**4. Discussion**

With this analysis, we offer a novel application of positive-outlier methods for identifying potential exemplars in reducing zero-dose children since 2000. The DRC, Ethiopia, and India showed among the largest absolute declines in both national no-DTP prevalence and subnational gaps between 2000 and 2019, while Bangladesh and Burundi demonstrated the largest percentage decreases in national no-DTP prevalence and subnational gaps during that time. Given the range of starting points, local contexts, and health system structures in these five countries, it is quite possible that the strategies used, and corresponding lessons learned in improving childhood vaccination—specifically around expanding service reach to unvaccinated children—may be applicable (or at least adaptable) to other settings. As highlighted by the so-called "neighborhood" analysis, comparing divergent no-DTP trajectories among peer locations could support deeper study around what catalyzed faster progress for some places—and how those lessons could be applied elsewhere. The combination of this positive outlier methodology and cross-country platforms supported by the Gavi Learning Hubs offers unique opportunities to better understand 'what works' for accelerating progress in reaching unvaccinated children worldwide.

Considering positive outliers—or potential exemplars—in reducing no-DTP prevalence for both absolute and relative progress can better reflect the range of successful strategies implemented from a range of different no-DTP prevalence starting points. After all, the types of programmatic and policy decisions that may occur when more than 30–50% of under-one children have had no doses of DTP could differ from those occurring when high zero-dose communities are more clustered and national levels of no-DTP are well below 10%. In health service delivery, these differences may unfold around more widespread intervention introduction and scale-up activities (e.g., addressing key infrastructure and personnel gaps that would otherwise impede adoption; mass mobilization and campaign-style outreach efforts) versus more tailored service provision to individuals or communities who still lack access to or demand for an intervention (e.g., hard-to-reach and hard-to-vaccinate populations [46]). For instance, in 2000, the DRC and Ethiopia started among the highest national levels of no-DTP observed across included countries in this study, as well as moderate-to-high levels of subnational gaps. Better understanding how the DRC and Ethiopia substantially reduced no-DTP metrics by 2019 could strengthen strategies adapted for countries that started at similar no-DTP measures in 2000 but had minimal or less pronounced reductions (e.g., Nigeria, Chad, Somalia).

Despite their marked progress since 2000, further improvements in vaccination reach and uptake are needed in the DRC, Ethiopia, India, and other countries still experiencing large populations of unvaccinated children. Accordingly, it is possible that lessons learned from countries with exceptional relative reductions from 2000 to 2019 could be applicable to countries such as the DRC, Ethiopia, and India today; after all, 2019 no-DTP estimates for the latter countries are quite similar to the 2000 estimates for countries such as Bangladesh and Burundi. For Bangladesh, reductions in national no-DTP and subnational gaps nearly paralleled each other time, charting a path toward nearly 0% no-DTP nationally and negligible subnational differences by 2019. These trends may reflect the country's concerted efforts to better reach rural communities with lower levels of vaccination [47], among other immunization and primary care strengthening interventions. For Burundi, levels of and subnational gaps in no-DTP markedly declined after the end of its civil war in 2005 [45]. From 2006 to 2010, Burundi adopted nationwide performance-based financing initiatives focused on improving child and maternal care [48], actions that have been associated with higher vaccination rates, particularly among the poor [49]. To better understand how different interventions and strategies may optimally align with current needs and barriers to vaccination, it is crucial to more deeply examine the programs and contexts in which past gains have occurred.

There is ample opportunity—and need—to characterize what drives successful vaccine delivery and uptake across the spectrum of past and current challenges, particularly around vaccination inequalities. One key consideration that emerged from this analysis involves the pathways by which no-DTP changed both nationally and sub-nationally from 2000 to 2019. Particularly among countries that started with higher levels of no-DTP (e.g., Nigeria and Mali; Ethiopia, and the DRC), subnational gaps often remained unchanged or increased while national no-DTP prevalence began improving. Such pathways suggest that explicit equity program targets and implementation practices may not occur until later. Other countries, including India and Bangladesh, had more consistent declines for both metrics from 2000 to 2019—a potential signal into the ways in which countries are concurrently addressing both national vaccination priorities and at least geographic inequalities. Nonetheless, it is also possible that countries such as India and Bangladesh experienced similar pathways of minimal changes in or rising subnational inequality amid decreasing national no-DTP prior to 2000. Developing a more formalized characterization or framework around 'pathways of progress' toward greater vaccination equity should be considered in future studies, both by geography and across other crucial factors (e.g., gender, wealth, education, religion, ethnicity).

