5 Discussion and Recommendations
5.1 Prioritized Intervention Strategies
The baseline survey covered fifteen priority Local Government Areas (LGAs) in Kano State: the three sentinel LGAs of Gaya, Gabasawa, and Nassarawa, plus twelve non-sentinel LGAs that together form a combined comparison group. This section draws on those findings to lay out where and how to intervene across all fifteen LGAs.
Zero-dose prevalence has three categories of drivers: demand, supply, and access. Ideally, a strategy to reduce zero-dose prevalence would work on all three at once, though the right balance will vary by setting. Some recommendations below — particularly those targeting missed opportunities and dropout — address vaccination uptake and completion more broadly, not zero-dose entry alone.
5.1.1 Demand-Side Priorities
The clearest demand-side gaps show up in caregivers’ own explanations for why their children were not vaccinated (Section 4.3). Social norms dominate — about 41% of zero-dose children had caregivers citing at least one norm-related reason — followed by beliefs (23%), access (14%), and health-facility issues (5%) (Table 4.76). Accordingly, reasons in the intent domain account for nearly six in ten zero-dose cases (Figure 4.59).
That points squarely to social-norms work as the lead intervention. Three approaches are worth prioritizing, some of which already exist in some shape or form: engaging community leaders, deploying trusted community health workers for household visits, and embedding vaccination counselling into routine health-system contacts such as postnatal care and birth registration. The evidence for community-leader engagement is particularly strong in Nigeria. A cluster-randomized trial in Cross River State, for instance, found that training traditional and religious leaders raised the odds of receiving any vaccine roughly twelvefold over control wards, at around US$60 per measles case averted — and a recent Cochrane review gives the broader combination of leader engagement, provider interventions, and outreach moderate-certainty backing (A. Oyo-Ita et al. 2021; A. E. Oyo-Ita et al. 2021).
Community-leader engagement is also a highly actionable LGA-level lever. Across survey areas, only about 41% of caregivers said community leaders had actively promoted vaccination in their area — but that figure ranges from 51% in Gabasawa and 46% in Gaya down to around 39–40% in Nassarawa and the non-sentinel LGAs (Figure 4.75). Nassarawa’s lower figure is probably less about under-mobilization than about the different way these channels function in urban settings; it also has the lowest zero-dose prevalence of the three sentinel LGAs (Figure 4.3, Table 4.51). The real opportunity for deeper leader engagement lies in the rural non-sentinel LGAs, especially the lowest-coverage ones like Tudun Wada and Bebeji (Table 4.51).
Unregistered birth status is one of the two dominant predictors of zero-dose status in the regression analysis, alongside caregiver belief in vaccine importance (Section 4.3). Children born outside a health facility are also significantly more likely to be zero-dose, pointing to a convenience pathway: families who deliver in a facility encounter birth registration and vaccination services at the same place and time. Strengthening the consistent application of that integrated protocol — and extending equivalent outreach to mothers who deliver outside formal facilities — could convert effects the data already show into a more reliable intervention package. The integration of vaccination counselling into birth-registration touchpoints has been shown to raise vaccination coverage in comparable settings (Cooper et al. 2015).
Reminder and recall systems are a useful complement with strong local evidence (Jacobson Vann et al. 2018; Ibraheem et al. 2021; Eze and Adeleye 2015), but the data put scheduling and forgetting at only about 4% of cited reasons for non-vaccination, so they should not be used as the primary demand-side lever. Similarly, the data support continued attention to calibrating messaging to the education and trust profile of each LGA’s caregivers — trust in health information varies sharply by caregiver education (Figure 4.56), and the caregiver population differs meaningfully across LGAs.
The three sentinel LGAs are the natural starting point for rollout: they have the baseline data needed to assess impact, and ward-level hotspot maps identify exactly where to focus first (Figure 4.10, Figure 4.13, Figure 4.16). Across the twelve non-sentinel LGAs, the Pathways vulnerability segmentation offers the most practical targeting tool: zero-dose prevalence rises monotonically from 11% in the lowest-vulnerability urban segment to 58% in the highest-vulnerability rural segment, with statistically significant gradients in both urban and rural areas (Figure 4.18). Most caregivers belong to segments R3.2 or R2, and these higher-vulnerability types are distributed across all LGAs rather than concentrated in a few. The typology is therefore most useful as a within-LGA tool, directing interventions toward the higher-vulnerability households inside each LGA, not as a basis for choosing which LGAs to prioritize in the first place.
