Use of death certificates in epidemiological studies, including occupational hazards: discordance with clinical and autopsy findings.

There has long been evidence of frequent inaccuracy of death certificates, with significant discordance between such designations and clinical and autopsy data. This exists for occupational diseases as well. The use of statistical rates based on death certificates has been seriously questioned despite their utility for total mortality. Programs to supplement death certificate data, particularly in occupational disease studies, may be helpful, and are reviewed.     

Causes of death in Huntington disease as reported on death certificates

Huntington disease (HD) is a late onset autosomal dominant neurological disorder. Two hundred fifty-three death certificates of HD-affected individuals from four midwestern states were examined to determine the completeness of reporting HD on the death certificates. Overall, 66% of death certificates indicated HD as a primary or contributory cause of death. There was significantly better reporting on more recent death certificates, but even since 1979, 16% did not report HD. The implications to those researching HD family histories and to the accuracy of mortality rates are discussed.     

The accuracy of liver cancer as the underlying cause of death on death certificates

Studies of liver cancer mortality are subject to confusion attributable to the changes in categories by which liver cancer is identified in successive revisions of the International Classification of Diseases. To determine the effects of these changes, diagnoses of 2,388 cases of primary liver cancer in the years 1973-80 were compared to the underlying causes of death recorded on the death certificates, using data from the National Cancer Institute's Surveillance, Epidemiology, and End Results Program. Results showed that only 53 percent of the deaths were attributed on death certificates to primary liver cancer. In a reverse comparison of 2,977 death certificates from the years 1973-85 with an underlying cause of death of primary liver cancer, 83 percent had been diagnosed as liver cancer. However, among the certificates that specified cancer of the liver, not specified as primary or secondary, as the cause of death, only 40 percent had been diagnosed originally as liver cancer. The mortality of liver cancer can be either underestimated or overestimated depending on which disease classification categories are used.     

Occupation as a risk identifier for breast cancer.

OBJECTIVES. Breast cancer mortality may be reduced if the disease is detected early through targeted screening programs. Current screening guidelines are based solely on a woman's age. Because working populations are accessible for intervention, occupational identification may be a way of helping to define and locate risk groups and target prevention. METHODS. We used a database consisting of 2.9 million occupationally coded death certificates collected from 23 states between 1979 and 1987 to calculate age-adjusted, race-specific proportionate mortality ratios for breast cancer according to occupation. We performed case-control analyses on occupational groups and on stratifications within the teaching profession. RESULTS. We found a number of significant associations between occupation and frequency of breast cancer. For example, white female professional, managerial, and clerical workers all had high proportions of breast cancer death. High rates of breast cancer in teachers were found in both proportionate mortality ratio and case-control analyses. CONCLUSIONS. These findings may serve as in an aid in the effective targeting of work-site health promotion programs. They suggest that occupationally coded mortality data can be a useful adjunct in the difficult task of identifying groups at risk of preventable disease.     

山东省1970-2005年食管癌死亡率变化趋势分析

目的分析山东省1970-2005年食管癌死亡率随年龄及时间的变化趋势。方法利用1970-2005年4次全死因调查资料,计算食管癌死亡率和标化死亡率,利用Joinpoint回归软件分析食管癌死亡率的升降速率。结果 1970-2005年间食管癌标化死亡率表现总体呈下降趋势,1970-1987,1987-1990和1990-2005年3个时期中的年均变化率分别为-0.61%(P0.05),2.33%(P0.05)和-1.42%(P0.05);食管癌死亡率随年龄的增长而升高,在各个年龄段上男性死亡率高于女性。结论山东省食管癌标化死亡率在该期间基本呈下降趋势,与食管癌高发区相似。     

An audit of a British sample of death certificates in which anorexia nervosa is listed as a cause of death

OBJECTIVE: Anorexia nervosa is associated with an increased mortality rate. National mortality statistics based on statutory death certification are potentially an important source of information. However, there are reasons to believe that these statistics may be subject to significant errors. An audit of the quality of information and diagnosis was conducted on death certificates in which anorexia nervosa was mentioned. METHOD: The current study examined data from death certificates of people who died in England and Wales between 1993 and 1999. RESULTS: There were 230 such deaths, but only 128--just over one half--were rated as likely to be deaths associated with true anorexia nervosa. DISCUSSION: National mortality statistics derived from death certificates are a flawed source of information on deaths from anorexia nervosa when taken at face value. There may be both underreporting and overreporting. Detailed examination may improve their usefulness by reducing the overerreporting. It seems likely that the association of deaths with anorexia nervosa is systematically underreported.     

