From the *Los Angeles County Department of Health Services, and the †University of California, Los Angeles, School of Public Health, Department of Epidemiology, Los Angeles, CA.
Submitted 6 December 2004; accepted 1 July 2005.
Supported by the Los Angeles County Department of Health Services.
Supplemental material for this article is available with the online version of the journal at www.epidem.com; click on “Article Plus.”
Correspondence: Matthew D. Redelings, 313 N Figueroa 127, Los Angeles, CA 90012. E-mail: email@example.com.
Background: Mortality statistics can be compiled using underlying cause-of-death data or multiple cause-of-death data, which include other contributing causes of death.
Methods: For the leading causes of death in the United States during 2000–2001, we compared underlying and multiple cause-of-death statistics.
Results: For some conditions, little difference was observed between the 2 estimates. For other conditions, up to 10 times more deaths were identified from multiple-cause data than from underlying-cause data. The 10 leading causes of death differed when using the 2 types of data.
Conclusions: Whenever possible, underlying and multiple cause-of-death statistics should both be presented. Analyses that use only the underlying cause of death ignore additional information that is readily available from multiple-cause data, and the more limited data may underestimate the importance of several leading causes of death.
Click on the links below to access all the ArticlePlus for this article.
Please note that ArticlePlus files may launch a viewer application outside of your web browser.
Analysis of mortality data is critical in efforts to assess the burden of disease, to develop policy initiatives, and to design effective public health interventions. Generally, cause-of-death statistics are compiled by examining the underlying cause of death as specified on death certificates by the examiner.1 The underlying cause of death is defined as “the disease or injury that initiated the train of events leading directly to death, or the circumstances of the accident or violence, which produced the fatal injury.”2
However, deaths rarely have only 1 cause, and relying on underlying cause-of-death data does not allow researchers to assess the role of other conditions that contributed to death. In many cases, these additional conditions may have been necessary factors in the chain of events leading to death, such that death would not have occurred in their absence. Yet because they did not initiate the chain of events resulting in death, they are overlooked in underlying cause-of-death analyses. Underlying-cause data are also limited in that different examiners often disagree as to which of the many factors contributing to death was the underlying cause.3 Multiple cause-of-death data, which specify both the underlying cause of death and as many as 20 contributing causes of death, offer an alternative.4
Mortality analyses that inform policy and funding initiatives at state and national levels still rely almost exclusively on underlying-cause statistics.5–11 Thus, policy decisions may be taken based on information that seriously underestimates the importance of some causes of death. However, the extent to which analyses of underlying causes of death and analyses of multiple causes of death differ for the leading causes of death has not been fully addressed.
We compared death statistics compiled using multiple-cause data with statistics compiled from underlying-cause data in the United States during 2000–2001 for the 50 rankable leading causes of death as defined by the National Center for Health Statistics (NCHS).12 We evaluated the degree of difference between underlying- and multiple-cause statistics.
Data were examined for all reported deaths occurring in the United States during the years 2000 and 2001. We used national multiple-cause-of-death data to identify both underlying and contributing causes of death.13 Mortality frequencies were tabulated for each condition comparing any mention of the condition with a listing of the condition as the underlying cause only.
A list of 50 rankable leading causes of death and cause-specific International Classification of Diseases, 10th Revision codes was obtained from NCHS.12,14,15 Rankable leading causes were formulated by sorting several common causes of death into 50 mutually exclusive categories that were thought to be useful from a public health perspective.6 We identified special conditions of interest not among the rankable leading causes of death from an NCHS-compiled list of 113 causes of death.12
A total of 4,827,153 deaths were reported in the United States in 2000–2001. An average of 2.70 causes of death per decedent were listed in multiple cause-of-death data (median = 2.00, lower quartile = 2.00, upper quartile = 4.00). Only 5% of decedents had more than 5 causes of death listed on the death certificate, and 25% had only 1 cause of death listed.
