In 1995, HIV infection was the eighth leading cause of death for the United States population, accounting for 1.8% of all deaths . In view of this small percentage, some analysts have asked whether politics rather than reason explains why the federal government spends more for research on AIDS than on any other cause of death . In a recent federal budget proposal, the proportion designated for HIV/AIDS was 11.8% (US$1.54 billion) of the US$13.1 billion total for the National Institutes of Health and 27.6% (US$634 million) of the $2.3 billion total for the Centers for Disease Control and Prevention . How can the appropriate percentage for HIV/AIDS research and prevention be rationally determined? HIV infection was the top cause of death among persons aged 25–44 years, accounting for 19.3% of deaths in this age group , but focusing on persons aged 25–44 years has been interpreted by some as distorting the importance of HIV infection , because fewer than 7% of all United States deaths occur in this age group. Simply considering all deaths in the population, however, without regard to age differences, does not adequately reflect the relative importance of HIV infection among causes of death. From a public health perspective, deaths at younger ages justify special attention, compared with deaths among the elderly, because they are more likely to be preventable, result in lost productivity and reduce support for dependent elderly persons and children . This view implies that the negative value of deaths should be quantified as a function of the age at which they occur, premature deaths counting more heavily than deaths in old age. Various formulae for measuring premature mortality have been proposed [6–8]. The Centers for Disease Control and Prevention have customarily assessed premature mortality with a simple definition of ‘years of potential life lost before age 65’ (YPLL) [9,10]. The utility of YPLL has been recognized by the United States General Accounting Office, whose health financing analysts found that premature mortality, measured as YPLL, was the best single indicator for reflecting differences in the health status of states' populations and recommended that it be used to assist the distribution of federal funding for core public health functions . In this study, we have examined the latest available data for the rank of HIV infection among the causes of YPLL nationally and by state and city and the trends in national population–based rates of YPLL over the period 1982–1995.
We used data on deaths of United States residents in 1982–1995 obtained by the National Center for Health Statistics (NCHS) from death certificates filed in all 50 states and the District of Columbia. Data for 1982–1994 were available on magnetic tape  and preliminary national (but not state or city) data for 1995 were available in published and unpublished tables . Causes of death were identified by their codes in the International Classification of Diseases, Ninth Revision (ICD-9) . HIV infection was identified by supplemental codes introduced in 1987 (042, 043, or 044) [14,15]. For the years before 1987, deaths due to HIV infection were approximated as deaths due to pneumocystosis (code 136.3), Kaposi's sarcoma (code 173), or deficiency of cell-mediated immunity (code 279.1). Underlying causes of death were classified into HIV infection and 38 other categories conventionally used by NCHS for ranking them (from the ‘List 72 Selected Causes of Death’) .
The YPLL for an individual was defined as the difference between 65 years and the age at death if the age was < 65 years, or zero if the age was ≥ 65years. For the population, YPLL due to each underlying cause of death was calculated as the sum of the YPLL for all individuals who died of that cause. Because data for 1995 were available only as tabulations for specific age groups, YPLL for that year were approximated by assuming that the age at death was the midpoint of the age range in each age group. The data were not adjusted to control for varying age distributions among the populations compared, because our objective was to measure the actual (unadjusted) burden of HIV infection on these populations. Data for cities were limited by NCHS to 193 cities that had a population of at least 100 000 in the 1990 census. Cities were defined by their political boundaries rather than their greater metropolitan areas. Results for city and state subpopulations are presented only for those with 25 or more deaths under the age of 65 years from all causes in 1994. Analysis by race/ethnicity excluded data from Oklahoma, because death certificates from Oklahoma did not include information on Hispanic ethnicity . Denominators for calculating YPLL rates per thousand population were mid-year intercensal estimates provided by the United States Bureau of the Census on magnetic tape. Census data for 1990 were used to calculate the 1994 rates for cities, because intercensal estimates were unavailable for cities.
Nationally, while the annual rates of YPLL per thousand population under 65 years of age due to most leading causes either decreased or were stable since 1982, the rate of YPLL due to HIV infection steadily increased up to 1994, reaching a peak of 4.77; the rate in 1995 (4.74) did not differ appreciably from that in 1994 (Fig. 1). On the basis of preliminary data for 1995, HIV infection was the fourth leading cause of YPLL, accounting for 8.8% (1 086 446) of the total YPLL (12 361 966) from all causes (53.93 per thousand population under 65 years of age), after unintentional injuries (accidents and adverse effects) (15.8%), cancer (15.7%) and heart disease (12.0%). Among males, HIV infection was also the fourth leading cause of YPLL (responsible for 11.0%) and among females it was the sixth leading cause (4.5%).
