Multiple lines of evidence demonstrate that with proper engagement in care and adherence to antiretroviral therapy (ART), persons with HIV/AIDS (PWHA) can lead long, healthy lives.1–4 However, HIV remains a leading cause of death (COD) in cities with high HIV burden, with HIV infection, and HIV-related deaths disproportionately affecting black and Latino/Hispanic populations.5–8
Many studies have evaluated HIV care outcomes among PWHA as a measure of the effectiveness of the care delivery and public health system in facilitating the progress of PWHA through stages of care, from diagnosis and linkage through retention and viral suppression (VS). These stages of care are often visualized as a “care continuum” or “cascade” that serves as a surveillance-based snapshot of the care status of the current population of PWHA.9 The continuum is a tool that estimates the number of undiagnosed persons, persons not linking or engaging in HIV care, and persons not achieving VS. Evaluating points in the HIV care continuum where individuals “fall off” can help identify policies and interventions to improve individual and population health.
Although VS is an important end point for the traditional continuum of care among people living with HIV, a similar methodology could be useful to understand deaths, a critical and often overlooked outcome measure among PWHA. To date, there has been little published work using the same markers of care in the traditional continuum of HIV care to evaluate PWHA who have died.
To address this knowledge gap, we propose the “HIV Mortality Reduction Continuum of Care” (HMRCC), an adaptation of the traditional continuum of HIV care. Our primary objective was to identify “drop-offs” in care during the period preceding death. Such “drop-offs” on the path to VS could be targets for novel or improved strategies to prevent HIV-related deaths. Our secondary objective was to describe the population of PWHA who died, including their care patterns per the HMRCC, to inform programs for people living with HIV who may be at risk of preventable, HIV-related death.
Data Sources and Study Population
The New York City Department of Health and Mental Hygiene (NYC DOHMH) conducts HIV surveillance of all newly diagnosed persons and PWHA receiving care in the 5 boroughs of NYC. The HIV Surveillance Registry (“the Registry”) contains named reports of all persons diagnosed with HIV/AIDS in NYC, in addition to electronic reports of all HIV-related laboratory tests ordered by NYC clinical providers regardless of the type of provider/facility or clinical visit (eg, inpatient vs. outpatient). The registry also tracks patients' vital status, including date of death, COD (when available), and residence at death. This information is updated regularly by matches with local and national death registries.
To generate the HMRCC, we selected all PWHA who had a recorded date of death from 2007 through 2013 and who were living in NYC at the time of death. To be included in the predeath care continuum outcome analysis, patients must have been alive for at least 15 months after their initial HIV diagnosis. We selected 15 months so that the HMRCC could be used to evaluate the care status of individuals who would have had the opportunity to use NYC's system of HIV-related care. Demographic characteristics (sex at birth, race/ethnicity, age at death, residence at death, and ZIP code-based poverty level of residence at death) and clinical characteristics (HIV diagnosis date, date of death, HIV transmission risk, CD4 and viral load (VL) test results, and COD) were obtained from the registry.
To assess the care and clinical outcomes in the period before death, we included all CD4 and VL tests that were reported to the Registry for these patients in the 15 months before death. Cause of death was based on the underlying cause as recorded in the decedent's death certificate. “HIV-related deaths” were deaths whose underlying cause was coded with one of 5 ICD codes that represent HIV disease, B20-B24 in ICD-9/-10 nomenclature.10 These codes correspond with opportunistic infections caused by infectious and parasitic diseases (eg, pneumocystis carinii pneumonia and cytomegaloviral infection), malignant neoplasms (eg, Kaposi sarcoma and Burkitt lymphoma), other specific diseases or infections associated with HIV infection, and with unspecified conditions consistent with HIV disease or AIDS. All other deaths were classified as “non-HIV–related” and included underlying causes, such as cardiovascular diseases, non–AIDS–defining cancers, viral hepatitis, diabetes mellitus, accidents, influenza and pneumonia, and other causes.