The degree of agreement between underlying and multiple cause-of-death statistics varied by condition. For intentional injuries (homicide, suicide), the underlying cause of death captured 99% of deaths identified in multiple-cause data (Table 1; supplementary table available with the online version of this article). Underlying-cause data also captured more than 90% of deaths reported in multiple-cause data for malignant neoplasms and HIV.
However, underlying-cause data captured only a fraction of deaths identified by multiple-cause data for some leading causes of death: diabetes mellitus (33%); influenza and pneumonia (29%); nephritis, nephrosis, and nephrotic syndrome (19%); and essential hypertension and hypertensive renal disease (8%).
For other leading causes of death, the degree of agreement between underlying-cause and multiple-cause statistics was moderate. Underlying-cause data captured 73% of deaths recorded by multiple-cause data from accidents or unintentional injuries, 61% percent of deaths from heart disease, 60% of deaths from Alzheimer disease, 59% percent of deaths from cerebrovascular disease, and 49% of deaths from chronic lower respiratory disease.
The 10 leading causes of death differed when we examined the 2 types of statistics (Table 2). Although the ranking of the first 4 leading causes of death is the same for both sources, the fifth leading cause according to underlying cause reports (accidents) is ranked tenth according to multiple-cause data. Conversely, the fifth most common cause according to multiple-cause data (essential hypertension and hypertensive renal disease) does not appear on the underlying-cause list. Alzheimer disease is the eighth most common underlying cause, but it is not among the 10 most common multiple causes. The rankings are similar on the 2 lists for the following: diabetes mellitus; influenza and pneumonia; nephritis, nephrosis, and nephrotic syndrome; and septicemia.
Our analyses indicate that the use of underlying cause-of-death data alone substantially underrepresents the mortality burden for several important causes of death. The increase in deaths observed when using multiple-cause data rather than underlying-cause data differed widely by condition. For some conditions, there was little difference between the 2 estimates. For other conditions, however, as many as 10 times more deaths were found from multiple-cause data than from underlying-cause data. The difference between the 2 sets of statistics was extreme enough to affect the ranking of the 10 leading causes of death in the United States.
Mortality analyses that drive policy and funding initiatives at state and national levels still rely almost exclusively on underlying cause-of-death statistics.5–11 Underlying-cause estimates of the mortality burden are only slightly lower than multiple-cause estimates for cancers, intentional injuries, and HIV. However, mortality from diabetes mellitus, influenza and pneumonia, nephritis, nephrotic syndrome, and nephrosis, and essential hypertension and hypertensive renal disease may be severely underrecognized in underlying-cause data. To a lesser degree, mortality from other leading causes of death (including unintentional injuries, heart disease, cerebrovascular disease, and Alzheimer disease) may also be underrecognized in underlying-cause data.
Analyses based on underlying-cause data alone may fail to recognize the full importance of these conditions, resulting in policy choices that may not be the most effective. Diabetes prevention, for example, may not be given the funding that it deserves if policymakers estimate mortality based solely on underlying-cause statistics, which capture approximately one third of diabetes deaths identified from multiple-cause statistics. Similarly, the impact of influenza and pneumonia may be substantially underestimated if researchers tabulate mortality based solely on underlying-cause statistics, which fail to capture approximately 70% of influenza and pneumonia deaths, as tabulated in multiple cause-of-death reports. Even for conditions such as lung cancer, in which underlying-cause statistics capture 94% of multiple-cause-identified deaths, the additional 6% of deaths not recognized by underlying-cause statistics represent more than 10,000 individuals per year.
