Beyond accountability: learning from large-scale evaluations

An article in the Lancet published online, October 11, 2011, which provides a useful overview of the evaluation of the Avahan HIV prevention programme which was rolled out across six states in India.

The authors explain the methodology behind the evaluation:

“The investigators took a fairly simple approach, focusing on two variables: the amount of money disbursed to the districts that are part of Avahan, as an indicator of intervention intensity; and HIV prevalence among pregnant women attending antenatal care, as an indicator of population trends in HIV infection. Applying a range of analytical approaches to assess the association between those two variables in Avahan and non-Avahan districts, Ng and colleagues noted a substantial and significant effect of Avahan in two of six states (and a small significant effect in a third state) and estimated that 100 178 infections were averted at the population level by 2008 (albeit with great uncertainty, 95% CI 25 897—207 713).”

They point to some potential weaknesses in this approach:

“The limited data on the strength of government and other programmes, the reliance on HIV prevalence among antenatal-care attendees to approximate population incidence trends, and the lack of pathway analyses to support the conclusions.”


“Rapid rollout and ethical discussions in India left no room for randomisation of districts, gathering new data in comparison districts, or even a stepped-wedge design in which the districts are covered sequentially. The few baseline trend data for key indicators, such as condom use among sex workers, and lack of good data for governmental and other prevention efforts in the same and other districts further compound the analysis.”

All Lancet articles are free to access online.

Year of publication: 
Health and HIV
Ties Boerma and Isabelle de Zoysa