We evaluated patient status within 2 time periods in the 15-month predeath period: “the intervenable period (IP)” and the “immediate predeath period.” The immediate predeath period was defined as 3 months before the date of death. Primary analyses excluded outcomes from the immediate predeath period to avoid overestimating care engagement and VS secondary to increased care-seeking behavior and access due to deteriorating clinical status and emergent clinical symptoms and events toward the end of life. Exclusion of this immediate predeath period from the 15 months of Registry data analyzed left 12 months before death during which continuum “drop-offs” could have been addressed with clinical or social interventions to prevent HIV-related mortality. This year-long period will be referred to as the “IP.”
During the IP, we used definitions for the HMRCC similar to those used in our traditional continuum among people living with HIV/AIDS, using CD4 and VL tests as proxies of HIV medical care.11 We determined the proportion of patients who: ever linked to care after HIV diagnosis, defined as the proportion of patients with any CD4 or VL test reported at least 8 days after the date of diagnosis; were retained in care, defined as ≥2 CD4/VL tests ≥90 days apart; were estimated to have been prescribed ART, measured as 95% of those retained in care based on data from the Medical Monitoring Project12 in NYC in 2013 (New York City MMP, Unpublished data, 2013); had a most recent VL reported in the IP of ≤1500 copies/mL13; and were virally suppressed, defined as the most recent VL reported in the IP was ≤200 copies/mL. Care continuum outcomes can fluctuate meaningfully depending on the definition of the various stages of care that are applied. To evaluate the impact of an alternate definition of retention in care, we conducted a sensitivity analysis that included 2 alternative retention definitions. The first was similar to our original definition, but with the added requirement that the 2 care visits signaled by laboratory results reported to the Registry having the same ordering care provider/facility. The second alternate definition incorporated this same requirement, and also required one additional laboratory test ordered by that provider/facility at a later date.
Subgroup analyses in this manuscript focus on HIV-related deaths, given that deaths due to HIV are preventable in the era of effective ART and guidelines supporting universal treatment. We evaluated primary care outcomes among persons with an HIV-related underlying COD stratified by sex at birth, race/ethnicity, age at death, and NYC borough of residence at death, according to data obtained from the Registry. In addition, we measured patients' earliest VL and CD4 count during the IP, as well as their VL and CD4 most proximate to death, by HIV-related vs. non-HIV–related COD. Finally, we compared characteristics of persons with an HIV-related COD who were retained in care by VS status (suppressed/unsuppressed) during the IP.
We quantified disparities in HIV-related age-adjusted death rates among specific demographic and other subgroups by area-based poverty level.14 We assigned persons to one of 4 ZIP code-level poverty groups (<10%, 10 to <20%, 20 to <30%, and ≥30% of residents below the Federal poverty level) based on their residence at death, combining HIV-related deaths across all years of the analysis, 2007–2013. We calculated HIV-related age-adjusted death rates per PWHA alive as of mid-year 2010, using 2010, because it was the middle year of the analytic period (age adjusted to the NYC Census 2010 population). Seven-year death rates were then averaged to create annual death rates. Finally, we calculated a “mortality disparity” metric for each category within the subgroup by subtracting the HIV-related age-adjusted death rate for the lowest (<10%) poverty group from the HIV-related age-adjusted death rate for the highest (≥30%) poverty group.
Between 2007 and 2013, 12,010 NYC PWHA died while living in NYC. Among these, 11,187 (93%) died at least 15 months after an initial HIV diagnosis and were eligible for analysis. This population included N = 5183 (46%) persons with a reported HIV-related underlying COD and N = 6004 (54%) with a reported non-HIV–related COD. Although the majority of deaths were due to non-HIV–related causes, a higher proportion of black and Latino/Hispanic decedents had an underlying HIV-related cause (N = 4655, 48%) compared with white decedents (N = 472, 35%, Table 1). In addition, the median age at death among HIV-related deaths was lower than among non-HIV–related deaths [51 vs. 54 years (data not shown)].
