Mortality among people who inject drugs: a prospective cohort followed over three decades in Baltimore, MD, USA

Abstract

Background and Aims:

During the past decades, people who inject drugs (PWID) have been impacted by the development of combination antiretroviral therapy (cART) to combat HIV/AIDS, the prescription opioid crisis and increased use of lethal synthetic opioids. We measured how these dynamics have impacted mortality among PWID in an urban US city.

Design:

Prospective cohort study using data from the AIDS Linked to the Intravenous Experience (ALIVE).

Setting:

Baltimore, MD, USA from 1988 to 2018.

Participants:

A total of 5506 adult PWIDs (median age at baseline 37 years).

Measurements:

Mortality was identified by linkage to National Death Index-Plus (NDI-Plus) and categorized into HIV/infectious disease (HIV/ID) deaths, overdose and violence-related (drug-related) deaths and chronic disease deaths. Person-time at risk accrued from baseline and ended at the earliest of death or study period. All-cause and cause-specific mortality were calculated annually. The Fine & Gray method was used to estimate the subdistribution hazards of cause-specific deaths accounting for competing risks.

Findings:

Among 5506 participants with 84 226 person-years of follow-up, 43.9% were deceased by 2018. Among all deaths, 30.5% were HIV/ID deaths, 24.4% drug-related deaths and 33.3% chronic disease deaths. Age-standardized all-cause mortality increased from 23 to 45 per 1000 person-years from 1988 to 1996, declined from 1996 to 2014, then trended upward to 2018. HIV/ID deaths peaked in 1996 coincident with the availability of cART, then continuously declined. Chronic disease deaths increased continuously as the cohort aged. Drug-related deaths declined until 2011, but increased more than fourfold by 2018. HIV/HCV infection and active injecting were independently associated with HIV/ID and drug-related deaths. Female and black participants had a higher risk of dying from HIV/ID deaths and a lower risk of dying from drug-related deaths than male and non-black participants.

Conclusions:

Deaths in Baltimore, MD, USA attributable to HIV/ID appear to have declined following the widespread use of combination antiretroviral therapy. Increases in the rates of drug-related deaths in Baltimore were observed prior to and continue in conjunction with national mortality rates associated with the opiate crisis.

INTRODUCTION

People who inject drugs (PWID) in the United States represent a marginalized population who are often confronted with health disparities, economic hardship and limited access to care [1–6]. These disparities lead to increased mortality related to infection (i.e. HIV and HCV), overdose, accidents and acute and chronic diseases [7–11]. Not surprisingly, excess mortality among PWID compared to the general population has been documented [7–12]. In particular, previous meta-analyses reported mortality rates in PWIDs were 10–14-fold higher than the expected mortality rate in the general population globally [7,13] and five to 16 times higher in the North American population [7,13].

The United States has experienced an unprecedented opioid crisis, with high rates of opioid prescribing, widespread non-medical oral opioid use and subsequent increases in injecting opioids. Penetration of fentanyl into the supply of the opioid drug market further intensified this crisis during the last decade [14,15]. Mortality due to substance use disorders and drug overdose has grown exponentially in the United States during the past three decades [16]. Nationally, drug overdose deaths reached a record high in 2017 [17–19] and, due to that, US life expectancy decreased for 3 consecutive years [20], leading to the declaration of a public health emergency in 2018 [21]. Previous studies suggested that the national opioid crisis was at least in part driven by chronic disease pain management, especially among those in rural and suburban areas [22–24].

The emergence of fentanyl and other synthetic opioids in the illicit drug market extended the impact to some urban communities with long-standing injection drug use (IDU; i.e. Baltimore, MD, USA). At a state level, substance-related intoxication deaths in Maryland have continued to rise (doubling from 1250 to 2406 between 2015 and 2018), driven primarily by opioid-related deaths [25]. Fentanyl-related deaths have tripled from 340 to 1119 between 2015 and 2016, mainly concentrated in the Baltimore area [22–25]. An earlier study found that the pattern of drug use among PWID, especially younger PWID, has changed to non-medical use of prescription drugs and high levels of polysubstance use [26], and these factors may have played a major role in overall mortality among younger PWID. Characterizing the impacts of long-standing endemic IDU and the current opioid crisis on overall and cause-specific mortality among PWID in the urban setting could help to guide public health strategies and interventions.

