The HIV care cascade: a systematic review of data sources, methodology and comparability

Abstract

Introduction: The cascade of HIV diagnosis, care and treatment (HIV care cascade) is increasingly used to direct and evaluate interventions to increase population antiretroviral therapy (ART) coverage, a key component of treatment as prevention. The ability to compare cascades over time, sub-population, jurisdiction or country is important. However, differences in data sources and methodology used to construct the HIV care cascade might limit its comparability and ultimately its utility. Our aim was to review systematically the different methods used to estimate and report the HIV care cascade and their comparability.

Methods: A search of published and unpublished literature through March 2015 was conducted. Cascades that reported the continuum of care from diagnosis to virological suppression in a demographically definable population were included. Data sources and methods of measurement or estimation were extracted. We defined the most comparable cascade elements as those that directly measured diagnosis or care from a population-based data set.

Results and discussions: Thirteen reports were included after screening 1631 records. The undiagnosed HIV-infected population was reported in seven cascades, each of which used different data sets and methods and could not be considered to be comparable. All 13 used mandatory HIV diagnosis notification systems to measure the diagnosed population. Population-based data sets, derived from clinical data or mandatory reporting of CD4 cell counts and viral load tests from all individuals, were used in 6 of 12 cascades reporting linkage, 6 of 13 reporting retention, 3 of 11 reporting ART and 6 of 13 cascades reporting virological suppression. Cascades with access to population-based data sets were able to directly measure cascade elements and are therefore comparable over time, place and sub-population. Other data sources and methods are less comparable.

Conclusions: To ensure comparability, countries wishing to accurately measure the cascade should utilize complete population-based data sets from clinical data from elements of a centralized healthcare setting, where available, or mandatory CD4 cell count and viral load test result reporting. Additionally, virological suppression should be presented both as percentage of diagnosed and percentage of estimated total HIV-infected population, until methods to calculate the latter have been standardized.

Keywords: HIV; HIV care cascade; HIV treatment cascade; antiretroviral therapy; cascade; population-based data; treatment as prevention; treatment coverage.

Figure 1

Study selection. ¹Published and unpublished literature were searched using the string HIV AND (cascade OR continuum); *PubMed, Medline (Ovid), CINAHL (Ebscohost); †unpublished literature: a. conference abstracts: CROI 2015, AIDS 2014, CROI 2014, CROI 2013, IAS 2013, HIV Drug Therapy Conference (Glasgow) 2014, b. specific websites (UNAID, WHO, Government Websites of OECD member countries) [14]. Authors were contacted to provide the complete paper where it was not available; ^¶additional records identified from search of reference lists; ^§some studies had more than one reason for exclusion.

Figure 2

(a) Rate (%) of virological suppression in estimated total populations of people living with HIV, including undiagnosed infection. Cascades reporting rate of virological suppression in the estimated population living with HIV, including undiagnosed infection, are shown. No cascades used comparable population-based data in these calculations. (b) Rate (%) of virological suppression in populations of people living with diagnosed HIV. Cascades reporting rate of virological suppression in the population living with diagnosed HIV are shown. The lighter bars indicate cascades that use comparable, population-based data. These results can be considered to be comparable. The darker bars indicate cascades using less comparable data. The results of cascades indicated with the darker bars cannot be considered to be comparable.