Return to the Table of ContentsMortality by Occupation, Industry, and Cause of Death: 24 Reporting States (1984-1988)
DHHS (NIOSH) PUBLICATION
NO. 97-114 JUNE 1997
Results and Discussion
Detailed results are presented in Tables B-1 through B-16, found in the zipped file included in Appendix B. These include the number of observed deaths and PMRs for the three age groups: aged 20 and over, aged 20 to 64, and aged 65 and over. The tables show different combinations of occupation and cause of death, or industry and cause of death, depending on three criteria: (1) the PMR for the aged-20-and-over group was greater than or equal to 120, (2) the lower limit of the 95% CI of the PMR for the aged-20-and-over group exceeded 100, and (3) the observed number of deaths in the aged-20-and-over group was greater than or equal to 10. Space limitations preclude the display of all PMRs, but they are available upon request.
All PMRs are also available on the CDC WONDER Page . Use the Searches and Queries function, log in to WONDER, and select New Query. Select NIO1 for industries, occupations, and causes of death listed alphabetically or NIO2 for industries, occupations, and causes of death listed according to their classification systems. Race- and sex- specific tables for an industry or occupation by all causes of death or a cause of death by all occupations or industries will be produced.
Four tables are shown for each race-sex group (see Appendix B for a complete list of the detailed tables in the zipped file):
In Tables 2-9 , to focus on associations that make substantial contributions to overall mortality and to de-emphasize the importance of chance or rare events, a minimum number of deaths per cause of death and occupation or industry combination was established. Similarly, to ensure the stability of the PMR estimates, each PMR included in these tables was required to have a minimum lower limit of the 95% CI. This minimum will eliminate highly elevated PMRs based on a small number of deaths. For each cause of death, the occupation or industry with the highest PMR meeting these criteria is shown in Tables 2-9. The minimum number of deaths and lower confidence limit vary according to sex and race to take into account the differences in the total number of deaths in each race-sex group in the data set. For white males the minimum number of deaths was 50 and the minimum lower 95 percent confidence limit was 150; for black males and white females, the criteria was 20 and 125; and for black females the criteria was 10 and 125. While Tables 2-9 show the occupation or industry with the highest PMR for selected causes of death, Tables B1-B16 show many other very interesting findings.
For black males, the highest PMR for each cause of death is shown in Table 3. Fewer of these associations can be classified as previously recognized: (1) coal workers pneumoconiosis in mining machine operators and (2) struck accidently by a falling object in forestry and logging occupations. Other associations have been reported previously in surveillance reports but not substantiated; for example, cerebrovascular diseases in farmers, except horticulture (seen in males in Great Britain,4 British Columbia,5 New York State,12 and Washington State19) and chronic obstructive pulmonary disease and allied conditions in molding and casting machine operators (which was elevated in Great Britain4). An association that may be due to lifestyle or socioeconomic factors or to chance is malignant neoplasm of the colon in sales occupations and personal goods and services.
For white females, the highest PMR for each cause is shown in Table 4. Few of the associations shown in this table have been previously recognized. Instead, two occupations experienced excess deaths for a number of causes which may indicate a need for further study of these groups or for targeted health promotion activities. Teachers, except postsecondary (the largest occupation group for white females) have the highest PMRs for four causes of death: (1) malignant neoplasm of other parts of the uterus, (2) Parkinson s disease, (3) anterior horn cell disease, and (4) multiple sclerosis and other demyelinating diseases of the central nervous system. Two of these associations support results from other studies. An elevated risk for Parkinson's disease in female teachers was reported in Washington State.19 As seen in Table 2, white male teachers in this study also had a highly elevated PMR for Parkinson's disease. Elevated risks for multiple sclerosis were reported in Washington State19 and Great Britain.4 Waitresses experience the highest PMRs for five causes of death: (1) malignant neoplasm of the lip, oral cavity, and pharynx, (2) malignant neoplasm of the esophagus, (3) malignant neoplasm of the trachea, bronchus, and lung, (4) alcohol-associated diseases, and (5) chronic obstructive pulmonary diseases and allied conditions. Smoking and alcohol consumptions are major risk factors for these causes of death, and the table illustrates the effect of lifestyle in this analysis and the need for health promotion programs for this occupational group.
For black females, the highest PMR for each cause is shown in Table 5. As with the white females, these are not recognized occupational associations. Two of the associations are similar to the results in other race-sex groups: the PMR for malignant neoplasm of the breast is elevated in teachers, except postsecondary, as it was for white female teachers, postsecondary; and the PMR for cerebrovascular diseases in farmers, except horticultural, is elevated for both black males and females. This may indicate a need for followup.
For black males (Table 7), the industry results similar to the occupation results include (1) human immunodefiency virus infection in workers in colleges and universities, (2) malignant neoplasm of the prostate in workers in elementary and secondary schools, (3) mental disorders in workers in horticultural services, (4) cerebrovascular diseases in agricultural production workers, (5) coal workers' pneumoconiosis in coal mining workers, and (6) struck accidently by falling object in logging workers.
For white females (Table 8), the industry results are quite different from those in the occupation analysis. Only malignant neoplasm of the cervix uteri in workers in eating and drinking places and anterior horn cell disease in workers in elementary and secondary schools are similar to the occupation results. Two of the associations have been reported previously: malignant neoplasm of the breast in workers in religious organizations12 and homicide in grocery store employees.19
As with the white females, few of the results for black females (Table 9)are similar to the occupation results. Those similar are malignant neoplasm of the breast in workers in elementary and secondary schools, alcohol-associated diseases in workers in private households, and cerebrovascular diseases in workers in agricultural production, crops.