PWID had a high risk of HIV transmission [27] coupled with inconsistent access to and engagement in HIV care, leading to higher HIV/AIDS related mortality rates compared to other risk groups [7,28]. Following the advent and widespread use of combination antiretroviral therapy (cART), AIDS-related mortality has decreased, and people with HIV (PWH) have life expectancies approaching the general population [29–31]. However, treatment interruption and suboptimal uptake of ART among PWID remained a major challenge [32] and the survival benefits were not experienced equally by PWID in the late-cART eras [30]. In addition to HIV/AIDS-related mortality, other studies observed increased mortality related to non-AIDS-defining conditions during the cART era [28,33–36]. HIV-uninfected PWID have been known to have an increased burden of aging and multi-comorbidities compared to other groups [37–39]. Cardiopulmonary diseases, including chronic obstructive pulmonary disease [40,41], are common among PWID. The high prevalence of viral hepatitis leading to chronic liver disease added further morbidity and mortality burden to this population [42,43]. Understanding the aggregative effect of HIV infection and chronic diseases on mortality among aging PWID is critical to guide long-term management of this population [42].

The mortality burden among urban PWID in the United States is therefore complex and impacted by multiple competing factors. National data on mortality among PWID have been largely unavailable due to systematic difficulty in identifying and obtain information on this target population. Most studies have been centered within drug treatment settings and often lacked the sample size or length of follow-up to reveal temporal trends of mortality. Longitudinal cohort data that were uniformly and systematically ascertained on mortality trend and granular data on risk factors among PWID is scarce in the literature [13]. The objective of this study was to estimate the temporal trends of overall and cause-specific mortality and to describe the associated determinative factors in a cohort of PWID with or without HIV over three decades.

METHODS

Study participants and data collection

The ALIVE (AIDS Linked to the Intravenous Experience) study is a long-standing community-based prospective study which has followed PWID in Baltimore, MD since 1988. The study was approved by the Johns Hopkins University Institutional Review Board, and all participants provided written informed consent. Overall study design, enrollment and data collection have been previously reported [44] and are described in the Supporting information. Briefly, participants (aged ≥ 18 years) with a history of IDU were recruited in 1988–89, 1994–95, 2005–08 and 2016–18. Semi-annual visits included standardized interviews, clinical examinations, laboratory testing and biorepository storage. Retention rates vary somewhat, but averaged 85–90% annual retention of surviving participants.

Outcome: cause-specific deaths and diagnostic categories

A detailed description on cause-specific deaths definition is presented in the Supporting information. All participants were annually linked to the National Death Index-Plus (NDI-Plus), with vital status and causes of death from 1988 to 2018. Three physicians (J.S., D.A.P. and E.M.B.) independently reviewed all International Classification of Diseases (ICD) codes identified as primary causes of deaths in NDI-Plus and classified them into four categories (Supporting information, Tables S1 and S2): HIV and infectious disease-related deaths (HIV/ID), overdose/drug/violence related (drug-related) deaths, chronic disease deaths and other miscellaneous cause of deaths.

Risk factors of interest

Socio-demographic factors (age, sex, race and educational level) were collected at baseline; risk factors including drug abuse screening test (DAST) [45] and self-reported IDU behavior were collected at baseline and at follow-up visit. Non-medical use of drugs was defined using the question: ‘Have you used drugs other than those required for medical reasons?’ from the DAST survey. At each visit, time-varying HIV infection was defined by detection of HIV-1 antibodies by the enzyme-linked immunosorbent assay and confirmed by Western blot. CD4⁺ T cell counts were measured using flow cytometry. Hepatitis C virus (HCV) serostatus was assessed at baseline and during follow-up; status was updated to seropositive based on follow-up test results.