Estimating cancer mortality rates from SEER incidence and survival data.

A method to estimate site-specific cancer mortality rates using Surveillance, Epidemiology, and End Results (SEER) Program incidence and survival data is proposed, calculated, and validated. This measure, the life table-derived mortality rate (LTM), is the sum of the product of the probability of being alive at the beginning of an interval times the probability of dying of the cancer of interest during the interval times the annual age-adjusted incidence rate for each year that data have been collected. When the LTM is compared to death certificate mortality rates (DCM) for organ sites with no known misclassification problems, the LTM was within 10 percent of the death certificate rates for 13 of 14 organ sites. In the sites that have problems with the death certificate rates, there were major disagreements between the LTM and DCM. The LTM was systematically lower than the DCM for sites if there was overreporting on the death certificates, and the LTM was higher than the DCM for sites if there was underreporting. The limitations and applications of the LTM are detailed.     

Use of multiple cause of death data in cancer mortality analyses

BACKGROUND: In a cancer mortality study, the decision of whether to define a study outcome via underlying cause of death (UCD) or via multiple cause of death (MCD) information may impact relative risk (RR) estimates and associated confidence intervals. METHODS: Simple equations are presented that relate RR estimates obtained in a cancer incidence study to the RR estimates obtained in mortality studies using UCD and MCD information. Data from the Surveillance, Epidemiology and End Results program were used to obtain information about the detection and confirmation rates of cancer diagnoses made via UCD. Data from US cause of death data tapes were used to obtain information on the ratio of UCD to MCD listings for cancer outcomes. Numerical examples illustrate the use of these equations. RESULTS: In our examples, the RRs obtained via analyses of MCD were close to those obtained via analyses of UCD (but of greater precision), even when assuming that the confirmation rate of cancer diagnoses made via MCD listing was substantially lower than that of diagnoses made via UCD. CONCLUSIONS: These finding are supportive of the use of MCD information in cancer mortality studies.     

Deaths: final data for 1999.

OBJECTIVES: This report presents final 1999 data on U.S. deaths and death rates according to demographic and medical characteristics. Trends and patterns in general mortality, life expectancy, and infant and maternal mortality are also described. A previous report presented preliminary mortality data for 1999. METHODS: In 1999 a total of 2,391,399 deaths were reported in the United States. This report presents tabulations of information reported on the death certificates completed by funeral directors, attending physicians, medical examiners, and coroners. Original records are filed in the State registration offices. Statistical information is compiled into a national data base through the Vital Statistics Cooperative Program of the National Center for Health Statistics (NCHS), Centers for Disease Control and Prevention. For the first time in a final mortality data report, age-adjusted death rates are based upon the year 2000 population and causes of death are processed in accordance with the Tenth Revision of the International Classification of Diseases (ICD-10). RESULTS: The 1999 age-adjusted death rate for the United States was 881.9 deaths per 100,000 standard population, a 0.7 percent increase from the 1998 rate, and life expectancy at birth remained the same at 76.7 years. For all causes of death, age-specific death rates rose for those 45-54 years, 75-84 years, and 85 years and over and declined for a number of age groups including those 5-14 years, 55-64 years, and 65-74 years. Aortic aneurysm and dissection made its debut in the list of leading causes of death and atherosclerosis exited from the list. Heart disease and cancer continued to be the leading and second leading causes of death. The age-adjusted death rate for firearm injuries decreased for the sixth consecutive year, declining 6.2 percent between 1998 and 1999. The infant mortality rate, 7.1 infant deaths per 1,000 live births, was not statistically different from the rate in 1998. CONCLUSIONS: Generally, mortality continued long-term trends. Life expectancy in 1999 was unchanged from 1998 despite a slight increase in the age-adjusted death rate from the record low achieved in 1998. Although statistically unchanged from 1998, the trend in infant mortality has been of a steady but slowing decline. Some mortality measures for women and persons 85 years and over worsened between 1998 and 1999.     