Stages of the HIV Mortality Reduction Continuum of Care
The HMRCC for all-cause deaths among NYC PWHA from 2007 to 2013 is shown in Figure 1. Although a large proportion of eligible patients were linked to care (N = 11,007, 98%), retained in care (N = 8992, 80%), and prescribed ART (N = 8497, 76%), only 40% (N = 4518) of decedents achieved VS during the IP. In the sensitivity analysis of alternate definitions of retention in care, the proportion of people considered retained in care dropped to 73% and 44%, respectively, for retention based on 2 and 3 labs ordered by the same provider. Furthermore, applying the more stringent retention in care definition based on 2 labs ordered by the same provider did not lead to higher levels of VS compared with the standard definition (38% vs. 40%, respectively).
Figure 2 presents the HMRCC stratified by the underlying COD. As expected, a higher proportion of persons who died of non-HIV–related COD were virally suppressed compared with persons who died of an HIV-related cause (46% vs. 34%, respectively). Median earliest VL during the IP among persons who died of a non-HIV–related COD was 113 copies/mL (range: 0–750,000) vs. 7760 copies/mL (range: 0–750,000) among those who died of HIV-related COD (Table 1). Median earliest CD4 count during the IP among persons who died of a non-HIV–related COD was 308 cells/μL (range: 0–2349) vs. 143 cells/μL (0–4475) among those who died of an HIV-related COD.
We examined the HMRCC for HIV-related deaths among specific subgroups (Fig. 3). Although most subgroups had similar rates of linkage to care, retention in care, and ART prescription, there were more pronounced differences in VS rates by demographic characteristics. Men had a higher VS rate compared with women (35% vs. 30%); whites (42%) had a higher VS rate compared with blacks (32%) and Latinos/Hispanics (33%). Viral suppression rates increased with increasing age at death: 13% of persons younger than 30 years were suppressed, compared with 15% of those aged 30–39, 22% of those aged 40–49, 40% of those aged 50–59, and 56% of persons aged 60 and older. By major HIV transmission risk categories, persons with a history of injection drug use (IDU) had a higher rate of retention (88%) and VS (37%) compared with men who have sex with men (MSM) (81% and 35%, respectively) and heterosexuals (81% and 28%, respectively). Persons with MSM-IDU transmission risk had the highest rate of retention (92%), but only the third highest rate of VS (29%).
We also saw variability in VS by persons' NYC county (borough) of residence at death, with the highest rate in Manhattan (39%) and the lowest in Queens (30%) (data not shown). There were corresponding differences in other care outcomes by borough, with Queens also having the lowest retention and ART prescription rates. By ZIP code-level poverty, VS decreased as poverty level increased (43% at the lowest poverty level versus 32% at the highest poverty level), but retention was the highest among the higher poverty groups (86% at the highest poverty level vs. 81% at the lowest poverty level, data not shown).
Characteristics of Persons Achieving Viral Suppression Before Death
Persons in our study with HIV-related COD who were retained in care and achieved VS during the IP were more likely than those who did not achieve suppression to be male (69% vs. 63%), white (11% vs. 7%), and reside in a low or medium poverty area at death (28% vs. 22%), and were less likely to have heterosexual transmission risk (16% vs. 21%) (Supplemental Digital Content Table 2, https://links.lww.com/QAI/B67).
Mortality Disparities by Subgroup and Area-Level Poverty
We identified a clear gradient in rising age-adjusted mortality rates with increasing area-based poverty level within categories of sex at birth, race/ethnicity, and HIV transmission risk (Fig. 4). By subgroup, females had higher age-adjusted mortality rates across poverty levels than males; blacks and Latinos/Hispanics had higher mortality rates than whites; and by risk, rates were the highest among IDU as compared to heterosexuals and MSM.