Statistical analysis

Baseline characteristics were stratified by individuals who died or survived during study follow-up; differences were evaluated using the χ² test for categorical variables and the Wilcoxon rank-sum tests for continuous variables. We conducted time-trend analysis for all-cause and cause-specific mortality rates. Person-time at risk for death accrued from the participants’ baseline visit, and ended at the earliest of date of death or end of study follow-up (31 December 2018). We assumed that no losses-to-follow-up (LTFU) occurred in the mortality rate calculations due to complete ascertainment of data from NDI-Plus. Age-standardized mortality rates were calculated based on the age distribution in the Maryland state census report by < 50 and ≥ 50 years [46]. All mortality rates were illustrated in curves over time using the locally weighted scatterplot smoothing (LOWESS) method.

Competing risk analysis estimated the subdistribution hazards of cause-specific deaths in each calendar period using the Fine & Gray method [47]. We stratified the overall study period into three eras (1988–97, 1998–2005, 2006–18), based on the observed mortality trends in the calendar periods corresponding to temporal changes in availability and uptake of cART regimens among PWID [48,49]. We constructed figures to illustrate the cumulative incidence function (CIF) of each cause-specific death and competing risk of other deaths by HIV serostatus (Supporting information, Figure S1). Within each competing risk model, we identified one of the four groups of cause-specific deaths as the primary outcome and all other causes of death as competing risk events. Person-time at risk within each calendar period in the competing risk analysis started at the later of participants’ baseline visits or the beginning of each calendar period, and ended with the earlier of date of death or last date of the calendar period. Participants who did not participate in an ALIVE study visit during a calendar period did not contribute person-years during that period. Covariates in adjusted models were selected based on a combination of literature review, evaluation of data availability and significance in the univariate analysis (Supporting information, Table S3). Time-varying covariates were carried forward from the last appearance during the calendar periods to the time of death or end of the period. All analyses were conducted using SAS version 9.4 (Cary, NC, USA) and Stata version 15 (Statacorp, College Station, TX, USA). This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines. The analysis was not pre-registered, and the results should be considered exploratory.

RESULTS

Study population

Among 5506 participants (aged ≥ 18 years) with 84 226 person-years of follow-up [median = 13 years, interquartile range (IQR = 5–27], 2417 (43.9%) were deceased by the end of 2018. The median age at baseline was 37 years and the majority were male (74%), of African American descent (81%), were HCV seropositive (81%) and, in general, had low socio-economic status (Table 1). Participants who died were more likely to be male, of African American descent, less educated, have lower income and had HIV and/or HCV infection.

TABLE 1.

Baseline characteristics of ALIVE study participants, overall and stratified by vital status on 31 December 2018

	Overall (N = 5506)	Alive (n = 3089)	Dead (n = 2417)	P-valuea
Age, median (IQR)	36.6 (30.8–43.8)	36.5 (30.2–45.0)	36.7 (31.6–42.5)	0.87
< 30, n (%)	1203 (21.9)	746 (24.2)	457 (18.9)	< 0.01
30–49, n (%)	3672 (66.7)	1907 (61.7)	1765 (73.0)
≥ 50, n (%)	631(11.5)	436 (14.1)	195 (8.1)
Male gender, n (%)	4056 (73.7)	2208 (71.5)	1848 (76.5)	< 0.01
Race				< 0.01
Black, n (%)	4409 (81.1)	2301 (74.5)	2108 (87.3)
White, n (%)	963 (17.5)	691 (22.4)	272 (11.3)
Other, n (%)	130 (2.4)	96 (3.1)	34 (1.4)
High school education, n (%)	2483 (45.2)	1451 (47.1)	1032 (42.9)	< 0.01
Annual income ≥ $5000, n (%)	1627 (30.1)	1028 (33.7)	599 (25.5)	< 0.01
Active injecting, n (%)	5366 (98.1)	2995 (97.7)	2371 (98.5)	0.04
Ever homeless, n (%)	2019 (39.0)	1109 (39.7)	910 (38.2)	0.27
Anti-HCV seropositive	3412 (80.6)	1854 (74.0)	1558 (90.2)	< 0.01
HIV-positive	1367 (24.8)	524 (17.0)	843 (34.9)	< 0.01

^aBy χ² test (for categorical variables) and Wilcoxon’s rank-sum test (for continuous variables).

ALIVE = AIDS Linked to the Intravenous Experience study; HCV = hepatitis C virus; IQR = interquartile range.

