Updates to the Spectrum/AIM model for the UNAIDS 2020 HIV estimates

Abstract

Introduction: The Spectrum/AIM model is used by national HIV programs and UNAIDS to prepare annual estimates of key HIV indicators. This article describes key updates to paediatric and adult models for the 2021 round of HIV estimates.

Methods: Potential updates to Spectrum arise due to newly available data, new analyses of existing data, and the need for new issues to be addressed. Updates are guided by experts through the UNAIDS Reference Group on Estimates, Modelling and Projections. Changes are tested and assessed for impact before being accepted into the final model.

Results: Spectrum tracks children living with HIV by CD4% for ages 0-4 and CD4 count for ages 5-14. Data from IeDEA treatment sites have been used to map the transition from CD4% to CD4 count at age 5. Breastfeeding patterns in sub-Saharan Africa have been updated with the latest survey data and estimates of continuation on antiretroviral therapy (ART) with breastfeeding have been revised based on recent studies. Model assumptions about the CD4 counts of people who drop out of ART have been revised to account for CD4 count increases while on treatment. If available, monthly data on numbers on ART can now be used to estimate the effects of COVID-19-related disruptions during 2020.

Conclusions: These changes are intended to provide more accurate estimates of HIV burden. The effects of these changes on paediatric indicators are small except in countries with new surveys that might have updated patterns of breastfeeding. Changes to the adult model have little effect on total new infections. AIDS-related deaths will be somewhat lower in countries that have data on ART drop out but might be increased by HIV care disruptions due to COVID-19. The updated model uses newly available data to improve the estimation of paediatric and adult HIV indicators.

Keywords: CD4; COVID-19; HIV; Spectrum; estimates; mortality.

Figure 1

Structure of Spectrum/AIM model.

Figure 2

Treatment dynamics.

Figure 3

CD4 counts of patient's pre‐ART, on ART and after treatment interruption (TI).

Figure 4

Modelled breastfeeding duration by region and HIV status in 2005, 2010 and 2015. Regional trends shown are based on average country effects. Maximum likelihood point estimates (solid curves) and 95% central credible intervals (shaded areas) are shown.

Figure 5

Numbers of people living with HIV by ART status. Based on reported numbers of people on ART by year and assumed annual retention on ART of 83% and the same rates of mortality and ART initiation for those previously on ART and never on ART. Note that the line for PLHIV not on ART includes those previously treated. ART, antiretroviral therapy; PLHIV, people living with HIV.