We also calculated disparities in age-adjusted mortality rates by poverty level (highest to lowest) within categories of sex at birth, race/ethnicity, and HIV transmission risk (Fig. 4). Males had a higher mortality disparity than females (5.7 vs. 2.9). By race/ethnicity, Latinos/Hispanics (5.5) had the highest mortality disparity, followed by blacks (4.1) and whites (2.0). By major HIV transmission risk categories, MSM-IDU had the highest mortality disparity (9.2), followed by IDU (6.3), MSM (3.8), heterosexuals (1.9), and perinatal (0.8).
We developed a novel HIV Mortality Reduction Continuum of Care and successfully applied it to persons with HIV who died in NYC during 2007–2013. The NYC HMRCC revealed striking disparities in care outcomes among NYC residents with HIV who died, both overall and for population subgroups. HIV VS rates in decedents with both HIV- and non-HIV–related causes of death were substantially lower than those measured in the NYC continuum of care of people living with HIV: in NYC in 2013, 64% of all persons with HIV were virally suppressed15; per the HMRCC, only 40% of persons with HIV who died of all causes during 2007–2013 had achieved VS. Furthermore, half of all decedents with an HIV-related COD had a last VL >7760 copies/mL. This finding underscores the need—and opportunity—for more effective clinical management and provision of supportive services to persons with unsuppressed HIV VL.
Looking across the continuum, the largest drop-off was seen in the later stages of care among decedents who seemed to have both linked and remained in care. Among NYC persons living with HIV who were ever prescribed ART, 84% have achieved VS; however, only 53% of persons with HIV who died and were ever prescribed ART achieved this stage of care. This suggests that comorbidities and other psychosocial or structural barriers to treatment adherence may be more prevalent in PWHA at risk of death.
Focusing on preventable, HIV-related deaths reveals high levels of linkage and retention among these individuals. Decedents with HIV-related COD were more likely to have been prescribed ART than decedents with non-HIV–related causes of death, and were more likely to have been prescribed ART than persons living with HIV in NYC. Progress from the ART prescription stage on the continuum to VS was suboptimal in HIV-related decedents compared with non-HIV–related decedents and people living with HIV on the traditional continuum: 43% of PWHA who died of an HIV-related COD and were ever prescribed ART were virally suppressed, compared with 63% of decedents who died of a non-HIV–related cause and were ever prescribed ART and 84% of persons living with HIV and ever prescribed ART.15
We found important poverty-related disparities in HIV-related mortality rates among demographic and risk subgroups, with men, Latinos/Hispanics, and IDU experiencing the most dramatic disparities by residential poverty level. These data underscore the role of poverty in influencing variability in death rates due to HIV across and among subgroups. Furthermore, poverty may be a factor underlying suboptimal care patterns and HIV outcomes before death among people with HIV. A number of studies have documented associations between neighborhood poverty and HIV risk, diagnosis, poor care outcomes, and poor treatment outcomes.16,17
Intervention strategies for improving HIV care outcomes and preventing HIV-related deaths could focus on alleviating conditions related to poverty, such as housing instability, food insecurity, and substance use. Our finding that decedents with an HIV-related COD who were residing in Queens in NYC at the time of death had the lowest retention, and VS rates could be partially driven by the higher proportion of foreign-born persons among HIV-related decedents in Queens compared with other NYC boroughs. Foreign-born New Yorkers have been shown to have poorer HIV-related outcomes compared with US-born New Yorkers.18
These drop-offs have several implications. First, standards for clinical follow-up of PWHA with poor health should be rigorous; laboratory monitoring should be done at least quarterly for such patients. Second, patients considered retained under the conventional definition19–20 that we applied in this analysis may be seeking care only intermittently. True engagement in care means consistent follow-up, not occasional or sporadic encounters with care; “doctor shopping,” use of episodic care venues (urgent care and emergency department), and inpatient care would all generate laboratory tests reported to the Registry that would signal retention, but may not actually represent longitudinal care connection. Interventions should seek to improve and promote continuity of HIV care among PWHA to reduce HIV-related mortality. Programs that provide adherence support to individuals to achieve and maintain VS are also critical. DOHMH has several ongoing projects that involve sharing HIV-related information from the Surveillance Registry with clinical providers in NYC to improve care outcomes among PWHA.