Proportionate mortality

Overall, the number of deaths and proportionate mortality was 738 (30.5%) for HIV/ID, 589 (24.4%) drug-related and 806 (33.3%) chronic disease-related deaths. Among PWH (25% in the ALIVE study), the percentage of deaths due to HIV/ID was 57.1%, 15.7% drug-related causes and 18% chronic disease-related. Detailed breakdown of cause-specific deaths (Supporting information, Table S1) showed that poisonings represented the most common drug-related death (66.9%), while malignancy (33.8%), heart disease (31.1%) and liver disease (16.5%) represented the most frequent chronic disease deaths. Over time, the proportion of HIV/ID deaths increased during the first few years and subsequently declined slowly corresponding to increases in chronic disease deaths; chronic disease deaths overtook HIV/ID as the most frequent cause of death in 2005 (Supporting information, Figure S2). The proportion of drug-related deaths was stable to decreasing before starting to increase again in 2014. Similar trends were observed among HIV-infected participants, with the notable and expected relative increase in the HIV/ID category (Supporting information, Figure S3).

Temporal trends in cause-specific mortality and key causal factors

The age-standardized all-cause mortality rate (Figure 1) first increased dramatically from 1988 to 1996 (23 to 45 per 1000 person-years), then decreased to 20 per 1000 person-years in 2014 before trending upward again (38 per 1000 person-years in 2018). The median age in the ALIVE study increased from 34 to 57 years, but the age-standardized mortality rate suggested that aging alone did not fully account for the upward trend of mortality. Figure 2 shows the temporal trend of the three major cause-specific death categories in relation to the primary causal variable for each. Crude HIV/ID mortality (Figure 2a) dramatically increased from 1989 to 1996 (2.5 to 18 per 1000 person-years), then dramatically decreased afterwards (to six per 1000 person-years in 2018). The observed mortality trend was inversely related to the median CD4⁺ count among PWH in the ALIVE study. The increase in median CD4⁺ count reflected sequential immune recovery after the availability and uptake of cART (Supporting information, Figure S4). Examining temporal trends of HIV/ID mortality versus HIV mortality alone showed similar trajectories (Supporting information, Figure S5). In Figure 2b, chronic disease mortality increased continuously as the cohort aged. In Figure 2c, after an initial increase in drug-related deaths, we observed a decrease until 2010 which roughly paralleled the decline in the proportion of active injecting in the cohort. Then, drug-related deaths increased more than fourfold from 3.8 to 17.6 per 1000 person-years from 2011 to 2018. Concurrently, the proportion reporting non-medical use of drugs nearly doubled (22.7 to 40.5%, Figure 2c). Sensitivity analysis that separated poisoning (illicit drug and alcohol) and violence-related deaths demonstrated the same temporal trends (Supporting information, Figure S6). Cause-specific mortality rates among PWH only displayed similar trends to those of the overall cohort (Supporting information, Figure S7).

FIGURE 1.

Age-adjusted all-cause mortality in AIDS Linked to the Intravenous Experience study (ALIVE), 1988–2018. Scatterplots represent the age-adjusted all-cause mortality in each year. The overall curve was constructed using the locally weighted scatterplot smoothing (LOWESS) method

FIGURE 2.

Trends in cause-specific mortality rates and key causal variables in AIDS Linked to the Intravenous Experience study (ALIVE), 1988–2018. All curves were constructed using the locally weighted scatterplot smoothing (LOWESS) method. (a) HIV/infectious disease (HIV/ID) mortality (per 1000 person-years) among all participants and median CD4⁺ cell count. among people with HIV in the cohort; (b) chronic disease mortality (per 1000 person-years) and median age among all survived participants of the cohort; (c) drug-related mortality (per 1000 person-years), proportion of active injecting and proportion of non-medical use of prescription drugs in the cohort

Risk factors for cause-specific mortality by calendar period

As presented in Table 2, the effect of HIV infection on HIV/ID-related deaths was reduced during the recent calendar period [subdistribution hazard ratio (sHR) = 8.18 in 2006–18 versus 16.97 in 1998–2005]. Active injecting, female sex, being black and HCV-seropositive were independently associated with increased risk of HIV/ID-related deaths in 2006–18. Drug-related deaths were primarily driven by active injecting. During the last calendar period, HIV and HCV infection were also associated with increased risk of drug-related deaths. Being black was inversely associated with risk of drug-related deaths throughout three decades and being female was inversely associated with risk of drug-related deaths in the last decade. Chronic disease-related deaths were driven primarily by older age.

