Amsterdam 2015
Menu
AIDSImpact.com
Conference Details
Authors
International Committee
Plenary Speakers
Presenting Speakers
Programme
Sessions
Scientific Committee
Acknowledgements
| Abstract #2323 - Strengthening HIV service provision | ||||
|
Session: 44.7: Strengthening HIV service provision (Parallel) on Thursday @ 16.30-18.00 in C202 Chaired by Simon Gregson, Michael Evangeli Authors: Presenting Author: Dr Romain Silhol - Imperial College London, United Kingdom |
||||
| Additional Authors: | ||||
| Aim: The Spectrum model is used by countries, with UNAIDS support, to produce national HIV estimates but is rarely validated against empirical data. Regional-level validations could be especially useful as Spectrum may be used to provide estimates of how local communities are affected by HIV, and for evaluating the potential impact of targeted interventions. | ||||
| Method / Issue: Antenatal surveillance (ANC) data (N˜1200 per round) and population survey data (N˜10,000) for adults aged 15-49 years, from 12 sites in 3 districts in Manicaland, east Zimbabwe, collected in 5 rounds between 1998 and 2011, were input into Spectrum to create a region-specific simulation on a consistent basis with national estimates. Model and empirical estimates were compared for key epidemiological and demographic indicators such as HIV incidence in adults, all-cause age-specific mortality, and orphan prevalence. Alternative scenarios for data availability were examined and sensitivity analyses were conducted for model assumptions considered important for regional estimates. The effects of migration were investigated as it may be important for regional HIV epidemics. Models were fit using the R-spline method. | ||||
| Results / Comments: The Spectrum model generally agreed with observed data, but over-estimated HIV incidence in the late 1990s and the subsequent pace of decline, and under-estimated young adult mortality. Age-specific HIV prevalence in children was very consistent with empirical estimates from Manicaland in 2010. Paternal and maternal orphanhood were lower and slightly higher, respectively, than empirical estimates. An estimated picture of the regional HIV epidemic relying on antenatal surveillance data alone under-estimated the size of the epidemic whilst including more population survey HIV prevalence data progressively improved the model fit. Adding ANC data from earlier in the epidemic did not significantly improve model accuracy, although the effect varied greatly depending on the way the extra information was provided. Altering levels of migration in the model had little effect on observed demographic and epidemic discrepancies. However, whilst the model allows for migration effects on population size and age-structure, in- and out-migrants are assumed to have the same HIV prevalence and risk behaviours as non-migrants. | ||||
| Discussion: The Spectrum model, using local population data, provided reasonable estimates of local-level HIV epidemics and its impact, although some discrepancies were noted. Differences in HIV prevalence between migrants and residents may need to be captured in the model. | ||||
| Go Back | ||||