5.1.2 Supply-Side Priorities
Three strands of baseline evidence shape the supply-side priorities: dropout patterns, Missed Opportunities for Simultaneous Vaccination (MOSVs), and how reliably caregivers say services are available. Two of the three point most clearly at Nassarawa; the third — MOSV occurrence — is high across all three sentinel LGAs.
Overall dropout along the immunization continuum is low, but concentrated in Nassarawa (Figure 4.19, Figure 4.20, Figure 4.21). The combined Diphtheria-Pertussis-Tetanus (DPT)-1 to DPT-3 dropout is 5.4%, and DPT-3 to Measles-Containing Vaccine (MCV)-1 is 3.4%. Gaya and Gabasawa run below those averages. Nassarawa exceeds them consistently: its Oral Polio Vaccine (OPV)-1 to OPV-3 dropout is 15.9%, against a combined rate of 10.6%, and its DPT-3 to MCV-1 dropout sits at 4.7%. This makes Nassarawa the primary LGA for increasing session frequency, establishing defaulter-tracing protocols, and strengthening routine immunization data quality. The other LGAs need these as standard practice rather than as priority investments.
MOSVs are more prevalent in the sentinel LGAs than in the non-sentinel stratum: between 74% and 80% of children in the sentinel LGAs had at least one MOSV for any dose, compared with 61% in the non-sentinel stratum (Figure 4.33; see also Figure 4.34, Figure 4.35, Figure 4.36). The most striking finding, however, is how rarely those missed opportunities get corrected. In Nassarawa, only 22% of MOSVs are eventually made up — compared with 38% in Gabasawa, 36% in Gaya, and 73% in the non-sentinel stratum (Figure 4.37, Figure 4.38). Closing that correction gap is a quality-of-service priority for all three sentinel LGAs, with Nassarawa first.
The fix involves a fairly well-defined package: ensuring providers consistently screen every contact for due doses, using defaulter-tracing to bring children back, and supervising facilities to audit whether screening is actually happening. This approach has been tested locally: a quality-improvement collaborative in five primary health-care facilities in Nassarawa applied two plan-do-study-act cycles to a baseline MOSV rate of 31.7% and produced a visible decline at two of the five sites (Adamu et al. 2019).
That said, the ceiling on what MOSV correction alone can achieve is modest. The counterfactual DPT-1 estimate (Figure 4.39) shows that eliminating MOSVs and early doses would raise overall DPT-1 coverage by only about 1.8 percentage points (from 51.4% to 53.2%), with the largest single-LGA gain in Nassarawa (~2.7 pp). Supply-side quality improvements matter for equity and service integrity, but they cannot substitute for demand-side and access-side work.
Caregiver-reported availability of vaccination services is broadly similar across areas — 88% in Nassarawa, 81–83% elsewhere (Figure 4.67). Cold-chain reliability, stock management, and provider supervision are worth maintaining and strengthening as part of the MOSV-correction effort, but on this evidence they are not the primary lever for reducing zero-dose prevalence.
The same ward-level hotspot maps used for demand-side targeting apply here, with Nassarawa’s hotspots taking priority on both dropout and MOSV-correction grounds (Figure 4.10, Figure 4.13, Figure 4.16). The study has also produced coverage maps that overlay the hotspot analysis with the locations and service-readiness ratings of immunization facilities at the time of the survey (Figure 4.8, Figure 4.11, Figure 4.14). Those maps give micro-planners a practical tool for cross-referencing zero-dose burden with service availability when sequencing supply-side activities within each ward.
5.1.3 Access-Side Priorities
Access priorities follow from two main evidence streams in the baseline: the spatial hotspot analysis and the travel-time and cost data.
The ward-level hotspot maps for the three sentinel LGAs (Figure 4.8 to Figure 4.13) give the most precise geographic signal of where access investments should land first. Hotspots here are defined as wards that combine elevated zero-dose prevalence with elevated absolute counts — both dimensions matter. The cutoffs used to define hotspot membership were chosen by the research team and differ across LGAs, so targeting decisions should consider the prevalence and absolute layers together rather than relying on either alone. In Gaya in particular, no ward’s lower confidence bound exceeds 50% zero-dose prevalence, which means the absolute-count layer should carry more weight there than the prevalence layer. Follow-up vaccination activities are best planned at the ward level: recent Nigerian geospatial work shows that ward- and catchment-level mapping reveals zero-dose pockets that district-level estimates simply hide (Utazi et al. 2024).