TABLE 2.

Determinant factors of cause-specific deaths by calendar period in ALIVE, 1988–2018

Time-period	Factorsa	HIV and infectious deaths		Chronic disease deaths		Drug and violence deaths
		sHR	95% CI	sHR	95% CI	sHR	95% CI
1988–97	Female	1.13	0.85–1.50	0.79	0.52–1.22	0.75	0.47–1.18
	Black	1.23	0.65–2.32	0.52	0.34–0.81	0.48	0.28–0.77
	Ageb	0.96	0.86–1.07	1.28	1.16–1.42	0.92	0.79–1.07
	High school education	1.14	0.89–1.45	1.08	0.79–1.46	0.65	0.45–0.94
	Active injecting	0.95	0.74–1.21	1.11	0.81–1.51	1.46	0.98–2.18
	HIV seropositive	17.0	10.9–26.5	0.53	0.37–0.75	0.87	0.60–1.25
	HCV seropositive	0.80	0.40–1.58	1.52	0.79–2.94	1.59	0.71–3.57
1998–2005	Female	0.89	0.64–1.23	1.10	0.75–1.61	0.73	0.47–1.13
	Black	1.48	0.65–3.39	0.96	0.47–1.96	0.38	0.21–0.68
	Ageb	1.04	0.93–1.16	1.39	1.24–1.56	0.91	0.78–1.06
	High school education	0.98	0.73–1.30	1.08	0.77–1.51	0.96	0.66–1.39
	Active injecting	0.96	0.72–1.27	1.20	0.85–1.70	2.05	1.41–2.96
	HIV seropositive	16.3	10.1–26.4	0.96	0.67–1.38	1.41	0.97–2.04
	HCV seropositive	2.09	0.77–5.64	1.05	0.54–2.05	1.08	0.54–2.17
2006–18	Female	1.82	1.14–2.30	1.09	0.82–1.44	0.56	0.39–0.81
	Black	2.59	1.11–6.08	1.06	0.66–1.70	0.62	0.41–0.94
	Ageb	1.02	0.90–1.16	1.40	1.28–1.54	0.97	0.88–1.07
	High school education	0.91	0.64–1.30	0.82	0.90–1.56	1.28	0.95–1.74
	Active injecting	1.79	1.26–2.54	1.18	0.89–1.58	2.13	1.55–2.93
	HIV seropositive	8.18	5.33–12.6	1.19	0.90–1.56	1.54	1.11–2.14
	HCV seropositive	2.34	1.26–2.54	1.48	0.92–2.36	1.39	1.55–2.93

Bold type indicates statistically significant results (P < 0.05).

^aSex, race (black versus other) and high school education was assessed at baseline. Age, active injecting and HIV serostatus were assessed at every visit. HCV serostatus was assessed at baseline and periodically during follow-up. Total percentage of missingness in time-varying covariates was < 3%.

^bPer 5 years increase in age.

sHR, sub-distribution hazard ratios, sHR were estimated based on multivariables competing-risk regression based on Fine and Gray’s proportional subhazard model. ALIVE = AIDS Linked to the Intravenous Experience study; HCV = hepatitis C virus; sHR = subdistribution hazard ratio; CI = confidence interval.

DISCUSSION

Our study of the ALIVE study of PWID, followed since 1988 in Baltimore, MD, documents the persistently high, albeit evolving mortality burden among urban PWID. By 2018, the age-adjusted death rate in the ALIVE study is more than fivefold compared to the US population (37.8 versus 7.2 [50] per 1000 person-years). Due to inherent difficulties in recruitment and long-term retention of PWID, few studies could address the complex interplay of exposures and multiple disease burden among an urban PWID population. We provided a comprehensive profile of mortality among PWID for three major cause-specific deaths (HIV-related deaths, drug and violence-related deaths and chronic diseases deaths) within our urban cohort during the last 30 years. We revealed the temporal trends of mortality impacted by the HIV epidemic, uptake of cART, population aging, and the recent opioid crisis. Findings for the current study could be generalized to other urban PWID populations in North America.