In 1994, the most recent year for which data on YPLL were available by race/ethnicity, HIV infection was the fourth (9.2%) leading cause of YPLL among white (non-Hispanic) males and the second (15.2%) among black (non-Hispanic) males, surpassed only by homicide (17.9%) in the latter group. Among Hispanic males, HIV infection ranked third (13.9%), after unintentional injury (20.5%) and homicide (15.1%) and among males in other racial/ethnic groups (e.g., Asian, Pacific Islander, American Indian), HIV ranked eighth (4.4%). The rank of HIV infection among causes of YPLL in females was tenth (1.7%) among whites, fourth (9.7%) among blacks, fifth (7.1%) among Hispanics and twelfth (1.2%) among other racial/ethnic groups. Despite similar percentages of YPLL due to HIV infection among blacks and Hispanics, the rates of YPLL due to HIV per thousand population under 65 years of age were two to three times as great among black males (22.4) and black females (7.2) as among their Hispanic counterparts (9.9 and 2.3, respectively). Rates were lower among white males (5.5) and white females (0.5) and lowest among males and females of other racial/ethnic groups (2.1 and 0.3, respectively).
By state of residence, HIV infection was the top cause of YPLL among males in New York, Florida, New Jersey and Maryland in 1994, where HIV caused 22.5, 18.1, 17.6 and 13.9% of YPLL, respectively. Among females, HIV infection was not the leading cause of YPLL in any state. Further stratified by race/ethnicity, HIV was the top cause of YPLL among white males in California (18.8%); black males in nine east-coast states: New York (29.4%), New Jersey (28.0%), Rhode Island (25.9%), Delaware (25.3%), Florida (24.0%), Connecticut (22.4%), Massachusetts (21.3%), Maryland (20.3%) and Georgia (15.4%); Hispanic males in seven mostly east-coast states: New York (32.7%), Massachusetts (24.5%), New Jersey (24.0%), Florida (23.3%), Louisiana (23.2%), Pennsylvania (20.8%) and Connecticut (19.2%); black females in six east-coast states: New York (25.8%), New Jersey (25.0%), Florida (21.9%), Rhode Island (21.2%), Delaware (16.1%) and Maryland (14.4%); and Hispanic females in New York (30.1%), Massachusetts (21.6%) and New Jersey (17.4%).
HIV infection was the leading cause of YPLL among males in 51 cities located mainly in states along the east, west and gulf coasts; the percentage of YPLL caused by HIV in these cities ranged from 12.6% in Worcester, Massachusetts, to 50.9% in San Francisco, California. Among females, HIV infection was the leading cause of YPLL in 11 cities in seven east-coast states, where the percentage of YPLL caused by HIV ranged from 11.6% in Atlanta, Georgia, to 31.4% in Newark, New Jersey.
In 1994, all the states and cities where HIV infection was the leading cause of YPLL belonged to a larger group of states and cities where HIV infection was the leading cause of death among men and women aged 25–44 years. For comparison with prior analyses [16,17], in 1994, among men aged 25–44 years, HIV infection was the leading cause of death in 13 states and 135 cities. Among women aged 25–44 years, HIV was the leading cause of death in two states (New York and New Jersey) and 32 cities.
After our mortality analysis was expanded from persons aged 25–44 years to include all years of potential life lost among persons under 65 years of age, HIV infection remained a leading cause of premature mortality in the United States. HIV infection was the fourth leading cause of YPLL nationally in 1995 and it ranked first among causes of YPLL for both males and females in a sizeable number of cities in 1994. These findings confirm the importance of HIV infection as a cause of mortality, which, some have argued, may have been overstated by the narrow focus on the age range 25–44 years in previous analyses. The rank of HIV infection would be even higher if it were compared with smaller subcategories of other causes (e.g., motor vehicle accidents, ischemic heart disease, lung cancer) instead of the large categories (e.g., unintentional injuries, heart disease, cancer) conventionally used by NCHS for ranking causes of death.