Across the twelve non-sentinel LGAs, the LGA-level coverage data identify the equivalent priorities: Tudun Wada and Bebeji have the lowest coverage in the non-sentinel group, while Gaya and Gabasawa hover near the Demographic and Health Survey (DHS)-Kano 42.2% zero-dose benchmark (Figure 4.3, Table 4.51, Table 4.52). These low-coverage LGAs are the natural first candidates for LGA-level access investment.
Distance to services is meaningfully worse in rural areas. More than half of children in Nassarawa live within one kilometre of a health facility, compared with about 40% in Gaya and Gabasawa and only 29% across the non-sentinel LGAs (Figure 4.60, Figure 4.61). The travel-time maps for each sentinel LGA localize the within-LGA pockets where physical distance is most likely to be driving non-vaccination (Figure 4.62, Figure 4.63, Figure 4.64). The practical implication is that rural LGAs — Gaya and Gabasawa among the sentinels, and much of the non-sentinel sample — need a combination of strengthened fixed sites and periodic mobile outreach to reach settlement pockets that fall outside walking distance of any service point.
The financial-burden picture is more nuanced than a simple story of cost as a barrier. Although vaccination services are officially free, attending a session carries real costs for caregivers — transport, food, time away from work or household duties, and sometimes informal payments at the point of care. Between 55% and 59% of caregivers across survey areas describe paying for vaccination as “very easy,” and only 5–8% describe it as “not at all easy” (Figure 4.78) — suggesting those costs are generally manageable, but not absent. But average per-visit spending varies considerably across LGAs — from ₦681 ($0.44) in Gaya to ₦1,824 ($1.19) in Nassarawa — and the composition of that spending differs too (Figure 4.79, Table 4.83). In Nassarawa, food and transportation are the main drivers of higher costs. That points toward urban informal-cost investigation and supply-side scrutiny as the more relevant response there, rather than rural transport-cost mitigation. Rural LGAs may still warrant some transport-cost support — particularly where outreach scheduling coincides with high-opportunity-cost periods for caregivers (Table 4.29, Table 4.30) — but as a complement to the broader access strategy, not its primary component.
5.1.4 Reaching Households Outside the Formal Health System
A pattern running through the demand, supply, and access analyses above is that the children most likely to be zero-dose are precisely those whose families rarely or never encounter the formal health system. Most of the interventions discussed so far operate within that system — provider supervision, microplanning, defaulter tracing, facility-side counselling. These are well-suited to converting families who do reach a facility into fully vaccinated children. But they hit a structural ceiling at the doorstep of households that never engage with the system in the first place. Reaching those families requires a complementary line of work that meets them on their own terms, before any formal-system contact. Three entry points are worth prioritizing.
The first is birth-attendant linkages. Traditional birth attendants remain the first point of contact with any kind of health system for a substantial share of rural Kano births. A formal referral pathway from traditional birth attendants to the routine immunization schedule — minimal training, simple referral cards, and a feedback loop tracking which referred children actually show up for their first dose — is the most direct way to convert a birth outside the system into a vaccinated child. This sits alongside, rather than replacing, the community-leader engagement and household counselling work on the demand side.
The second is place-of-birth-informed outreach. Ward-level outreach planning typically draws on coverage and dropout indicators. It should also explicitly factor in the share of recent births that took place outside a health facility. Wards with a high non-facility birth share need outreach sessions, mobile teams, and household visits at higher frequency than coverage statistics alone would suggest — because their coverage gap is concentrated in households that standard surveillance never sees.
The third is non-system touchpoints. Religious institutions, market days, and civil-registration events are regularly visited by families who rarely set foot in a health facility. Rather than leaving these to ad hoc campaign use, they should be systematically evaluated as opportunistic vaccination touchpoints. Where they already operate at scale, they offer a low-cost route to first contact with precisely the population the rest of the strategy struggles to reach.
None of this replaces the demand, supply, and access priorities above — it’s the complement that makes the broader strategy reach the people it’s designed for. Operationally, these activities should be sequenced first in the rural sentinel LGAs and across the rural non-sentinel LGAs, where non-facility births are most common and where the existing within-system levers are most constrained.
As interventions roll out, the full portfolio can be re-evaluated against the same baseline indicators. The sentinel LGAs carry the strongest evidence base for sub-LGA targeting today. The non-sentinel LGAs should be sequenced according to their LGA-level profiles across coverage, dropout, MOSV rates, community-leader promotion, and behavioural and social drivers.
