For now, however, this and the better-known SENSOR
system9 for occupational injury reporting are the most
credible surveillance systems available for the general
industrial population.
A system of on-going surveillance should be con-
tinuously renewed so that longitudinal trends can be
tracked. Ideally, it would have the following character-
istics:
• comprehensive coverage of a defined population, with
the capacity to capture individual demographic
characteristics;
• reasonably accurate determination of diagnosis to
support reasonably certain estimates of incidence,
morbidity and mortality;
• accurate reporting of both occupation and industry;
• accessibility to users, such as workers' compensation
agencies;
• audit and quality assurance;
• linkage capability, to allow linkage to other data sets,
such as cancer registries and health care utilization
data;
• strict confidentiality with respect to individual identi-
fiers;
• compatibility with extant coding systems and conven-
tions, such as ICD-10;
• continuous reporting for on-going surveillance, with
automatic data updating and reporting;
• interpretive features to ensure valid use of the data.
Additional features that would be desirable but
probably not practical would be some surrogate measure
of individual exposure level (such as duration of
Downloaded from https://academic.oup.com/occmed/article/50/2/141/1396507 by guest on 03 November 2021
employment), impairment level for living subjects (to
support estimates of disability) and multiple occupations
and avocational activities. Capturing individual exposure
levels would not be necessary because a job exposure
matrix based on hygiene data could also be built into the
system and would provide a valid estimate for a
population. Obviously the practicality of adding these
features is limited by the degree of intrusiveness and cost.
Data linkage on this magnitude is technically possible
but beset by practical problems and issues of public
policy. Such a comprehensive system may not be
possible today, particularly under privacy legislation in
many countries in Europe and North America. However,
it should be possible to develop such a system in a few
jurisdictions to serve as models and as data resources.
Indeed, data linkages such as those described are already
under discussion in Canada to support administration
and resource utilization among the provincial health
insurance schemes and the newly regionalized health
authorities in several provinces. Workers' compensation
agencies have not been a major part of these discussions
despite their obvious interests.
A PARTIAL SOLUTION FOR IMPERFECT
TIMES
Given the large issues and substantial obstacles that
would block early development of such a system, one
may propose an interim arrangement to at least improve
the quality of information currently used as the basis for
adjudication and social decision-making involving in-
jured workers:
• data sharing among workers' compensation and
industrial insurance agencies on an on-going basis
through a central statistical clearing house;
• actuarial analysis of pooled claims data on a much
larger scale, beyond the scope of any one scheme or
insurance carrier;
• analysis of the data, taking into account employers'
performance and socio-economic status (because the
business of insurance carriers is often segmented to
different levels of the market, reflecting characteristics
important in assessing risk);
• changes in legislation to declare these data a public
resource to be accessed in the public interest;

• protection for competitiveness among insurance
carriers, with universal access to the pooled data
(forcing competitiveness on the basis of efficiency
rather than proprietary data);
• access to these data, by legitimate investigators from
academia, industry and labour (but not to data
specific to one company);
• support by industrial insurance schemes, perhaps
through payment into a pooled fund, for an on-going
programme of surveillance, cross-jurisdictional (and
international) comparisons and selected targeted
studies;
• use of these data in setting priorities among occupa-
tional health agencies such as the Health and Safety
Executive (HSE) and Occupational Safety and Health
Administration (OSHA) (notwithstanding that other
considerations also affect priorities).
At the very least, workers' compensation and disability
insurance data can be used more effectively in the
assessment of occupational risks. For example, Gold-
smith has demonstrated that data on silica-exposed
workers from workers' compensation data in California
yielded estimates of risk for lung cancer similar to those
of more conventional cohort studies.10 These data are
not easy to work with and have problems of bias. These
problems should be a challenge to epidemiologists rather
than a reason for neglecting this rich source of data.
One option that is not recommended as a remedy is
meta-analysis. Re-analysis of existing published studies,
with all the limitations of that method,11 is likely to
further confuse the picture rather than clarify the
contemporary risk profile. Meta-analysis has serious
drawbacks even when sufficient studies of reasonable
quality do exist. It is unlikely to be of assistance where
data are lacking or unreliable.
In summary, the excellent foundation in occupational
epidemiology on which our occupational health and
safety and compensation structures are based is no
longer timely or adequate. Ideally, a comprehensive
surveillance mechanism that would be automatically
updated for changing risk in a changing economy would
be ideal and may be attainable with further developments
in health information technology. However, there are
many obstacles to such a system. A more eclectic plan for
T.L. Guidotti: Occupational epidemiology 145
cooperation and data-sharing would help in the near-
term and would establish a pattern of collaboration that
could both place adjudication on a more solid founda-
tion.
REFERENCES
1. Guidotti TL. Applying epidemiology to adjudication. Occ
Med, State of the Art Review 1998; 13: 303-314.
2. Guidotti TL. Evidence-based medical dispute resolution in
workers' compensation. Occ Med, State of the An Review
1998; 13: 289-302.
3. Guidotti TL. Mortality of urban firefighters in Alberta,
Downloaded from https://academic.oup.com/occmed/article/50/2/141/1396507 by guest on 03 November 2021
1927-1987. AmJIndMed 1993; 23: 921-940.
4. Andersson NR, Schilling RSF. Occupational epidemiology
in developing countries. In: Jeyaratnam J, ed. Occupational
Health in Developing Countries. Oxford: Oxford Medical
Publications, 1992; pp. 169-200.
5. Guidotti TL, Baser M, Goldsmith JR. Comparing risk
estimates from occupational disease monitoring data. Publ
Health Rev 1987; 15: 1-27.
6. Beaumont JJ> Singleton JA, Doebbert G, Riedmiller KR,
Brackbill RM, Kizer KW. Adjustment for smoking, alcohol
consumption, and socioeconomic status in the California
Occupational Mortality Study. Am J Ind Med 1992; 21:
491-506.
7. Beebe GW. Prospects for national data on occupational
mortality. In: Chiazze L, Lundin FE, Watkins D, eds.
Methods and Issues in Occupational and Environmental
Epidemiology. Ann Arbor, MI: Ann Arbor Science, 1983;
pp. 75-82.
8. Burnett C, Maurer J, Dosemeci M, Rosenberg HM.
Mortality by Occupation, Industry and Cause of Death: 24
Reporting States, 1984-1988. Vital and Health Statistics
Report. Washington DC: US Government Printing Office,
1997.
9. Maizlish N, Rudolph L, Dervin K, Sankaranarayan M.
Surveillance and prevention of work-related carpal tunnel
syndrome: an application of the sentinel events notification
system for occupationalrisks.AmJIndMed 1995; 27: 715—
729.
10. Goldsmith DF. Use of workers' compensation data in
epidemiology research. Occ Med, State of the An Review
1998; 13: 389^116.
11. Myers JE, Thompson ML. Meta-analysis and occupational
epidemiology. Occ Med 1998; 48: 99-101.