More than 40% of the ALIVE participants died during follow-up (median = 13 years), primarily from HIV/AIDS, overdose and chronic diseases. Our study showed that the rates and proportion of HIV/infectious disease-related deaths declined over time, although these remain a notable cause of death particularly among HIV-infected PWID. Furthermore, with improved survival among PWH, there has been a persistently rising challenge of chronic disease-related deaths in the cohort. Most importantly, we showed an alarming increase in the rate of drug-related deaths post 2011 linked to the opioid crisis. Notably, the current age-standardized mortality rate is approaching the mortality rate of the cohort at the peak of AIDS epidemic in 1996, prior to cART availability (Figure 1). The mortality rates we observed were comparable to a recent meta-analysis among people with extramedical opioid use [13]. In the meta-analysis [13], the authors synthesized data from 99 cohorts of PWIDs and reported a pooled all-cause crude mortality of 15.9 per 1000 person-years, drug-related mortality of 5.2 per 1000 person-years and AIDS-related deaths of 1.9 per 1000 person-years. Although drug-related overdose deaths at the national level might have plateaued in 2017 [19,51], we have observed a continued rise in the ALIVE study. In 2018, the crude drug-related death rate in the ALIVE study was more than 80 times higher (17.6 versus 0.22 per 1000 person-years) than the national level [50]. Even excluding violence-related deaths in our drug-related death category, the crude estimate still exceeded 70-fold (16 versus 0.22 per 1000 person-years) compared to the national level, despite substantive efforts to reduce overdose deaths in Baltimore city [19,52]. These data call for continued vigilance and support for a comprehensive public health intervention [24] that include providing low threshold access to treatment for opioid use disorder [53].

The current opioid crisis is often depicted as having major impact on white men and women living outside large urban areas, minimizing the history of opioid use among diverse urban populations [54,55]. However, we observed that drug-related mortality has been exacerbated within a population of PWID. This excessive risk is probably driven by non-medical use of prescription opiates, transition to injection and the emergence of highly potent synthetic opioids such as fentanyl in the illicit drug supply [56]. Despite notable declines in the proportion of active injecting (down to < 30% in 2010 but with a concerning rise in 2018), it remains the strongest risk factor for drug-related mortality. Additionally, similar to previous reports from the ALIVE study [26], our observations (Figure 2c) provide further evidence that active injecting does not fully reflect the drug use burden among urban PWID in recent years. Non-injection drug use has penetrated this population and plays a major role in drug-related mortality. Drug-related mortality risk was also increased among males and individuals with HIV or HCV infection; both viral infections are markers of increased injection intensity in the past. Although female and black race were inversely associated with drug-related mortality, this may be due to the residual competing risk of dying from HIV/ID causes. Previously we have documented that active injection increased risk for non-fatal overdose, but those at highest risk have received inadequate harm reduction services (e.g. nalaxone) [57]. Strategies to mitigate the harms (e.g. naloxone distribution, fentanyl test strips) should be targeted to the highest-risk indiviudals, including active PWID in general but particularly males, PWH or with HCV infection.

Although moderated, HIV infection still plays a major role in the mortality burden in the ALIVE study. While we observed a dramatic decline in HIV/infectious disease deaths in the ALIVE study after 1997 owing to the widespread uptake of cART, such a decline demonstrated a 1–3-year lag compared to other developed countries and regions. PWID with HIV have substantially lower life expectancy than men who have sex with men and other transmission groups [30]. The burden of HIV/ID deaths is especially high among female or black participants. A previous study suggested PWH in the ALIVE study had lower ART adherence compared to other cohorts [58]. The treatment continuum and engagement in HIV care remains a major challenge in this population [32,59–67]. Active injecting has detrimental effects at each stage of the HIV care continuum and, as we observed in the current study, it has independent effects on HIV/ID deaths in the modern cART era in addition to the association with drug-related mortality. These data suggest that interventions to improve the treatment continuum and engagement in care among PWID infected with HIV remains critical in ameliorating the mortality burden and decreasing health disparities between PWID and other groups. Targeted intervention programs for HIV care continuum among female or black PWIDs, including community outreach and harm reduction services, might be critical. They also echo the recent report of the National Academy of Medicine to integrate care, programs and strategies to target both opioid disorders and infectious diseases in response to the dual epidemics [68].

