Integration of a Digital Health Intervention Into Immunization Clinic Workflows in Kenya: Qualitative, Realist Evaluation of Technology Usability

**Background:** Approximately 25% of children in Africa remain underimmunized against vaccine-preventable diseases. Digital health interventions (DHIs) such as electronic immunization registries (EIRs) have been introduced to improve vaccination coverage in low-resource settings, yet few have been institutionalized at scale. This study aimed to describe how health care workers (HCWs) in Siaya County, Kenya, integrated an EIR into their immunization clinic workflows and to develop a refined program theory on EIR usability.

**Methods:** Informed by realist methodology, the authors developed an initial program theory based on existing health-technology adoption frameworks (Fit Between Individuals, Task, and Technology framework; Smith ergonomics balance theory). A qualitative study was designed using semistructured interviews and direct workflow observations. Twelve facilities were purposively selected from all public facilities in Siaya County performing point-of-care (POC) data entry, stratified by facility size (small, medium, large based on 33rd and 66th quantiles of children aged <1 year in catchment area) and length of EIR experience (<3 months vs ≥3 months). Nineteen HCWs were interviewed (74% with 1-5 years at facility, 53% with ≥3 months EIR experience, 63% nurses). Fifty-eight workflow sessions were observed across 11 facilities (one facility had no children presenting). Data were analyzed using thematic coding in ATLAS.ti, with context-mechanism-outcome (CMO) configurations used to develop rules for implementers.

**Key Results:** Of 58 observed workflows, 42 (72%) involved alternating between EIR and paper tools at POC, 5 (9%) used paper only with later EIR entry, and 11 (19%) used paper only (due to system performance issues). User agreement with EIR usability was generally high (≥75% agreement on most heuristics), but 32% (6/19) disagreed that workflow was good during dual data entry, and 42% (8/19) disagreed that there were enough staff to adequately use the EIR. All 19 users (100%) agreed the EIR was easy to use, would recommend it, and were satisfied overall. Four rules were developed: (1) ensure users complete training to build familiarity, understand system value, and know where to find support (74% agreed they received adequate training; 79% knew where to find user guides); (2) confirm the system captures all needed data and is easy to navigate (84% agreed EIR captured correct information; 89% agreed errors could be prevented); (3) identify work-arounds for poor network, system performance, and too few staff or resources (53% agreed they had enough staff; 74% agreed system functioned when needed; 89% said downtime was minimal); (4) make users aware of expected workflow changes and how these differ over time and by facility size. The refined program theory added 'system-reality alignment' as a key intermediary condition, defined as congruence between system design, user needs, and real-time workflow.

**Clinical Implications:** The EIR had high acceptability but faced significant barriers including software bugs (incorrect vaccine schedules, incorrect defaulter classification), poor system performance after a software upgrade, network connectivity issues, and staffing shortages. Users preferred paperless data entry but were required to maintain dual records. The study demonstrates that DHI success depends on closing the 'design-reality gap'—systems must align with actual workflows, infrastructure realities, and user capacity. The authors recommend using human-centered design during pilot stages and again after scale-up, combined with implementation science methods, to continuously assess and improve system fit. The four rules provide actionable guidance for future DHI implementers in low-resource settings.