As human immunodeficiency virus (HIV)–infected persons are surviving to older ages, the spectrum of causes of death among them is changing—the proportion of deaths in which HIV infection was the underlying cause has decreased, while the proportion of deaths due to causes not clearly attributable to HIV has increased [1-3]. Retention in medical care and effective treatment to achieve a suppressed viral load are essential to reduce morbidity and mortality, as well as the potential for onward transmission of the virus [4]. However, little information is available about care and outcomes before death among persons living with HIV who eventually died of either HIV-attributable or other causes. Assessment of predeath care by cause of death can inform opportunities for intervention. This study is an update on a recent study that examined HIV care within the year before death [5]. We expanded the investigation to compare HIV-infected persons who died of an HIV-attributable underlying cause with those who died of another underlying cause with respect to (1) disease stage within 12 months before death and (2) measures of care in terms of frequency of CD4 T-lymphocyte counts or viral load measurements and viral suppression.

All US states and the District of Columbia require reporting of cases of HIV infection to their health departments; however, not all have mandatory reporting of all values of CD4 T-lymphocyte cell counts and viral load test results by laboratories. We used data reported to the National HIV Surveillance System (NHSS) of the Centers for Disease Control and Prevention through July 2015 from 13 jurisdictions with mandated laboratory reporting of all results from HIV-related tests to their HIV surveillance programs and that also collected cause of death information (California, District of Columbia, Hawaii, Iowa, Louisiana, Maryland, Michigan, Missouri, New Hampshire, New York, South Carolina, Texas, West Virginia). The analysis was restricted to persons who died in 2012, were ≥13 years old at the time of death, and resided in the selected areas at both diagnosis and death. The purpose of the residential restriction was to enhance the completeness of data on laboratory test results, because laboratories report test results to the health departments of the jurisdictions corresponding to the patient’s residential address reported by the health care provider. Among the health departments of these 13 jurisdictions, at least 85% of the deaths they reported to NHSS had data on the underlying cause, which they obtained by linking HIV surveillance data with either state or local vital records data or the National Death Index. The health departments reported the data to NHSS without key personal identifiers (eg, name, Social Security number) that they used to link HIV cases to death records and laboratory test results. Causes of death were identified by codes in the International Classification of Diseases, Tenth Revision (ICD-10) [6]. We classified a death as “HIV-attributable” if the reported underlying cause was HIV infection, indicated either explicitly (by an ICD code for HIV infection itself) or implicitly (by an ICD code for an AIDS-defining opportunistic illness or immunodeficiency—cell-mediated or unspecified type, not an antibody-mediated or congenital type). We assumed HIV infection was underlying an opportunistic illness or immunodeficiency because all the decedents had HIV infection reported to NHSS, even if HIV was not mentioned on the death certificate. If the underlying cause was known but was not HIV infection, we classified the death as “non–HIV-attributable death” (and the underlying cause as “other” than HIV infection; Multimedia Appendix 1). Persons missing information on underlying causes of death were excluded from the analyses. Stage of disease was based on the most recent CD4 cell count or percentage, and was defined in a reversible way, so that a person whose HIV disease had previously met the criteria for stage 3 (AIDS) could be reclassified in stage 1 if the most recent CD4 cell count was ≥500 cells/µL [7]. Persons were considered to be “in care” within 12 months before death if they had ≥1 CD4 or viral load test result in that period and “in continuous care” if they had ≥2 CD4 or viral load test results at least 3 months apart within the last 12 months before death [8]. Viral suppression (defined as <200 copies/mL) was based on the most recent viral load in the 12 months before death.

