Despite decades of immunization investment, an estimated 14.3 million children worldwide remain zero-dose (ZD). Kenya’s reported national immunization coverage of about 72% conceals important subnational disparities and limits the ability of routine methods to identify context-specific drivers of ZD and under-immunized (UI) status. The Zero-Dose Learning Agenda aimed to apply human-centered design (HCD) and continuous learning to answer three questions: Who are ZD children, why they are ZD, and how to co-design and pilot tailored interventions in high-burden settings.
The initiative was implemented between November 2023 and December 2025 in two Kenyan sub-counties with distinct contexts: Turkana Central, characterized as rural and nomadic, and Rachuonyo North, characterized as semi-urban. The work used PATH’s Living Labs 4D HCD methodology (Discover, Define, Dream, Design) to structure a cycle of inquiry, co-creation, prototyping, and iterative testing of interventions in selected facilities and communities.
Data collection combined qualitative and quantitative approaches: interviews, observations, focus group discussions (FGDs), facilitated discussions, and quantitative surveys. Analysis included qualitative coding, quantitative survey analysis, and triangulation across data sources. Co-creation activities included ideation and concept validation with local stakeholders; prototyping included development of detailed concept sheets, measurement plans, and implementation plans. Iterative testing permitted adaptation of interventions in response to emergent findings.
Across both sites the team engaged 798 participants and identified six distinct caregiver archetypes per site. The analysis showed different primary drivers by setting: gender norms were a dominant driver in Rachuonyo North, while geographic access and marginalization were dominant in Turkana Central. The project emphasized that ZD children share a label but not a single pathway to exclusion, underscoring the need for interventions tuned to local determinants such as household gender dynamics, health system trust, and physical access constraints.
Three conceptually distinct interventions were co-designed, prototyped, and iteratively tested in selected facilities and communities:
Chanjo Talks Kazini: a community-based engagement intervention that enrolled 188 participants and included a completion and satisfaction pathway for caregivers.
Quality Household Assessment and Sensitization: a community health promoter–led household assessment and sensitization activity to detect and refer ZD and UI children.
Chanjo Mashinani: a low-cost outreach model delivering services closer to communities through multiple small events.
Each intervention included measurement and implementation planning to permit rapid learning and modification during testing.
Reported performance metrics for the interventions included the following outcomes:
Chanjo Talks Kazini (n=188) achieved a 60.1% completion rate and 92.5% participant satisfaction. The intervention was associated with an increase in perceived spousal support for child vaccination from 37.1% to 74.4%.
Quality Household Assessment and Sensitization trained 68 community health promoters, reached 874 households, and identified 15 zero-dose and 39 under-immunized children; all identified children were subsequently vaccinated according to the report.
Chanjo Mashinani conducted 30 low-cost outreach events, reached 266 zero-dose children, and reduced costs by more than 70% compared with conventional outreach approaches.
These outcomes illustrate that tailored, context-informed interventions can increase identification and vaccination of ZD and UI children while improving perceived support and reducing delivery costs in tested settings.
The iterative HCD approach surfaced barriers that baseline instruments had not detected. An example reported was male caregivers’ fear of HIV testing, which influenced vaccination decisions and had been invisible to earlier assessments. The continuous learning cycle allowed teams to diagnose context-specific obstacles and adjust interventions in real time to address those barriers.
Combining human-centered design with continuous learning enabled the program to diagnose context-specific barriers to immunization and to correct course during implementation. The study’s findings support three overarching programmatic implications:
Zero-dose and under-immunized children require tailored strategies because drivers of exclusion vary by location and caregiver type.
Engaging local voices through co-creation is essential not only for problem identification but also for designing practical, acceptable interventions and adapting them during testing.
Resources and support are needed to strengthen weaker health systems so that tailored interventions can be scaled and sustained.
The authors emphasize that a shared label (ZD) does not substitute for granular, context-driven strategies to reach excluded children.
Ethical approval for the work was provided by the Ethics committee of Maseno University. The authors declared no competing interests. Funding was declared from the Gates Foundation. The report states that all data produced in the study are available upon reasonable request to the authors.