Similar analyses have been done on premature mortality in other countries. For example, in 1990, HIV infection was found to have caused 8% of YPLL among men in West Berlin, Germany and 9% of YPLL among men in Catalonia, Spain [18,19]. Although the reporting of HIV infection on death certificates of persons with AIDS has been fairly complete in the United States (about 90% in one study ), in other nations, especially developing countries, where the reporting of HIV infection and perhaps other important conditions on death certificates may be considerably less complete, a reliable YPLL analysis may be difficult. Despite high rates of death from HIV infection in some developing countries, the still higher rates of death from other causes, particularly among children, may give HIV infection a lower rank among causes of YPLL than it has in developed countries .
Among blacks and Hispanics in the United States, the rank of HIV infection as a cause of premature mortality and the percentage of YPLL due to HIV infection were higher than they were among non-Hispanic whites and other racial/ethnic groups. In addition, the number of states in which HIV infection was the leading cause of YPLL was greater for blacks and Hispanics than for whites and other racial/ethnic groups. This is consistent with the higher rates of AIDS among blacks and Hispanics, particularly in the northeast [22,23]. Interpretation of these findings is hampered by the absence of standard definitions of racial/ethnic categories and the means to ascertain them . More importantly, apparent racial/ethnic differences probably reflect social, economic, behavioral and other factors, rather than race/ethnicity itself. Unfortunately, we could not control our analysis for potential confounding between race/ethnicity and these other variables, because data on socioeconomic and behavioral factors are not recorded on death certificates. Despite the difficulty interpreting racial/ethnic comparisons and the possibility that they may further stigmatize disadvantaged groups, they may help to target prevention efforts to populations at increased risk .
The similarity of the findings from the analysis of YPLL to those from the analysis of mortality among persons aged 25–44 years does not mean that the analysis of YPLL was unnecessary. For officials to decide rationally how to allocate financial resources for prevention of morbidity and mortality in the entire population, they need to know more than the distribution of causes of death among persons aged 25–44 years or any other particular age group. Conversely, it is not enough to look simply at total mortality without regard to age. To increase the cost-effectiveness of a comprehensive plan to prevent illness and death, the fact that preventability generally decreases with age should be taken into account and analyzing YPLL is one way to do so.
Although 65 years is the customary age of retirement from work in the United States, since overall life expectancy in this country in 1995 was 75.8 years , the age below which death is considered premature could reasonably be chosen to be at least 75 years, as was done for an analysis of YPLL in Australia . An age cut-off as high as 85 years has also been used to calculate YPLL . Raising the age cut-off reduces the percentage of YPLL due to causes of death that occur predominantly at younger ages and increases the percentage attributable to causes that occur mainly at older ages. Thus, changing the cut-off from 65 years to 75 or 85 years reduces the percentage of YPLL due to HIV infection in 1995 from 8.8% to 7.3% or 5.9%, respectively. With these older age cut-offs, the ranks of some conditions among causes of YPLL change. Unintentional injury falls from first to third place, below cancer and heart disease, but the rank of HIV infection remains at fourth place.
The method we used to calculate cause-specific YPLL is not ideal for measuring premature mortality. It incorrectly assumes that elimination of any cause of death would allow a person to live to the arbitrarily chosen age cut-off, whereas in fact other conditions could cause the death of the person before that age. For the age cut-off to take into account the effect of other potential causes of death, it should theoretically be the life expectancy based on deaths due to all causes other than the one under consideration . It should, moreover, not be the life expectancy at birth but the life expectancy at the age of death (based on persons who lived beyond that age and died subsequently). Calculation of this life expectancy should also take into account its variation among demographic groups (by sex, race/ethnicity, geographic area) and over time. A single summary measure that includes premature morbidity and disability as well as premature mortality would be even more useful . These more complex approaches are probably impractical for routine use. In contrast, our relatively crude method is simple enough to be used by persons with little training in biostatistics or life table analysis.
Regardless of whether one uses 65 years of age or an alternative cut-off age, the calculation of years of potential life lost provides a comparison of the various causes of death that should be useful to public officials for allocating scarce public health resources. The high rank of HIV infection among causes of premature mortality indicates the priority that HIV infection should be given in such considerations. If only YPLL were considered, the proportion of public health resources that should be devoted to HIV infection would seem to equal the percentage of YPLL due to HIV (using the 65 years of age cut-off, 8.8% for the nation as a whole; 50.9% for males in San Francisco). However, planners and policy makers must also recognize the value of maintaining successful prevention programmes (e.g., immunization of children), which cannot be measured in terms of current levels of morbidity, mortality, or YPLL. Similarly, if YPLL due to HIV infection were reduced by advances in antiretroviral therapy and prevention of opportunistic infections, as recent trends in AIDS deaths suggest may be starting to occur , that would not mean that funding for prevention and treatment of HIV infection should be reduced. Furthermore, the increasing prevalence of HIV infection and the plausible ease of prevention by changing behavior imply that the prevention of HIV infection should perhaps be allotted a greater proportion of resources than the amount estimated from YPLL alone.