The PMRs presented in this report can be used in several ways. They may be evaluated by researchers and used as leads for further study. They may be used as additional information in the evaluation of previously hypothesized associations. They may identify new occupations and industries not previously recognized as experiencing an excess risk for a known occupational disease. They may be used to target health promotion and intervention activities to the appropriate workers for non-occupational disease.
A statistically significant elevation of a PMR cannot be interpreted directly as indicating a causal relationship between the industry or occupation and the cause of death. Since a very large number of PMRs were tested for statistical significance, many of the elevated PMRs will occur due to chance. Other elevated PMRs will be due to confounding factors. Cigarette smoking, which is more prevalent among blue collar workers compared with white collar workers in the United States,24 is potentially a strong confounder for a number of causes of death, including malignant neoplasms of the lung, larynx, and bladder, and ischemic heart disease. Alcohol consumption is a potential confounder for liver disease and malignant neoplasms of the mouth, pharynx, esophagus, and larynx. Socioeconomic factors such as availability of health care and diet are also potential confounders for a number of diseases.
Individual PMRs must be evaluated to determine which associations are likely to reflect cause- and-effect relationships.25 Several questions should be asked about the association:
Several studies have compared the death certificate information about occupation and industry with employment information from interviews conducted before death or interviews with next-of- kin.26,27,28,29,30 Except for one,27 the studies are small, with fewer than 400 comparisons. They are all limited geographically, and some are limited to specified causes of death. For white males, agreement for occupation ranged from 53% to 69%; and for industry, it ranged from 62% to 84%. Percent agreement for white females was a little higher than for white males. In one study of black males and females,27 the percent agreement for black females was similar to that for white females. The percent agreement for black males for occupation was 35%, and for industry it was 69%.
Another limitation is the quality of cause-of-death information reported on the death certificate. The medical certification of cause of death is made only by a qualified person, usually a physician, a medical examiner, or a coroner. The reliability and accuracy of cause-of-death statistics are, to a large extent, governed by the ability of the certifier to make the proper diagnosis and by the care with which he or she records this information on the death certificate. A number of studies31 have focused on the quality of this information by comparing the specified underlying cause of death to various sources, including autopsy reports and hospital records. The studies of the quality of the cause-of-death information have limitations similar to those found in the studies of employment information quality. Most of the studies are small, the alternative source of information may not be accurate, and there is no standard definition of agreement. In general, the accuracy of the certification differs among disease outcomes. A small study comparing the death certificate information with autopsy data32 found an 87% agreement for neoplasms, 82% for circulatory diseases, 50% for digestive diseases, and 33% for respiratory diseases. A second autopsy study33found an overall concordance of 85%. This ranged from about 95% concordance for neoplasms and vascular diseases to 88% for digestive and 76% for respiratory diseases.
For accidental deaths, the information about injury at work was not routinely coded for NCHS before data year 1993. The NIOSH Division of Safety Research has constructed a database that contains information from death certificates with an indication that the injury occurred at work.6The NTOF database contains records from all U.S. States from 1980 through 1993 for decedents aged 16 or older with an external cause of death and a Yes response to the Injury at work item on the death certificate. These data more accurately identify occupations and industries with a high risk of death due to work-related injuries than do the data in this report.
Death certificates have little information about potential confounding factors, including tobacco and alcohol use and socioeconomic status. Tobacco and alcohol use are known to be more highly associated with some occupations than with others.24,34 Socioeconomic status is usually determined by income and years of education, which are also associated with occupation. Without this information, it is not possible to control in the analysis for these possible confounders, resulting in spuriously elevated or decreased PMRs. However, studies comparing crude risk estimators with smoking-adjusted estimators suggest that the absence of information about smoking will seldom lead to serious errors in risk estimation.35,36 The lack of information about confounders is starting to be partially addressed. The Standard Certificate of Death, as revised in 1989, now collects information about the education of the decedent. The coded information was added to the 1989 mortality files for 21 States. Four States have added items collecting information about the contribution of tobacco use to the death.
Several authors have discussed the advantages and limitations of the PMR method.37,38,39,40 Briefly, the advantages are as follows: (1) the population at risk is not needed to calculate the PMRs and, therefore, additional misclassification is not introduced by using a different source of occupation information in the denominator, and (2) the computer programming is relatively simple and inexpensive. The main limitation is that without a population at risk, it is not possible to estimate death rates. The PMR indicates only whether the age-standardized proportion of deaths from a specific cause appears to be higher or lower than the expected proportion for a particular occupation or industry. The PMR will be a poor estimate of the risk of death if the population-based SMR for all causes for an occupation or industry group is greatly above or below 100. For example, the SMR is often affected by socioeconomic factors, with high socioeconomic groups having lower all-cause SMRs than low socioeconomic groups.41 Therefore, a high socioeconomic occupation could have elevated PMRs for causes of death for which the death rates themselves were not elevated. The PMR may also be misleading if the death rate for a major cause of death is much higher or lower than expected in the occupation or industry. For example, if persons are selected into an occupation according to health status or level of fitness, the mortality risk for cardiovascular disease is likely to be low. This may cause a spurious increase in the PMRs for other causes of death.