5.2 Endline and Impact-Measurement Design
The intervention portfolio eventually leads to a sharper question that state partners and implementing organisations will need to answer: did the Zero-Dose initiative actually reduce zero-dose prevalence in Kano, relative to where this baseline left them?
Answering that question well requires a single measurement event, designed in advance, and powered to detect a change of the size the programme is realistically expected to produce (Lakens, Scheel, and Isager 2018; WHO 2018). The Head-to-Head Comparison Report (2026) narrows the instrument choice considerably: a gold-standard probability survey is the only design evaluated there that produces defensible prevalence estimates at the stratum level, with controlled error and full design-based validity. Lot Quality Assurance Sampling (LQAS) and the modelled-surface estimates are useful between baseline and endline, but they are not substitutes for a benchmark prevalence figure at endline. Network Scale-Up Method (NSUM) and Rapid Convenience Monitoring (RCM) are explicitly ruled out as unsuitable for this role by the methods-comparison findings (Mindset 2026).
The recommendations below take the pre-post-on-hotspot-LGAs design as the recommended endline design and specify the protocol decisions that flow from it.
EL1. Adopt the pre-post-on-hotspot-LGAs design as the endline core. This design aligns the endline directly with the inferential question the Zero-Dose initiative will be evaluated against. The endline’s primary job is to test whether the prioritized interventions moved the four pre-defined baseline benchmarks. That test is sharpest at the strata where a benchmark exists and where baseline variances are large enough to support a pre-post difference test without inflating sample sizes in pursuit of unnecessary precision. Expanding to a full fifteen-LGA repeat would widen the analytic scope without strengthening the test of the headline hypothesis — and it would pull budget away from the lighter monitoring layers — LQAS ward classification, modelled-surface refresh, hotspot re-checks — that produce more decision-relevant signal in the intervening years. For programmes that do want LGA-level endline estimates beyond the four baselined strata, the pragmatic compromise is to keep the recommended pre-post core and add a targeted top-up sample in a small number of additional LGAs where a specific decision genuinely turns on the result, rather than expanding uniformly across all fifteen.
One important caveat applies to any design that selects analytic areas based on baseline hotspot status: by definition, hotspots are chosen because they had elevated zero-dose prevalence, so any follow-up comparison is susceptible to regression to the mean — apparent improvement that reflects statistical reversion toward the population average rather than genuine intervention effect. The recommended remedy is to avoid using baseline hotspot data as the “pre” observation in a naïve pre-post comparison. Instead, vaccine histories recorded at the endline visit can reconstruct uptake for children whose vaccination window fell after the intervention start date, allowing a within-endline pre/post split that does not condition on baseline extremes.
It is also worth distinguishing between two related but analytically distinct activities. A hotspot follow-up is a targeted, lower-cost resurvey of the highest-burden wards, designed to test whether conditions improved where intervention was most concentrated. A full endline repeats the baseline sampling frame across the sentinel LGAs to support LGA-level pre-post comparisons against the four baseline benchmarks. Both serve legitimate purposes but answer different questions; the pre-post core recommended here is an endline, not a hotspot follow-up.
EL2. Use the 6–23-month age cohort as the primary analytic frame. The baseline used the standard 12–23-month reporting cohort. The underlying survival curves, however, show that zero-dose status stabilises by around six months of age, with almost no further movement between six and 23 months once the early-infancy window for Penta-1 timely receipt has passed. Building the endline on the 6–23-month cohort recovers essentially the same prevalence signal while drawing on roughly 50% more eligible children per cluster. This means the endline can either hold effective sample size constant and reduce the number of doors knocked — a real saving on a design where building-footprint enumeration governs field duration — or hold door-knocking volume constant and use the broader cohort to sharpen the minimum detectable change. The 12–23-month subset still serves as the calibration check and as the figure that aligns with DHS/Multiple Indicator Cluster Survey (MICS) and Gavi reporting conventions.