Although underlying-cause statistics by themselves can fail to describe the full importance of a disease or condition as a cause of death, multiple-cause statistics by themselves can fail to convey the sometimes marginal importance of the condition in generating the deaths it is reported to have caused. Conditions with extremely low case fatality such as anemia may be reported as contributing causes of death in multiple-cause data even if they have played only a negligible role. They may even be reported simply to reflect their prevalence at the time of death. In this analysis, for example, multiple-cause data showed both anemia and prostate cancer as having caused approximately 90,000 deaths from 2000 to 2001. However, prostate cancer was reported as the underlying cause of death in approximately 70% of cases, whereas anemia was only reported as the underlying cause in 10% of cases. Although multiple-cause data reports almost the same number of deaths from prostate cancer as from anemia, additional information from underlying-cause data indicates that prostate cancer is more important as a cause of death.
This report has important limitations. No gold standard exists for cause-of-death statistics. Causes of death may be incorrectly reported on death certificates. It is not possible to discern based on the coding how important a role the condition played in causing death or whether death could have been postponed if the condition had been prevented. Death certificates may not contain information on conditions such as depression that cause substantial morbidity but contribute to death only indirectly. Although there are inherent differences between the 2 types of cause-of-death data, some discrepancy may also be due to the accuracy of reporting, and these effects are very difficult to assess.
We suggest that both multiple and underlying cause-of-death statistics should be presented to provide readers with a more complete picture of the mortality burden from different conditions. Exclusive use of either type of cause-of-death statistic ignores important information that is readily available and is helpful in characterizing cause-specific mortality. However, for conditions such as anemia in which case fatality is low, caution should be exercised in interpreting additional cases found in multiple-cause data.
1. Chamblee RF, Evans MC. New dimensions in cause of death statistics. Am J Public Health. 1982;72:1265–1270.
3. Smith Sehdev AE, Hutchins GM. Problems with proper completion and accuracy of the cause-of-death statement. Arch Intern Med. 2001;161:277–284.
4. Israel RA, Rosenberg HM, Curtin LR. Reviews and commentary. Analytical potential for multiple cause-of-death data. Am J Epidemiol. 1986;124:161–179.
5. Arias E, Anderson RN, Kung H, Murphy SL, Kochanek K. Deaths: final data for 2001. Natl Vital Stat Rep. 2003;52:1–115.
6. Anderson RN, Smith BL. Deaths: leading causes for 2001. Natl Vital Stat Rep. 2001;52:1–85.
7. Trends in Lung Cancer Morbidity and Mortality. American Lung Association, Epidemiology and Statistical Unit; 2003.
8. Centers for Disease Control and Prevention. Annual smoking-attributable mortality, years of potential life lost, and economic cost—United States, 1995–2000. MMWR Morb Mortal Wkly Rep. 2002;51:300–303.
9. Cancer Statistics 2004. A Presentation From the American Cancer Society. American Cancer Society, Inc; 2004.
10. Centers for Disease Control and Prevention. National Diabetes Mellitus Fact Sheet: General Information and National Estimates on Diabetes Mellitus in the United States, 2002. Atlanta: US Department of Health and Human Services, Centers for Disease Control and Prevention; 2003.
11. Mokdad AH, Marks JS, Stroup DF, Gerberding JL. Actual causes of death in the United States, 2001. JAMA. 2004;291:1238–1245.
12. ICD-10 Cause-of-Death Lists for Tabulating Mortality Statistics (updated October 2002 to include ICD codes for terrorism deaths for data year 2001 and WHO updates to ICD-10 for data year 2003). Hyattsville, MD: National Center for Health Statistics; 2002.
13. National Center for Health Statistics (1997–2004). Data File Documentations, Multiple Cause-of-Death, 2000–2001 (machine readable data file and documentation, CD-ROM Series 20). Hyattsville, MD: National Center for Health Statistics.
14. Goldacre MJ, Duncan ME, Cook-Mozaffari P, Griffith M. Trends in mortality rates comparing underlying cause and multiple cause coding in an English population 1979–1998. J Publ Health Med. 2003;25:249–253.
15. International Classification of Diseases, 10th Revision. Geneva: World Health Organization; 1992.
© 2006 Lippincott Williams & Wilkins, Inc.