Causes of Death among Persons with AIDS in the Era of Highly Active Antiretroviral Therapy: New York City

Context As HIV treatment becomes more effective, AIDS-related deaths are decreasing and HIV-infected patients are dying of other causes. Better information about these other causes will help to determine appropriate health care for this population. Contribution The authors used death certificates to identify the causes of death in 68669 residents of New York City reported with AIDS. The percentage of deaths from nonHIV-related causes increased from 19.8% to 26.3% between 1999 and 2004. The principal causes of nonHIV-related deaths were cardiovascular disease, substance abuse, and nonAIDS-defining cancer. Cautions Death certificates are an imperfect way to identify cause of death. Implications Health care for HIV-infected patients must include prevention and management of common diseases as well as HIV-focused care. The Editors Over the past 20 years, AIDS has been transformed from a disease that was almost inevitably fatal to a chronic condition that is manageable for many people in the United States (1). The evolution began modestly in the early 1990s with prophylaxis against common opportunistic illnesses and accelerated in the mid-1990s with the introduction of protease inhibitors and highly active antiretroviral therapy (HAART). Between 1996 and 1998, HIV-related morbidity and mortality decreased by 60% in the United States (24). Along with increases in survival, the spectrum of underlying causes of death among persons with AIDS has gradually shifted. Between 1987 and 1999, the proportion of deaths due to nonHIV-related causes increased from 10.6% to 22.9% in 2 U.S. metropolitan areas (5). The most common nonHIV-related causes of death reported in the literature are alcohol and drug dependence, cardiovascular disease, and nonHIV-related cancer (69). The distribution of these causes varies with the sociodemographic characteristics of the persons studied, notably the prevalence of injection drug use (1012). In recognition of the increasing importance of nonHIV-related causes of death, the Infectious Diseases Society of America (IDSA) has argued that health care for people with HIV infection should expand from a primary focus on HIV-related illnesses to include preventable conditions that account for an increasing proportion of deaths (13). Thus, analyses that contribute to a fuller understanding of the underlying causes of death in subpopulations of persons with AIDS are needed. Many previous analyses are limited by small sample size, lack of generalizability, a focus on specific causes of death, and a failure to distinguish between deaths of persons with AIDS and deaths of persons with HIV infection (non-AIDS) (8, 1419). New York City is the single largest HIV/AIDS-reporting jurisdiction in the United States, accounting for 15.3% of AIDS cases and 16.4% of deaths among persons with AIDS (20). Thus, we had a unique opportunity to conduct a population-based analysis of the spectrum of underlying causes of death in a large and heterogeneous population. The data are drawn from 2 population-based registries, the New York City HIV/AIDS Reporting System and Vital Statistics Registry, and cover the period of 1999 through 2004. Methods Population The population was made up of persons 13 years of age or older who received; a diagnosis of AIDS; were alive at any time between 1999 and 2004; were reported to the New York City HIV/AIDS Reporting System as of 30 September 2005; were residents of New York City at the time of diagnosis; and, among those who died, had a known underlying cause of death (98.2% of all deaths). Data Sources The New York City HIV/AIDS Reporting System is a population-based registry of persons who received a diagnosis of AIDS (beginning in 1981), as defined by the Centers for Disease Control and Prevention (CDC), or HIV infection (non-AIDS) (beginning in 2000) (21). The current AIDS case definition includes a positive test result for HIV plus 1 or more of 26 opportunistic illnesses or a CD4+ lymphocyte count less than 0.200109 cells/L or less than 14% of total lymphocytes. The New York City HIV/AIDS Reporting System receives reports of possible AIDS diagnoses through an electronic laboratory reporting system or physician reports and investigates them by chart review. Reporting of AIDS in New York City is estimated to be 95% complete (22). The vital status of persons with AIDS is ascertained by semiannual matches between the HIV/AIDS Reporting System and the Vital Statistics Registry. The underlying cause of death is coded at the New York City Department of Health and Mental Hygiene (DOHMH) Office of Vital Statistics by a nosologist who is certified by the National Center for Health Statistics. The nosologist codes the cause of death using the International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10) (23). We classified persons as living in an area of poverty if they lived in a ZIP codetabulation area with more