EL3. Keep Penta-1 as the headline zero-dose indicator, and report the true zero-dose rate when available and practical. The baseline reported two parallel zero-dose definitions: a true zero-dose definition based on receipt of any vaccine (card or recall), and a Penta-1 zero-dose definition aligned with Gavi’s operational target. After six months of age the Penta-1 proxy captures approximately 88% of truly zero-dose children. Penta-1 should remain the headline endline indicator — it is the Gavi-aligned reporting figure, it maps directly to what the intervention package targets, and it is the most comparable figure across the international zero-dose literature (Aheto et al. 2023; Sbarra et al. 2021). Where a full vaccination-history tool is administered, reporting the true zero-dose rate alongside adds little marginal burden, since both indicators come from the same questionnaire — though not all methods (such as LQAS) routinely collect a full routine immunization history, and the additional respondent and interviewer time should be weighed against the survey’s overall burden. If the two diverge at endline — a widening or narrowing of the ≈12-percentage-point gap the baseline documents — that is a useful diagnostic that signals whether the programme shifted catch-up vaccination behaviour more broadly or only timely Penta-1 receipt. A technical note from this study’s vaccine-correlation analysis is worth noting alongside this convention: Bacille Calmette-Guérin (BCG)-0 is in fact the antigen most highly correlated with true zero-dose status in our data. The gain over Penta-1 is not large enough to justify breaking with the global Gavi reporting convention, but it is worth flagging in the protocol so that future analyses can choose between them with full information.
EL4. Plan for a smaller endline sample than the baseline. Two considerations support a smaller endline envelope. First, due to conservative assumptions and misleading signals from the pilot, this baseline sampled more children than needed to meet the original sampling goals; there is no methodological reason to repeat that at endline. Second, and more substantively, the pre-post test does not require the endline to be powered as a first-cycle standalone benchmark. Because the baseline already contributes its share of the change-detection variance, the endline sample can be considerably smaller and still deliver adequate power for a minimum detectable change of programme-relevant size (Lakens, Scheel, and Isager 2018; Kish 1965). The corresponding effective sample sizes can be backed out from the baseline bootstrap replicates, and the endline protocol should report the minimum detectable change explicitly so that the implied resource envelope and the implied power are visible to all parties.
EL5. Confine endline reach to the existing sentinel and combined non-sentinel strata. The endline reuses the sentinel-LGA arm at full design intensity and the combined non-sentinel stratum as the comparison group, without creating new LGA-level strata. This keeps variance comparable across the two rounds and avoids structural changes that would muddy the pre-post difference test.
EL6. Use baseline prevalence estimates to right-size the endline sample. When planning a survey, sample sizes are typically calculated using a conservative assumption of 50% prevalence — the point at which uncertainty is highest and sample requirements are largest. The baseline now provides area-specific zero-dose estimates that are far more informative than that default. Where prevalence is well below 50% — as it is across most sentinel LGAs — fewer interviews are needed to achieve the same statistical precision, and the endline can be sized accordingly rather than applying a uniform conservative envelope. Building-footprint sampling should be reused as in the baseline to preserve frame comparability and the implementation-fidelity audit trail that geofencing provides (Jean Baptiste et al. 2024; Utazi et al. 2024).
EL7. Add intervention-interaction probing questions to the endline instrument. Before the endline is fielded, the Gates Foundation (GF) and implementing partners should provide a comprehensive list of which interventions have been delivered in which communities since baseline. The endline instrument should then include probing questions about whether the surveyed household has, to its knowledge, had any interaction with each major intervention category. This serves two purposes. First, it adds an analytic lever for the impact evaluation that does not depend on observing vaccination services per se: even where vaccination is the outcome, intervention contact is the most direct programme-level mediator, and tracking it explicitly enables subgroup analyses than a vaccination-status-only instrument can support. Second, it gives implementing partners a feedback loop on community-level uptake of their own programming, which is useful for adaptive management between endline rounds.
EL8. Time the endline two to three years after intervention rollout completes, on the same survey-month calendar as the baseline. The endline should be fielded approximately two to three years after intervention rollout completes in the sentinel LGAs, using the same survey-month calendar as the baseline so that age-cohort definitions and seasonality remain comparable. This recommendation is qualitative rather than calendar-dated because the rollout schedule is not yet fixed; the locking decision is length of post-rollout exposure, not absolute date, and should be agreed with implementing partners before sample-size finalisation. A common pitfall in pre-post designs is fielding the endline too close to rollout completion: the effect under measurement has not yet had time to express itself in 12–23-month prevalence, and the comparison reads as a null that is really a timing artefact.
EL9. Lock the protocol decisions before fieldwork begins. A short endline protocol document — fixing the minimum detectable change, the post-stratification controls, the sample-size envelope, the post-rollout exposure window, the intervention-interaction probe set, and the agreed numeric coverage targets per stratum — should be agreed between the GF, the state task force, and the implementing partners before fieldwork begins (Lakens, Scheel, and Isager 2018; WHO 2018). That protocol is the bridge between the intervention prioritization in this report, the methods-comparison conclusions in the head-to-head report, and a defensible answer to the headline question the Zero-Dose initiative will ultimately be held to.