This study has limitations. Although we limited the analysis to persons who (to the best of our knowledge) were living in NYC at the time of death, patients could have been accessing care at facilities located outside city limits. Because HIV-related tests conducted at such facilities are not reportable to the Registry, our analysis might have underestimated the extent of care patients received during the IP. Although retention in HIV care was relatively high in our study population, out-migration from NYC related to care-seeking did not likely have a large impact on our analysis. Although it is also theoretically possible that resistance to HIV medications could explain, in part, our finding of low VS rates coupled with high retention rates in this cohort, data among newly HIV-diagnosed individuals in NYC during this period show relatively low proportions of transmitted drug resistance (eg, 11.7% in 2009, 15.2% in 2012).21
In addition, there is potential for misclassification of the underlying COD in our data. In our analysis, some decedents had an undetectable HIV VL at death and during the IP, but were classified as having HIV as their underlying COD. This could represent disease progression despite VS or potential misclassification of the underlying COD or over reporting of HIV as the underlying COD. Similarly, deaths classified as non-HIV–related (eg, cardiovascular disease and certain malignancies), although not being directly attributable to HIV, may still have been a result of HIV infection. However, although engagement in high-quality HIV care should improve CD4 count and VS, it is less clear to what extent HIV care could prevent death from such HIV-associated conditions. Finally, although our analysis sheds light on the populations that are most at risk of death and should be targeted for intensive outreach services, this analysis excluded persons who died within 15 months of their HIV diagnosis. Given the lateness of their diagnosis, this group is likely an even more difficult population to reach. In the cohort of PWHA who died during 2007–2013 in NYC, there were 823 persons (6.9% of all deaths among PWHA) who died within 15 months of their HIV diagnosis. This is an important area for further analysis.
Although retention in HIV care was high among NYC PWHA who died during 2007–2013, VS was low, at nearly half that among persons living with HIV. High retention coupled with low VS suggests the need to develop strategies to improve suppression and address psychosocial and structural barriers to optimal clinical management. The HIV Mortality Reduction Continuum of Care is a novel framework for evaluating predeath care patterns among PWHA and identifying opportunities for intervention. Future interventions should include strategies to investigate HIV-related deaths as sentinel events. Although disease progression may occur even in the setting of viral suppression, improved medications and guidelines supporting universal treatment of HIV regardless of disease stage should be associated with a significant and durable reduction in the probability of AIDS-related death.22 Beyond using data to identify individuals who may be out-of-care, our analysis implies that these data should be used to inspire and support novel strategies to better support the progress of people engaged in care through the latter stages of the care continuum to VS. Like the traditional continuum of care, the observed “drop-offs” seen in the HMRCC represent opportunities to address the barriers to care experienced by people living with HIV to prevent premature death related to their infection.
The authors acknowledge Karen Coeytaux for her assistance with the analysis of mortality disparities, and the staff of the HIV Epidemiology and Field Services Program at the New York City Department of Health and Mental Hygiene for their contributions to the NYC HIV surveillance system, the data source for this analysis.
1. Antiretroviral Therapy Cohort Collaboration. Causes of death in HIV-1-infected patients treated with antiretroviral therapy, 1996–2006: collaborative analysis of 13 HIV cohort studies. Clin Infect Dis. 2010;50:1387–1396.
2. Danel C, Moh R, Gabillard D, et al. A trial of early antiretrovirals and isoniazid preventive therapy in Africa. N Engl J Med. 2015;373:808–822.
3. Lundgren JD, Babiker AG, Gordin F, et al. Initiation of antiretroviral therapy in early asymptomatic HIV infection. N Engl J Med. 2015;373:795–807.
