Characterizing Retention in HAART as a Recurrent Event Process

Insights Into 'Cascade Churn'

Bohdan Nosyk; Lillian Lourenço; Jeong Eun Min; Dimitry Shopin; Viviane D. Lima; Julio S.G. Montaner


AIDS. 2015;29(13):1681-1689. 

In This Article

Abstract and Introduction


Objective: The benefits of HAART rely on continuous lifelong treatment retention. We used linked population-level health administrative data to characterize durations of HAART retention and nonretention.

Design: This is a retrospective cohort study.

Methods: We considered individuals initiating HAART in British Columbia (1996–2012). An HAART episode was considered discontinued if individuals had a gap of at least 30 days between days in which medication was prescribed. We considered durations of HAART retention and nonretention separately, and used Cox proportional hazards frailty models to identify demographic and treatment-related factors associated with durations of HAART retention and nonretention.

Results: Six thousand one hundred fifty-two individuals were included in the analysis; 81.2% were male, 40.6% were people who inject drugs, and 42.8% initiated treatment with CD4+ cell count less than 200 cells/μl. Overall, 29% were continuously retained on HAART through the end of follow-up. HAART episodes were a median 6.8 months (25th, 75th percentile: 2.3, 19.5), whereas off-HAART episodes lasted a median 1.9 months (1.2, 4.5). In Cox proportional hazards frailty models, durations of HAART retention improved over time. Successive treatment episodes tended to decrease in duration among those with multiple attempts, whereas off-HAART episodes remained relatively stable. Younger age, earlier stages of disease progression, and injection drug use were all associated with shorter durations of HAART retention and longer off-HAART durations.

Conclusion: Metrics to monitor HAART retention, dropout, and reentry should be prioritized for HIV surveillance. Clinical strategies and public health policies are urgently needed to improve HAART retention, particularly among those at earlier stages of disease progression, the young, and people who inject drugs.


Modern HAART has transformed the HIV epidemic in two profound ways:[1] First, at an individual level, persons living with HIV (PLHIV) today are living longer, healthier lives.[2,3] Second, at a population level, it has been established that the expansion of HAART coverage can lead to a population-level reduction in HIV incidence.[4–7] The extent of these benefits rely on the continuous lifelong use of HAART, as treatment interruptions have been associated with inferior patient outcomes, increased risk of HIV transmission, and emergence of antiretroviral drug resistance.[8–10] Structured treatment interruptions were once applied in clinical practice but have been discontinued as of 2005 following evidence-based recommendations discouraging this practice.[11] However, unstructured interruptions and poor or intermittent retention in HAART remain common.[12–15] It is therefore of vital interest to those engaged in the care of PLHIV to monitor lifelong retention on HAART.[16]

The cascade of HIV care is a well established monitoring tool that measures retention in care and highlights attrition along the HIV care continuum from HIV infection to diagnosis, linkage to HIV care (often defined as the first HIV-related physician visit or diagnostic test), retention in HIV care (regular HIV-related care), to HAART retention and the ultimate goal of virologic suppression.[17,18] Attrition or 'leakage' at different stages along the cascade is heterogeneous, composed of individuals never accessing a given level of care and those disengaged from regular care. The latter itself is a heterogeneous group comprised of those dropping out of treatment for the first time, those with multiple treatment discontinuations, as well as those lost to HIV care entirely (i.e. discontinued regular physician follow-up and diagnostic testing). The sometimes cyclical process of engagement, disengagement, and reengagement in HIV care has been referred to as cascade 'churn', and has been cited as a crucial component of HIV monitoring.[19–22] HAART persistence or sustained engagement (as opposed to adherence, which measures the number of prescribed doses of HAART that were ingested)[23,24] is of particular interest as the ultimate goal of sustained viral suppression cannot be achieved without it.

Although determinants of HAART retention have been studied exhaustively,[25,26] less is known about the process of HAART persistence, dropout, and reengagement over time. Episodes of HAART persistence and disengagement vary in length and the determinants of these episode durations may have critical implications for clinical practice and public health policy. Dynamic characterization, as afforded by advances in longitudinal data analysis,[27–29] can help to shed light on the longitudinal process (i.e. pattern of HAART persistence and disengagement) in the interest of informing clinical practice. This form of analysis may also provide important information to capture within mathematical models used to estimate transmission dynamics and make resource allocation decisions.

We aimed to identify and characterize possible determinants of successive durations of HAART retention and nonretention over time. We applied Cox proportional hazards frailty models to link health administrative data captured for the population of PLHIV in British Columbia (BC) between 1996 and 2012 to address our objectives.