1. Rosenberg HM, Ventura SJ, Maurer JD, Heuser RL, Freedman MA: Births and deaths: United States, 1995
. Monthly Vital Statistics Report
2. Brimelow P: Who decides
1997, 27 January 159
3. Weiss R: Medical researchers hope to see funding figures rise
. The Washington Post
1997, 10 February, Section A:17.
4. Fumento M: Decoding the AIDS numbers
. The Washington Times
1995, 14 February, Section A:19.
5. Michaels D, Levine C: Estimates of the number of motherless youth orphaned by AIDS in the United States
6. Morrow RH, Bryant JH: Health policy approaches to measuring and valuing human life: conceptual and ethical issues
. Am J Public Health
7. Centers for Disease Control: Premature mortality in the United States: public health issues in the use of years of potential life lost
8. Gardner JW, Sanborn JS: Years of potential life lost (YPLL) – what does it measure
9. Centers for Disease Control and Prevention: Years of potential life lost before age 65 – United States, 1990 and 1991
10. Desenclos JCA, Hahn RA: Years of potential life lost before age 65, by race, Hispanic origin, and sex – United States, 1986–1988
11. General Accounting Office: A health status indicator for targeting federal aid to states (GAO report no. GAO/HEHS-97-13)
. Gaithersburg: General Accounting Office, 1996.
12. National Center for Health Statistics: Catalog of Electronic Data Products from the National Center for Health Statistics
. Hyattsville: U.S. Department of Health and Human Services; 1995.
13. World Health Organization: Manual of the International Statistical Classification of Diseases, Injuries, and Causes of Death, Based on the Recommendations of the Ninth Revision Conference, 1975
. Geneva: World Health Organization; 1977.
14. Singh GK, Kochanek KD, MacDorman MF: Advance report of final mortality statistics, 1994
. Monthly Vital Statistics Rep
15. Centers for Disease Control and Prevention: Human immunodeficiency virus (HIV) codes: official authorized addendum ICD-9-CM, effective January 1, 1988
16. Selik RM, Chu SY, Buehler JW: HIV infection as leading cause of death among young adults in US cities and states
17. Selik RM, Chu SY: HIV infection as leading cause of death among young adults in US cities and states: 1991 [letter]
18. Baier WK, Haberland J, Schott J, Bergmann KE: Die Bedeutung von AIDS fur die Sterblichkeit in Berlin (West)
19. Casabona Barbara J, Blanch Mur C, Vall Mayans M, Salvador Vilalta X: Premature mortality related to AIDS among men and women in Catalonia
20. Chu SY, Buehler JW, Lieb L, et al.
: Causes of death among persons reported with AIDS
. Am J Public Health
21. Murray CJL, Lopez AD: Mortality by cause for eight regions of the world: Global Burden of Disease Study
22. Centers for Disease Control and Prevention: HIV/AIDS Surveillance Report
23. Hu DJ, Fleming PL, Mays MA, Ward JW: The expanding regional diversity of the acquired immunodeficiency syndrome epidemic in the United States
. Arch Intern Med
24. Hahn RA. The state of federal health statistics on racial and ethnic groups
25. Centers for Disease Control and Prevention: Use of race and ethnicity in public health surveillance: summary of the CDC/ATSDR Workshop
26. Gold J, Li Y, Kaldor JM: Premature mortality in Australia 1983–1992, the first decade of the AIDS epidemic
. Med J Aust
27. Centers for Disease Control and Prevention: Years of potential life lost before ages 65 and 85 – United States, 1989–1990
28. Lai DJ, Tsai SP, Hardy RJ. Impact of HIV/AIDS on life expectancy in the United States
29. Murray JL, Lopez AD, eds. The Global Burden of Disease: a Comprehensive Assessment of Mortality and Disability from Diseases, Injuries, and Risk Factors in 1990 and Projected to 2020
. Cambridge: Harvard University Press; 1996.
30. Centers for Disease Control and Prevention: Update: trends in AIDS incidence, deaths, and prevalence – United States, 1996