4. Samji H, Cescon A, Hogg RS, et al. Closing the gap: increases in life expectancy among treated HIV-positive individuals in the United States and Canada. PLoS One. 2013;8:e81355.
5. New York City Department of Health and Mental Hygiene. Epiquery: NYC interactive health data system—vital statistics death/mortality data 2000–2013. Available at: http://nyc.gov/health/epiquery
. Accessed February 12, 2016.
6. Siddiqi AE, Hu X, Hall HI. Mortality among blacks or African Americans with HIV infection–United States, 2008–2012. MMWR Morb Mortal Wkly Rep. 2015;64:81–86.
7. Los Angeles County Department of Public Health Office of Health Assessment and Epidemiology. Mortality in Los Angeles County 2012: Leading Causes of Death and Premature Death With Trends for 2003–2012. Los Angeles, CA: The Los Angeles County Department of Public Health; 2015.
8. District of Columbia Department of Health HIV/AIDS, STD, and Tuberculosis Administration. Annual Epidemiology & Surveillance Report, 2013. Washington, DC: 2015.
9. Gardner EM, McLees MP, Steiner JF, et al. The spectrum of engagement in HIV care and its relevance to test-and-treat strategies for prevention of HIV infection. Clin Infect Dis. 2011;52:793–800.
10. Sackoff JE, Hanna DB, Pfeiffer MR, et al. Causes of death among persons with AIDS in the era of highly active antiretroviral therapy: New York City. Ann Intern Med. 2006;145:397–406.
11. Sabharwal CJ, Braunstein SL, Robbins RS, et al. Optimizing the use of surveillance data for monitoring the care status of persons recently diagnosed with HIV in NYC. J Acquir Immune Defic Syndr. 2014;65:571–578.
12. Centers for Disease Control and Prevention. Medical Monitoring Project (MMP). 2015. Available at: https://www.cdc.gov/hiv/statistics/systems/mmp/
. Accessed June 26, 2017.
13. Quinn TC, Wawer MJ, Sewankambo N, et al. Viral load and heterosexual transmission of human immunodeficiency virus type 1. Rakai Project Study Group. N Engl J Med. 2000;342:921–929.
14. Toprani A, Hadler J. Trends in mortality disparities by area-based poverty in New York City, 1990-2010. Paper presented at: 2015 Northeast Epidemiology Conference; October 1, 2015; New Brunswick, NJ.
15. New York City Department of Health and Mental Hygiene. Care and Clinical Status of Persons With HIV in NYC in 2013 as Based on HIV Surveillance Data. 2013. Available at: http://www1.nyc.gov/site/doh/data/data-sets/epi-surveillance-slide-sets.page
. Accessed June 15, 2016.
16. Burke-Miller JK, Weber K, Cohn SE, et al. Neighborhood community characteristics associated with HIV disease outcomes in a cohort of urban women living with HIV. AIDS Care. 2016;28:1–6.
17. Wiewel EW, Bocour A, Kersanske LS, et al. The association between neighborhood poverty and HIV diagnoses among males and females in New York city, 2010–2011. Public Health Rep. 2016;131:290–302.
18. Wiewel EW, Torian LV, Nasrallah HN, et al. HIV diagnosis and utilisation of HIV-related medical care among foreign-born persons in New York City, 2001–2009. Sex Transm Infect. 2013;89:380–382.
19. U.S. Department of Health and Human Services Health Resources and Services Administration. HAB HIV Core Clinical Performance Measures for Adult/Adolescent Clients: Group 1. Rockville, MD: 2008.
20. U.S. Department of Health and Human Services Health Resources and Services Administration. HIV/AIDS Bureau Performance Measures. Rockville, MD: 2013.
21. HIV Epidemiology and Field Services Program. HIV Surveillance Annual Report, 2013. New York, NY: New York City Department of Health and Mental Hygiene; 2014.
22. Walensky RP, Paltiel AD, Losina E, et al. The survival benefits of AIDS treatment in the United States. J Infect Dis. 2006;194:11–19.