Background
Manitoba is a central Canadian province where annual rates of new HIV infections are consistently higher than the national average (7.9 vs. 6.9 per 100,000 population, respectively, in 2018) [
1]. Injection drug use (33.9%), condomless anal sex between men (24.4%), and condomless vaginal (heterosexual) sex (20.9%) are the most commonly identified HIV risk exposures in Manitoba [
2], and new infections in 2018 were disproportionately high, compared to the rest of Canada, among individuals identifying as Indigenous (50% vs. 19.3%) and female (40% vs. 29.3%) [
1,
3]. Additionally, notable heterogeneity in rates of new HIV infection exists across the province by geography, age, and sex [
3]—in 2018, 77.6% of new diagnoses occurred in Winnipeg, the provincial capital and main urban centre, and among newly diagnosed females, 11.6% were ≤ 19 years (compared to 1.6% of males) and 14.0% were ≥ 60 years (compared to 3.1% of males) [
3]. At the end of 2018, Manitoba Health, Seniors and Active Living (MSHAL), estimated that 1572 people were living with HIV in the province (personal communication, J. Paul, April 16, 2020), and the Manitoba HIV Program—the primary provider of HIV care in the province—estimated that approximately 1400 people were in care in the same year [
2].
Over the past decade of HIV research, heavy emphasis has been placed on the HIV care cascade (“the cascade”)—a framework and analytic tool providing insights into the continuum of care services for people living with HIV [
4‐
6]—and its simplified counterparts the 90–90-90 Initiative [
7]. Conventionally, cascades use aggregate data to illustrate the proportion of individuals in a population of people living with HIV who have been diagnosed, linked to HIV care services, retained in care, then initiated and sustained on HIV treatment to, ultimately, reach virologic suppression. Using aggregate data to illustrate the continuum of HIV care for an entire population is useful insofar as it can provide a general picture of points of “leakage” or “bottlenecks” within a health system or care program. However, relying on aggregate data to paint a picture of an entire population risks obscuring the underlying heterogeneity among and between individuals and groups who make up the population. To generate evidence that can help to inform the development and optimization of interventions and programs addressing inequities in HIV care, it is crucial to conduct equity analyses that generate disaggregated cascades to showcase nuances and highlight inequalities across the cascade steps within a population.
In 2015, all 193 Member States of the United Nations agreed upon the 2030 Agenda for Sustainable Development, comprising seventeen Sustainable Development Goals (SDGs), which built upon the Millennium Development Goals (MDGs) introduced fifteen years earlier [
8]. At the core of the SDGs is the notion of
leaving no one behind, which, “represents the unequivocal commitment … to reduce the inequalities and vulnerabilities that leave people behind and undermine the potential of individuals and of humanity as a whole” [
9 , p.
6]. This idea underscores the interconnectedness of the SDGs and principles of health equity [
10]—a noted limitation of the MDGs [
11]. As such, under the auspices of the SDGs [
12], there is a need for research that focuses on identifying (health) inequalities that exist, examining the factors that perpetuate and exacerbate these inequalities, understanding how specific inequalities are related to broader health inequity [
13], and developing strategies to minimize or, ideally, eliminate them. As noted in SDG 17, to adequately assess (in)equities, it is necessary that data are disaggregated by socioeconomic, demographic and other relevant, context-specific characteristics [
11,
12].
Publicly available HIV epidemiological data in Manitoba are limited to reports published by the Public Health Agency of Canada (PHAC) [
1] and MHSAL [
3], which focus solely on surveillance data. As such, local understandings of inequalities in HIV care and clinical outcomes among different groups in the province are rudimentary. In 2013—through the support of a multi-site program of research,
Advancing Primary Healthcare for Persons Living with HIV in Canada (the LHIV Study), funded by the Canadian Institutes of Health Research [
14]—a prospective clinical cohort of people living with HIV and/or receiving HIV care in Manitoba was established as an embedded research project within the Manitoba HIV Program [
15]. The establishment of the LHIV-Manitoba cohort opens up numerous analytic opportunities to better understand HIV epidemiology in Manitoba and, for the first time, provides access to de-identified, individual-level clinical data, allowing for disaggregated analyses to take place.