A Bayesian approach to assess heart disease mortality among persons with diabetes in the presence of missing data

Some states’ death certificate form includes a diabetes yes/no check box that enables policy makers to investigate the change in heart disease mortality rates by diabetes status. Because the check boxes are sometimes unmarked, a method accounting for missing data is needed when estimating heart disease mortality rates by diabetes status. Using North Dakota’s data (1992–2003), we generate the posterior distribution of diabetes status to estimate diabetes status among those with heart disease and an unmarked check box using Monte Carlo methods. Combining this estimate with the number of death certificates with known diabetes status provides a numerator for heart disease mortality rates. Denominators for rates were estimated from the North Dakota Behavioral Risk Factor Surveillance System. Accounting for missing data, age-adjusted heart disease mortality rates (per 1,000) among women with diabetes were 8.6 during 1992–1998 and 6.7 during 1999–2003. Among men with diabetes, rates were 13.0 during 1992–1998 and 10.0 during 1999–2003. The Bayesian approach accounted for the uncertainty due to missing diabetes status as well as the uncertainty in estimating the populations with diabetes. The findings and conclusions in this report are those of the authors and do not necessarily represent the views of CDC.     

Underlying and multiple cause mortality in a cohort of workers exposed to aromatic amines

Background The National Institute for Occupational Safety and Health (NIOSH) has previously conducted studies of bladder cancer incidence and mortality at a synthetic dye plant that manufactured beta-naphthylamine from 1940 through 1979. This report extends the period of mortality follow-up 13 years and analyzes both underlying and nonunderlying causes of death. Methods The vital status of each cohort member, as of December 31, 1992, was determined by using the National Death Index and information from the Internal Revenue Service and the U.S. Postal Service. The NIOSH life table analysis system (LTAS) was used to generate person-years-at-risk and the expected numbers of death for 92 categories of death, using several referent rates (U.S. underlying, Georgia underlying, U.S. multiple cause). Results There were three bladder cancer deaths listed as underlying cause, yielding a standardized mortality ratio (SMR) based on U.S. rates of 2.4 (95% confidence interval (CI) = 0.5, 7.0) and a total of eight bladder cancers listed anywhere on the death certificates (SMR based on multiple cause referent rates = 5.6; 95% CI = 2.4, 11.1). Mortality from esophageal cancer, which had been significantly elevated in the previous study, was no longer significantly elevated (SMR = 2.0; 95% CI = 0.8,4.1). Mortality from all causes was significantly higher than expected (SMR = 1.5; 95% CI = 1.3, 1.6). Conclusions The elevated bladder cancer risk in this cohort was detected by the multiple cause, but not the underlying cause, analysis. Elevated mortality from other causes of death, especially among short-term workers, may be related to regional and lifestyle factors. Am. J. Ind. Med. 34:506–511, 1998. © 1998 Wiley-Liss, Inc.     

Estimating the out-of-hospital mortality rate using patient discharge data

This paper explores the hospital quality measures based on routine administrative data such as patient discharge records. Most of the measures used in the literature are based on in-hospital mortality risks rather than post-discharge events. The in-hospital outcomes are sensitive to the hospital's discharge policy, thus could bias the quality estimates. This study aims at identifying out-of-hospital mortality risks and disentangling discharge and re-hospitalization rates from mortality rates using patient discharge data. It is shown that these objectives can be achieved without post-discharge death records. This is an example of the use of public use administrative data for estimating empirical relations when key dependent variables are not available. Using data on the lengths of hospitalizations and out-of-hospital spells, the mortality rates before and after discharge are estimated for a sample of heart-attack patients hospitalized in California between 1992 and 1998. The results suggest that the quality assessments that ignore the variation of discharge rates among hospitals could be misleading.     