Chronic disease deaths have been steadily increasing during the past 30 years as the cohort has aged. Since 2006, mortality attributed to chronic diseases has replaced HIV/ID as the greatest proportionate cause of death. PWID may suffer from accelerated aging [37–39] due to multiple factors, including long-term drug use [69], chronic life stressors and infection with HCV [70,71], HIV [72] or both [73]. We documented a large proportion of chronic disease deaths due to cardiovascular diseases, malignancy and liver diseases, which tracks closely with hospitalization events in the ALIVE study [74]. The shift from HIV related to non-HIV-related chronic comorbidities as primary causes of hospitalization and mortality parallels trends in other populations [75,76]. Chronic inflammation from long-term HIV infection could further intensify multiple aging phenotypes, such as frailty [77] and cardiovascular diseases [78]. Further investigation into the pathways precipitating the heightened chronic disease burden in this population can inform the appropriate long-term management and healthy aging of the PWID population.

Our study may be limited by several factors. First, ALIVE study participants are a community-based prospective cohort, and the traditional limitations for such a study design apply when interpreting our findings. However, historically, our community samples have been representative and comparable to the source population in HIV infection rates and drug use behavior [44,79,80]. Secondly, we relied upon NDI-Plus data for the underlying cause of death definition. Therefore, potential misclassification could occur, especially when NDI switched from using the ICD-9 to the ICD-10 system [81]. Certain causes of death (e.g. liver-related mortality [82]) may be preferentially underestimated with sole reliance on NDI data. However, we have evaluated each of the ICD codes independently and we do not expect that such changes will alter our observed trends. Thirdly, to study the determinant factors impacting cause-specific mortality among all calendar periods, we were limited to data that have been consistently collected since the beginning of the cohort. Certain factors that might be important to cause-specific mortality (e.g. body mass index to chronic disease-related deaths, types of illicit drugs, polysubstance uses and drug treatment to drug-related deaths) from the early decades of the cohort are missing in our analysis. Future studies using granular data (e.g. polysubstance use, drug use intensity, HIV viremia) from the ALIVE study available in more recent years to assess risk of cause-specific death is warranted. Lastly, we may have underestimated the prevalence of HCV seropositive in the ALIVE study because HCV serostatus was determined at baseline and retested periodically among participants during follow-up. However, we previously reported that in this population HCV acquisition generally occurred within 2 years of initiating injection [83,84], and our baseline prevalence estimate was 80%.

Our study also has certain strengths. By linking participants to NDI-Plus with essentially complete ascertainment of death, we exclude differential mortality resulting from LTFU. Moreover, benefiting from the long-standing ALIVE study infrastructure, we were able to report detailed data on more than 5000 PWID, an otherwise hard-to-reach population [82], in a 30-year time-period as they experienced the HIV/AIDS epidemic, population aging and the opioid crisis. Quality data on mortality among PWIDs in a well-defined cohort are still scarce [13,82]. Our study fills a gap in the literature [13] by providing individualized cohort data that were uniformly and systematically ascertained over three decades.

CONCLUSIONS

During the past three decades, mortality due to HIV and other infectious diseases have substantively declined among PWID, with corresponding increases in chronic disease mortality rates with aging of this population. However, overall age-adjusted mortality rates among urban PWID remain notably high and are approaching the peak mortality observed prior to the availability of cART. Drug-related mortality rates have displayed dramatic increases in recent years corresponding to the opioid crisis. These results highlight the urgent need for public health and clinical programs to provide integrated prevention and care (services dually targeting both opioid disorders and infectious diseases; HIV care that emphasizes comorbidity management) to respond to the parallel epidemics facing this vulnerable population.