While assessing progress metrics is a necessary first step to better identify potential exemplars in reducing zero-dose burdens, they alone cannot shed light on what countries have executed and how such actions were associated with further improvements. Formally applying methods such as that of the EGH program, with qualitative examination of policy and programs alongside quantitative analyses around drivers of progress [32], should be prioritized for countries and/or subnational locations with notable advances in reducing zero-dose children. Furthermore, the learning and evaluation platform offered through the Gavi Learning Hubs [43], wherein characteristics of immunization programs and factors contributing to their impact will be examined in prospective manner with country researchers and leadership, will enable greater cross-country or subnational engagement around what works to reach unvaccinated children across contexts. This is particularly important for larger countries where subnational locations started at similar starting points but experienced different trajectories over time. For instance, in Nigeria, bordering states Kaduna and Plateau had fairly high levels of no-DTP prevalence in 2000 (69.4% and 50.6%, respectively; Supplementary Figure S2A). By 2019, Plateau decreased its no-DTP prevalence to 14.5%, whereas Kaduna reduced no-DTP prevalence to 31.9%. Another pair of bordering states—Kogi and Enugu—had no-DTP prevalence of 48.1% and 40.7% in 2000; by 2019, Engu recorded a much larger decline by 2019 (to 7.5%) whereas Kogi still exceeded 20% no-DTP prevalence. In Ethiopia, three regions—Afar, Somali, Benshangual-Gomez—had the country's highest no-DTP prevalence in 2000, at 75% or higher, followed by Oromia (69.3%) (Supplementary Figure S2B). By 2019, Benshangual-Gomez and Oromia reduced regional levels of no-DTP to 8.3% and 11.1%. Although Afar and Somali also recorded

substantive declines in overall no-DTP prevalence, each region still had no-DTP prevalence exceeding 25%—and experienced widening gaps in no-DTP among zones. Given these trends and patterns, it is likely that many countries—especially larger ones—could benefit from so-called neighborhood analyses and positive outlier research at the subnational level.

Past work has sought to synthesize and/or assess particular characteristics of successful immunization programs; nonetheless, few studies have expressly focused on both zero-dose children and incorporating mixed-methodologies with a positive outlier lens. For example, qualitative research in Senegal, Zambia, and Nepal points to factors including strong community engagement, integrated delivery, adaptive service provision, and robust data systems as central to improving and/or maintaining high levels of DTP1 and/or DTP3 [35,50–52]. However, the degree to which these approaches are fully transferable to communities with high zero-dose burdens remains unclear. Integrated service delivery, particularly for key primary care interventions for mothers and infants, may have an important role in addressing zero-dose burdens given the high overlap of missing vaccine doses with other essential health services [11]. Strengthening community engagement may require taking a longer-term lens and multifaceted investments, especially in areas of prolonged conflict and/or distrust of health systems and providers. Innovative programs such as the DRC's Mashako Plan, which was launched in 2018 and has sought to improve vaccination completion rates among select provinces through a mixture of supervision support, supply chain improvements, and monitoring efforts [53], may also provide valuable implementation lessons for countries with equally large and/or dispersed populations.

It is worth noting that declines in no-DTP prevalence—and thus increased coverage of DTP1, a marker of program reach—do not inherently equate to gains in broader program retention or complete immunization. For instance, in much of Ethiopia, DTP3 coverage has not improved in parallel amid sizeable increases in DTP1 [7,54,55]. This means that while more children are being reached by vaccination services—an unequivocally crucial milestone—an increasing percentage of them remain under-vaccinated and thus may still be vulnerable to preventable disease. Although some parts of immunization programs can support both vaccination initiation and completion well (e.g., sufficient availability of qualified health workers, strong supply, and cold chain systems), other factors can differentially affect how or whether children finish vaccination series after receiving their first doses [7,56,57]: the availability of defaulter tracking systems, provider-client relationships and trust, flexibility in scheduling for multiple vaccine doses and/or other health services, among others. As global immunization agendas such as IA2030 [36] and Gavi 5.0 [37] rightly bring more attention to zero-dose populations and programs strategies to reach them, it is crucial that political and funding commitments around addressing gaps in under-vaccination also are maintained.