Here, building upon previous work [
15‐
17], we use equiplots to present disaggregated cascade analyses (by age, sex, geography, ethnicity, immigration status, and HIV exposure category) that visualize inequalities in service uptake and clinical outcomes among LHIV-Manitoba cohort participants who were alive as of 31 December 2017 and had received an HIV diagnosis on or before that date. Exploratory multivariable logistic regression analyses are used to quantify these inequalities, generate hypotheses, and provide guidance for future cascade research in Manitoba. In conjunction with future research to understand
why identified inequalities exist across the cascade [
6] and how these inequalities contribute to health inequities [
13], our examination of the cascade through an equity lens [
18,
19] will provide Manitoba’s provincial care program with evidence needed to develop patient-centred care plans that meet the needs of heterogeneous client subgroups, and to advocate for policy changes addressing inequities in HIV care across the province.
Discussion
In general, our data indicate that Manitobans living with HIV are progressing well along the cascade and are nearly meeting 90–90-90 targets—81.5% of those
alive and diagnosed are on treatment and 91.3% of those
on treatment are virologically suppressed (Fig.
1). However, equity analyses highlight important inequalities in progression along the HIV care cascade among cohort participants within different sociodemographic groups. Disaggregating our cascade data calls attention to clear inequalities in HIV care and outcomes by age, geography, and ethnicity among cohort participants. In general, individuals who are younger and non-white are less likely than their counterparts to reach subsequent cascade steps. Notable heterogeneity also exists along the cascade based on participants’ reported HIV exposure categories, with MSM progressing relatively well participants reporting IDU having relatively poor odds of reaching virologic suppression. These trends are not unprecedented; similar inequalities across the cascade have been noted in a variety of contexts [
28‐
32]. While our multivariable logistic regression analyses highlight specific inequalities of statistical significance across the cascade, the use of equiplots to analyze disaggregated cascade data is an innovative and important method for identifying inequalities that, although not statistically significant, are highly relevant and should be considered during programmatic planning and design to address population-level inequities in HIV-related health outcomes. On the quest to generate evidence that can inform policy and program development aimed at minimizing health inequities, using cascade data, both aggregated and disaggregated, to identify inequalities in health outcomes and service access, delivery, and utilization is necessary, but insufficient. As Seckinelgin [
6] and Zamora and colleagues [
11] have argued, employing additional methodologies, such as qualitative inquiry and community-based participatory research, to inform policy and program design is crucial.
When it was first introduced in 2011 by Gardner and colleagues [
4], the spectrum of engagement of HIV care, which ultimately became the HIV care cascade, was framed as an analytic tool for mapping individual- and population-level progression through the continuum of HIV care services. Specifically, Gardner’s model [
4] provides a framework through which to determine the proportion of individuals in various stages along the continuum, and to “explore the potential impact of interventions to improve engagement in care” (p.795). However, over time, the cascade, and its simplified counterparts, the 90–90-90 Initiative [
7] and the 95–95-95 Fast-Track targets [
33], have been adopted or endorsed by global technical and policy normative bodies (e.g. UNAIDS [
7,
33] and the World Health Organization [
34]), and used to guide and influence international HIV policy development [
6]. Expanding the utility of the cascade framework from an analytic tool to a large-scale decision- and policy-making framework is problematic because, as Seckinelgin [
6] notes, “the model itself does not analyse the broader socio-political and economic conditions that interact with individuals’ experiences of HIV and that inform their decisions to engage with health services” [
6]. In the process of developing health policies that align with the principles of health equity and the SDG commitment to leaving no one behind, it is essential for decision-makers to thoroughly consider how social determinants of health influence and manifest inequalities in health outcomes and access to health services [
10,
35].
