Table 4 Coding Process

Expert panel findings inform initial codebook

Expert panel findings regarding the timeline of events during the observation period informed both in-depth interview data collection instruments and the initial codebook. These events included community discussions to identify local health priorities for action, mentorship activities between MCH nurses and TBAs, and the use of evaluative monitoring processes by community leaders, among others. Examples of resulting identified CFIR constructs include: Patient Needs & Resources, Networks & Communications, and Formally-Appointed Leaders.

Inter-rater reliability exercise

Analysts coded a diversity of respondent transcripts independently using CFIR constructs, comparing each other’s coding until 80% reliability was reached.

Combined deductive and inductive coding

Analysts used initial CFIR constructs for coding, and identified emergent themes as they evolved in the coding process. For example, though the CFIR provided codes for implementers’ perception of the intervention and motivation to implement (e.g. Knowledge & Beliefs about the intervention), the framework did not address beneficiaries’ motivation for institutional delivery.

Twice-weekly coding debrief discussions for alignment

As transcripts were coded, the analysts developed narrative summaries of each transcript and codes applied. The analysts then jointly debriefed each transcript, using these as a means of continual alignment in coding and identification of emerging themes. This process was carried through until completion of coding.