than 20% of the population below the 1999 federal poverty level or if they were homeless (24). All other variables were patient-level and were collected as part of routine surveillance. We derived demographic data from medical record reviews and provider reports and computed age at the end of 2004 or at the time of death for persons who died. We classified race or ethnicity as Hispanic, black (non-Hispanic), white (non-Hispanic), or other or unknown. The HIV transmission categories were injection drug use, men who have sex with men, and high-risk heterosexual sex. The high-risk heterosexual category included heterosexual sex with a partner who had HIV infection, with an injection drug user, or with a bisexual man. We classified men who were injection drug users and who had sex with men as injection drug users. Otherwise, when more than 1 risk factor was reported, we classified persons on the basis of the CDC hierarchy of transmission categories (25). We defined borough as the borough of residence at the time of AIDS diagnosis. We grouped the year of the AIDS diagnosis into 3 periods: pre-HAART (before 1996), early HAART (19961998), and late HAART (19992004). We obtained CD4+ lymphocyte counts primarily through an electronic laboratory reporting system. The CD4+ lymphocyte count used in the analysis was the lowest count in the second half of 2004 or within 6 months of death. Outcome The outcome was the underlying cause of death. Persons with an unknown underlying cause of death (n= 233 [1.8%]) were excluded from cause-specific analyses. HIV-Related Underlying Causes of Death We classified deaths as HIV-related if the ICD-10 code for the underlying cause of death was between B20 and B24 (HIV disease) or if the ICD-10 code was for an opportunistic illness in the CDC case definition. The latter criterion ensured that we did not misclassify deaths of people with AIDS as nonHIV-related because HIV was not mentioned on the death certificate (26). We did not further categorize these deaths in the main analysis because 70.9% of deaths were assigned a nonspecific underlying cause, for example, HIV disease resulting in other specified conditions (ICD-10 code B23.8) (Appendix Table 1). Appendix Table 1. Categories of Underlying Causes of HIV-Related Deaths in Persons with AIDS in New York City, 19992004* NonHIV-Related Underlying Causes of Death We classified deaths with a known underlying cause that did not meet the criteria described earlier as nonHIV-related. We further classified underlying causes into 9 major categories based on those used by the New York City DOHMH Office of Vital Statistics (27). Appendix Table 2 shows these categories and their associated ICD-10 codes. The substance abuse category included heterogeneous conditions that were associated with alcohol and drug abuse, including drug dependence (that is, overdose), alcoholic liver disease, cirrhosis, hepatitis C, and liver cancer (2732). The cardiovascular disease category comprised all ICD-10 codes between I00 and I78, except cardiac arrest codes. The cancer category comprised malignant types of cancer, except liver cancer and neoplasms that are part of the CDC case definition. We further classified nonHIV-related causes into 16 specific subcategories to better characterize the cause of death. Appendix Table 2. Codes for Major Categories of NonHIV-Related Causes of Death and Selected Specific Causes within Categories* Statistical Analysis We calculated the age-adjusted mortality rates per 10000 persons with AIDS for each year from 1999 to 2004 and for the entire time period. Mortality rates were age-standardized to the U.S. Census population in New York City in 2000 (33). We tested trends in rates of HIV-related deaths, nonHIV-related deaths, and specific nonHIV-related causes by using linear regression models. The model that tested trends in HIV-related and nonHIV-related deaths pooled all deaths to allow for differential trends and an explicit statistical test of whether they differed. We compared crude and age-standardized mortality rates by using methods developed for mortality vital statistics (34). We tested the association between time to death and patient characteristics in separate Cox proportional hazards regression models for HIV-related and nonHIV-related deaths. Independent variables in the model were age, sex, race or ethnicity, HIV transmission category, borough, residence in an area of poverty, year of AIDS diagnosis, and lowest CD4+ lymphocyte count. Date of cohort entry was 1 January 1999 or the date of AIDS diagnosis if diagnosis was after this date. We followed cases until death or we censored cases on 31 December 2004 if patients were still alive on that date. Those who died of a nonHIV-related cause were censored on the date of death in the model that assessed time to HIV-related death. Similarly, those who died of an HIV-related cause were censored at death in the model that assessed time to nonHIV-related death. We verified the proportional hazards assumption by