BACKGROUND: There is general agreement amongst major international policy makers that sustainability is a key component of health interventions in developing countries. However, there is little evidence on the factors enabling or constraining sustainability. Diffusion of innovation theory can help explain how the continuation of activities is related to the attributes of innovations. Innovations are characterised by five attributes: (i) relative advantage; (ii) compatibility; (iii) complexity; (iv) triability; and (v) observability. An eye care programme was selected as a case study. The programme was implemented in the Brong Ahafo region of Ghana and had been funded over a ten-year period by an international organisation. METHODS: Sustainability in the study was defined as the level of continuation of activities after the end of international funding. Measuring the continuation of activities involved checking whether each eye care activity continued (i.e. out-patient consultation, cataract surgery, outreach, school health, and statistics) or was interrupted after the end of Swiss Red Cross funding the 11 district hospitals where the programme was implemented. RESULTS: The results showed a relationship between the level of sustainability and the attributes of every activity. The activities with the lowest score for the attributes were less sustained. School health screening was the least sustained activity after the end of international funding. This activity also held the smallest score in terms of attributes: they were the most incompatible and most complex activities, as well as the least triable and observable activities, amongst the four district activities. In contrast, compared to the three other district activities, facility-based consultations were more likely to be routinised because they were perceived by the hospital managers as very compatible, and not complex. CONCLUSIONS: Using diffusion of innovations theories can help predict the sustainability of specific activities within a health programme. The study also highlighted the need for disentangling the various components of a health programme in order to identify which activities are more likely to be continued within a health system. The same methodology could be used in a different setting and could help predict which innovations are more likely to be adopted and maintained over time.