We assessed indicators overall and by sex, age, race/ethnicity, and transmission category (male-to-male sexual contact, ie, men who had sex with men or MSM; injection drug use; MSM and injection drug use; heterosexual contact; and other). We also determined care and viral suppression by disease severity (whether a person’s infection had ever been classified as stage 3 disease, AIDS), the length of time since diagnosis of HIV infection, and urban versus rural area of residence at the time of diagnosis (metropolitan statistical area population ≥500,000; metropolitan statistical area population 50,000-499,999; and nonmetropolitan population <50,000). Using log-binomial regression, we calculated prevalence ratios and 95% CIs to determine statistical differences on measures of care between persons who died of HIV-attributable underlying causes and those who died of non–HIV-attributable underlying causes. To control for potential confounding covariates, the analyses were adjusted for sex, age at death, race/ethnicity, transmission category, and ever AIDS. Analyses were adjusted for missing risk factor information. Analyses were performed using SAS version 9.3 statistical software (SAS Institute Inc), with the GenMod procedure for the log-binomial regression.

Among 6594 persons who died in 2012 with diagnosed HIV infection, 48.88% (3223/6594) died of HIV-attributable underlying causes and 51.12% (3371/6594) died of non–HIV-attributable underlying causes. Of those who died of HIV-attributable causes, most were male (2374/3223, 73.66%) and ≥40 years old at death (2673/3223, 82.94%); their racial/ethnic distribution was 47.35 % (1526/3223) non-Hispanic black or African American, 23.30% (751/3223) non-Hispanic white, 22.25% (717/3223) Hispanic or Latino, and 7.11% (229/3223) other; 40.80% (1315/3223) were MSM (Table 1). Among those who died of non–HIV-attributable causes, most were male (2493/3371, 73.95%) and ≥40 years old at death (3064/3371, 90.89%); their racial/ethnic distribution was 45.57% (1536/3371) non-Hispanic black or African American, 29.01% (978/3371) non-Hispanic white, 18.30% (617/3371) Hispanic or Latino, and 7.12% (240/3371) other; 36.78% (1240/3371) were MSM (Table 2). Overall, the percentage of persons with late-stage disease (stage 3, AIDS, based on the most recent indicator, ie, CD4 test or opportunistic illness diagnosis) in the 12 months before death was more than twice as great among persons who died of HIV-attributable causes (65.28%, 2104/3223; Table 1) as among persons who died of non–HIV-attributable causes (30.88%, 1041/3371; Table 2). The percentage with stage 3 disease was similarly higher in almost all demographic groups and transmission categories of persons who died of HIV-attributable causes than in their counterparts who died of other causes.

Overall, 91.34% (2944/3223) and 82.74% (2789/3371) of persons who died of HIV-attributable or non–HIV-attributable causes, respectively, had care within 12 months before death (≥1 CD4 or viral load test; Table 3). The percentage of persons who received continuous care (≥2 CD4 or viral load tests, 3 months apart) was similar among persons who died of HIV-attributable and non–HIV-attributable causes (66.15%, 2132/3223 and 65.89%, 2221/3371, respectively; Table 4). The percentage of persons with viral suppression was substantially lower among those who died of HIV-attributable causes (33.76%, 1088/3223) than among those who died of non–HIV-attributable causes (50.96%, 1718/3371; adjusted prevalence ratio 0.69, 95% CI 0.65-0.73; Table 4). Prevalence of viral suppression for persons who ever had stage 3 (AIDS) was lower among those who died of HIV but did not differ by cause of death for persons who never had stage 3 (AIDS).

The percentages with ≥1 CD4 or viral load test was ≥74% for most demographic groups and transmission categories but somewhat lower for persons who never had stage 3 disease (AIDS; 69.0%, 191/277 and 64.7%, 522/807 among persons who died of HIV-attributable and non–HIV-attributable causes, respectively) and persons whose diagnosis of HIV infection was ≤12 months before their death due to non–HIV-attributable causes (67.8%, 124/183).The percentage of persons with viral suppression was lower in most demographic and behavioral groups of persons who died of HIV-attributable causes than in their counterparts of persons who died of other causes.

Most persons in this study had HIV infection diagnosed more than 5 years before death (2368/3223, 73.47% and 2762/3371, 81.93% among those who died of HIV-attributable and non–HIV-attributable causes, respectively; Table 3).