• Evaluation
• community mobilisation
• HIV prevention
• targeted interventions
• India
• anthropology
• ethnography
• HIV
• social science

The case that engaging communities at a high risk of HIV to act in reducing their own vulnerability is essential to prevention programmes has gained widespread acceptance as a public health principle. Producing unequivocal evidence of its effectiveness in reducing HIV transmission is problematic however and the means and mechanisms through which such effects may operate remain a matter of debate. The papers in this special issue contribute to our understanding of both these issues, while indicating that it is feasible to mobilise high risk, marginalised communities for HIV prevention on a large scale. Overall, the papers document remarkable successes across diverse regional contexts and local populations in establishing community groups and building collective solidarity among group members. Yet, some findings are less robust and the papers also show how systematic measurement of community mobilisation and the production of reliable evidence concerning its effects on HIV risk remain fraught with difficulties. In part, these arise from a tension between the research design requirements for producing consistent and reliable evidence acceptable to the public health community and the nature of the very activities that may contribute to effective HIV prevention.

The mechanisms through which community mobilisation exerts effects are addressed in a model1 which conceptualises community mobilisation (glossed as ‘identification, collectivization and ownership’) as an outcome of increasing community participation that will, in turn, lead to better programme outcomes (ie, a reduction in HIV transmission). Arguably however, a reverse pathway in which community participation rests on mobilisation activities is equally plausible and, indeed, other community-based approaches in health and development, including evidence from the Sonagachi project upon which community mobilisation in Avahan was modelled, indicate that collectivisation and identification with a common purpose (‘mobilisation’) are prerequisites for community-level activism. Given the dialectic between these two dimensions, attempts to pin them to distinct semantic domains may be fruitless; indeed, Wheeler et al2 treat mobilisation as essentially synonymous with ‘participation’. However, some consistency in their use is necessary given the positing of a theory of change to account for intervention effectiveness that proposes directionality from one (participation) to the other (mobilisation). Discerning direction of causation is, of course, a classic problem in epidemiology but without a clear understanding of such pathways, it is difficult to assess, for example, the grounds for a claim2 that observed increases in service coverage and health-seeking behaviour are attributable to community mobilisation.

Many of the papers report on the development and use of two innovative survey instruments, the Behavioural Tracking Survey (BTS) and the Community Ownership and Preparedness Index (COPI), which provide tools for systematic measurement of the degree and reported effectiveness of community mobilisation. One set of papers uses results from the BTS to evaluate community mobilisation as an activity in its own right, though the instrument was not designed primarily for this purpose. Three papers report increases in self-efficacy and collective efficacy among sex workers as a consequence of community mobilisation.3–5 However, some refinements to an earlier tool upon which BTS is based6 are puzzling. ‘Collective efficacy’, for instance, is now reported as individual ‘confidence’ in collective efficacy, so that ‘collective action’ is essentially an operational indicator of this (perceived) efficacy. In some papers ‘collective agency’ and ‘collective action’ are both reported although these simply index different kinds of group activity and are not qualitatively distinct domains. Given the significant likelihood of reporting (social desirability) bias in such a resource-intensive intervention, heavy reliance on self-report items focusing on individual perceptual states to measure change gives rise to difficulties in data interpretation.

A second set of papers demonstrate COPI's validity in documenting the formation of community-based organisations, but highlight tensions between the need to produce generalisable indicators of effectiveness and the requirement to demonstrate programme success within a limited time-frame. The COPI was developed as much to monitor ‘transition-readiness’ (from Avahan support to the Government of India's National AIDS Control Programme) as to measure community mobilisation; that is, it seeks to quantify organisational preparedness and, as such, is not primarily an evaluation instrument but rather, or also, an advocacy tool that constitutes an intervention in itself.7 The growing programme emphasis over time on organisational preparedness is apparent in findings8 that show increases in organisational capacity (programme management, governance, engagement with the state) at the cost of a decline in actual community mobilisation activities (networking). Monitoring and evaluation inevitably entail compromises between enabling inductive, context-specific modifications to programme design and the needs of measurement, and the valuable account of programme evolution by Wheeler et al describes how early attempts to monitor community-based organisation development simply diverted attention from primary programme activities to fulfilling targets—an issue that has plagued India's health and family welfare programmes for decades.2 The emphasis on organisational capacity is again illustrated in contributions which describe the development of the programme; particularly striking is the proliferation of bureaucratic structures that community mobilisation has entailed.3 ,4 This may be unavoidable when implementing such interventions on a large scale, particularly when one aim is democratic representation, but there are uncomfortable parallels with the heavily bureaucratised structures of the Indian state and associated potential dangers, including entrenchment of vested interests and institutional inertia.

One notable absence is any account from the two northeastern Indian states covered by Avahan where the key target population for intervention was injecting drug users (IDUs), while only one paper discusses another targeted population, men who have sex with men. The other papers all focus on female sex workers and the logic model presented here relates solely to this population.1 Since the community mobilisation component of Avahan was based on the Sonagachi project which focuses on female sex workers, this emphasis is unsurprising. Nonetheless, Avahan as a programme also targeted men who have sex with men and IDUs and the possibility that mobilisation activities may have had less success in these groups because of the qualitatively different character of these diverse ‘communities’ as potential collectivities is fundamental to assessing potential transferability to other populations and settings. The terms conventionally used to describe engagement processes imply that high risk populations already exist as communities, whereas evidence suggests that the formation of collective identity is often essential to mobilisation efforts.4 ,9 Where commonalities between members of targeted populations consist solely in shared high risk behaviours (as in IDUs) rather than in a common occupational identity (as in sex workers), the potential for building collective solidarity using the same model may be more constrained.

Clear evidence for the effectiveness even of conventional targeted intervention components for preventing HIV infections in high risk populations remains surprisingly limited.10 While demonstrating effects through the measurement of standardised, rigorously comparable, quantified ‘hard’ outcomes remains an ultimate evaluation goal, there are especially significant methodological challenges to achieving this for complex interventions including community mobilisation that, by their nature, entail social structural change and require ongoing inductive and context-responsive adaptation to achieve their aims.11 Other forms of evidence, including longitudinal, ethnographic and case study accounts, to document less readily measurable dimensions of intervention design and implementation can also be invaluable for gaining insight into why interventions do or do not work12–14 and for assessing the potential transferability of models across settings and populations. Further scaling up of community mobilisation initiatives needs to build in evaluation from the start15 ,16 and incorporate research methodologies which do not assume the categories ‘community’ and ‘mobilisation’ are self-evident. This would facilitate clearer explication of pathways that link social interventions with epidemiologically observable effects and are socially and culturally as well as biologically plausible.17 ,18