As we demonstrated, performing equity analyses using HIV care cascade data, and illustrating inequalities along the cascade using equiplots, concisely draws attention to points along the cascade at which specific groups of individuals are unable to optimally engage in HIV care or reach target health outcomes. Still, these analyses provide insufficient context or explanation for leakages along the cascade. In order to appreciate nuances in observed inequalities across the cascade, and to understand, for example, why people are having a hard time engaging in their HIV care and how to best support equitable access for all, complementary research approaches, namely qualitative inquiry and community-based participatory research, are necessary [
6,
11,
36].
Next, using our analyses in this paper as an example, we demonstrate one way to expand the utility of innovative data visualization techniques, such as the equiplot. In Fig.
4, obvious discrepancies exist in the proportions of cohort participants from different geographic regions in Manitoba falling within each cascade step. Of particular interest to the Manitoba HIV Program may be the relatively low proportions of individuals living in western Manitoba categorized as
in care and the substantial leakage at the
virologically suppressed step among participants living in southern Manitoba. Indeed, previous research has also identified substantial geographic heterogeneity in engagement in HIV care [
31,
37], some of which may be attributable to limited access to services due to physical distance [
38], or other individual-, community-, and structural-level barriers [
32,
37,
39]. Although our disaggregated analyses have provided a useful starting point for understanding geographic heterogeneity in HIV care in Manitoba, further mixed methods explorations will be necessary to delve into understanding the complex circumstances that shape inequities along the cascade before meaningful recommendations can be made to inform local programming or policy. A next step to understand geographic inequalities will require further disaggregating data (e.g. by sex, age, socioeconomic status) to uncover whether specific groups
within geographic regions are further vulnerable to suboptimal engagement. Once a reasonable level of granularity is achieved in identifying “key groups” who may require additional support to engage in HIV care, program adjustments and policy development should then be based upon further-contextualized understandings of barriers to engagement in care through meaningful community involvement in program and policy decisions [
11,
36]. Policy and programmatic decisions aimed at reducing health inequities must incorporate nuanced conceptualizations of how the various factors that influence access to and engagement with necessary and appropriate health services interact and overlap [
39] to create specific conditions that prevent individuals from progressing through the HIV care cascade and other care continua [
6]. If these complexities are not considered, and instead the linear logic inherent to the cascade [
6] is privileged, policies intended to reduce gaps in equity will continue to miss the mark.
Study limitations
This study has a few limitations that must be noted. First, a number of limitations inherent to the design of our clinical cohort have been described in detail elsewhere [
14‐
17]. Opportunities to participate in the cohort are introduced to individuals in the context of their clinic appointments with the Manitoba HIV Program; participation is optional and does not impact the way that HIV care and other services are received. Still, we have to assume that selection bias may be influencing our analyses, and actual engagement in HIV care among the clinic population may be lower than we are able to assess from the cohort. For the same reasons, we cannot presume that our findings are generalizable to the broader population of people living with HIV in Manitoba, although previous work suggests that these data are reasonably representative of larger population in HIV care in Manitoba [
15]. Second, using the LHIV-Manitoba cohort as a starting point for the first step of the cascade means that we cannot ascertain information about the proportion of people living with undiagnosed HIV in Manitoba and thus limits our ability to generate provincial estimates for all 90–90-90 targets. Finally, available data were limited such that we were unable to analyse the Manitoban HIV care cascade by income, level of education, or other socioeconomic status (SES) indicators. This will be an important addition to this work, which we will undertake as we move forward with more detailed analyses of our cohort data.
Acknowledgements
This work was conducted on Treaty 1 and Treaty 2 territories, the original lands of the Anishinaabeg, Cree, Oji-Cree, Dakota, and Dene peoples and the homeland of the Métis Nation. We thank our colleagues the Health Information Research Governance Committee of Nanaandawewigamig, the First Nations Health and Social Secretariat of Manitoba for their support. We also wish to acknowledge our collaborators at Manitoba Health, Seniors and Active Living’s Information Management & Analytics team. The results and conclusions are those of the authors and no official endorsement by MHSAL or other data providers is intended or should be inferred. Thank you, also, to providers and staff at the Manitoba HIV Program and the LHIV Study team, in particular Dr. Claire Kendall. Finally, and most importantly, we gratefully acknowledge the time and contributions of our cohort participants, without whom this work could not exist.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.