*5.2.3 Qualitative analysis of value for money*

In Rusizi, the results were provided to all stakeholders involved in the training with Ministry of Health and 25 EHOs through Focus Groups Discussions at District Level. The EHOs were asked to identify and discuss reasons for the variation

between CHCs and to provide contextual rational for some of the anomalies, or where targets were under or over-reached. These insider observations from the grass roots provides the explanation for various challenges and shortcomings, as well as reasons for success of the CHC Model allowing some recommendations to achieve better Value for Money in future CHC programs based on the CHC Theory of Change [30].

In Mberengwa, an in-depth observation was taken on a small sub-set of six CHCs using an interpretivist approach. This was triangulated with participant observation, key informant interviews and focus group discussions involving Environmental Health staff, local leaders, CHC members and others. Field work was done over 2 weeks in Ward 19, which had 39 CHCs in the 43 villages and a population of 9245, in 1481 households. In addition, two villages without CHCs were sampled to serve as control groups to enable comparison [10].

#### **5.3 Limitations and possible sources of bias**

We use project monitoring data which, we accept could be open to interviewer bias as the field officers who managed the programme also assisted the facilitators in the collection of the village data. However, an effort has been made to minimize this bias, by using an external researcher in each country to clean data, excluding all incomplete data and verifying all records and findings in Rwanda [32, 45] and Zimbabwe [44] through spot observations. It is also not ideal that all that co-authors of this paper have been associated either with the design of the CHC approach and the implementation of the intervention in both Zimbabwe and Rwanda and may not be strictly impartial. However, in the interests of our genuine concern to improve learning in the sector, we have attempted to provide only such programming evidence which has been verified by external observers conducting research for their own theses which have subsequently been properly peer reviewed.