Mortality from ischemic heart disease in Iceland, 1951-1985.

The objective of this study was to evaluate mortality rates from ischemic heart disease among Icelanders during the period of 1951 to 1985. In some developed countries, the number of deaths from ischemic heart disease declined markedly in this time period, and it is interesting to study whether the same has occurred in Iceland. The study was based on information obtained from the Statistical Bureau of Iceland, which keeps records of deaths based on death certificates as well as other population records. Nonparametric tests were used to correlate death rates and calendar years. Rates per 100,000 were calculated and plotted. The results indicated that the mortality rates from ischemic heart disease among Icelanders have not yet peaked.     

Mortality from dementia in Norway, 1969-83.

From 1969 to the end of 1983 in Norway, dementia was coded as the underlying cause of death from 2058 death certificates, and as a contributory cause from 19,459. This is 3.56% of the total number of deaths. It seems that a considerable proportion of dementia cases are noted on death certificates in Norway. Death rates based on dementia as the underlying cause of death have increased with time, but when including contributory causes, rates have declined. The data may be useful in epidemiological studies, eg, to search for aetiological clues for Alzheimer's disease. Due to the inclusion of contributory causes of death in the registers and to the high number of dementia cases noted on death certificates, Norwegian mortality data on dementia are probably of better quality than in most other countries.     

Mortality in a cohort of antimony smelter workers

Animal studies show that antimonymay cause lung cancer and heart and lung disease in rodents. In exposed humans, ECG abnormalities and heart and lung disease have been reported. This mortality study of 1,014 men employed between 1937 and 1971 in a Texas antimony smelter consisted primarily of workers of Spanish ancestry (n = 928, 91.5%). Hispanics are known to smoke at much lower rates than non-Hispanics, and their lung cancer and heart disease mortality is generally low. When ethnic-specific Texas lung cancer death rates were used for comparison, mortality from lung cancer among antimony workers was elevated (SMR) 1.39, 90% CI 1.01-1.88), and we observed a significant positive trend in mortality with increasing duration of employment. When ischemic heart disease death rates from three different Spanish-surnamed populations were used for comparison, the rate ratios for mortality from ischemic heart disease were 0.91 (90% CI 0.84-1.09), 1.22 (90% CI 0.78-1.89), and 1.49 (90% CI 0.84-2.63). Pneumoconiosis/ other lung disease death rates for Spanish-surnamed men were unavailable and so calculation of rate ratios used white males as a comparison population (SMR 1.22; 90% CI 0.80-1.80). These data suggest some increased mortality from lung cancer and perhaps nonmalignant respiratory heart disease in workers exposed to antimony. However, conclusions are limited by possible confounders and the difficulty of identifying appropriate referent groups.     

Applications of extraction mapping in environmental epidemiology

This paper discusses a new method which allows the extraction of a background disease rate from a data set consisting of spatial co-ordinates of morbidity or mortality events. We demonstrate its application with two data sets, one based on cancer registry data and the other on death certificates.     

Death certificate-based occupational mortality surveillance in the United States

Surveillance of cause-specific mortality patterns by occupation and industry through the use of death certificate records is a simple and relatively inexpensive approach to the generation of leads as to potential occupational disease problems. Researchers from the National Institute for Occupational Safety and Health (NIOSH) have been working with the National Center for Health Statistics, other federal agencies, and state health departments on a number of programs to foster the development of standardized, routine coding of occupation and industry entries on death certificates by state health departments. Thirty-one states and the District of Columbia are now doing such coding. These data are being analyzed currently by investigators at NIOSH and at individual state health departments for the purpose of hypothesis generation on occupation-disease relationships. The proportionate mortality ratio method is the predominant method being used, as appropriate denominator data are not generally available. This type of surveillance is particularly useful for the study of occupation and industry groups for which it is difficult to assemble cohorts, such as groups that are predominantly non-union and in small workplaces. Limitations of this surveillance include its inappropriateness for monitoring those occupational diseases which are not often fatal, and the limited scope and accuracy of death certificate information.