Vaccination strategies for measles control and elimination: time to strengthen local initiatives

WHO-UNICEF national estimates of immunization coverage (WUENIC) show that global coverage of the first dose of measles-containing vaccine (MCV1) coverage soared from 16% in 1980 to 68% in 1989, rose slightly to 71% in 1999, increased to 83% by 2009, and then stabilized at 85% from 2015 to 19. All 17 countries in AFR and EMR with WUENIC MCV1 < 50% in 2000 had increased coverage by 2019 to a median of 60%, but their coverage remained far below levels needed for elimination. Furthermore, coverage remained very low in 2019 in Angola (51%), Cameroon (60%), Central African Republic (49%), Chad (41%), DRC (57%), Ethiopia (58%), Guinea (47%), Somalia (46%), South Sudan (49%) and Nigeria (54%) [11]. A second dose of MCV (MCV2) was rarely part of routine schedules in countries eligible for support from the GAVI Vaccine Alliance (GAVI) until 2010 since when introductions have accelerated with 60% WUENIC MCV2 achieved by 2019.

Geospatial analyses of survey data have allowed estimation of subnational and local patterns of MCV1 coverage [12‐14] and show that coverage can vary substantially within countries and over time. Figure 1 shows the results of geospatial analyses of survey data from 2000 to 2019; details of the methods are described elsewhere [12]. Although MCV1 coverage in most districts and countries increased from 2000 to 2019, there is a band of districts across AFR from Nigeria to Somalia, additional large areas of Guinea and Angola and parts of Afghanistan and Pakistan where coverage was estimated to be below 50% in 2000 and still below 50% in 2019 [12]. MCV1 coverage remained below 80% in both 2000 and 2019 at the district level in much of AFR, EMR and parts of most countries in SEARO. By contrast, few districts in the Americas had estimated MCV1 coverage persistently below 80% in both 2000 and 2019.

Clustering of unvaccinated individuals poses risks for local disease outbreaks but could also facilitate targeted interventions. Known contributors to spatial inequity include remoteness, conflict and urban slums. For countries with available data, Fig. 2 shows the estimated geospatial distribution of children of the target age who did not receive MCV1 in 2017, in relation to conflict-affected, urban and remote rural areas. Direct comparison between countries is limited by potential differences in the completeness of data on conflict, but it is clear that most unvaccinated children in EMR live in conflict-affected areas, as do those in some of the largest African countries such as DRC, Ethiopia and Nigeria. A high proportion of unvaccinated children live in remote rural locations in Chad, DRC, Ethiopia, Mauritania and the Republic of Congo (Fig. 2a). Overall, a relatively small proportion (~ 10%) of unvaccinated children lived in urban areas, but analyses to date have not distinguished the urban poor from other urban populations. Of note, none of these factors is identified for over half the unvaccinated children in many of the higher coverage countries (Fig. 2b). Other socio-economic and cultural factors have recently been found to be consistently associated with higher vaccination uptake around the globe, including high confidence in vaccines, high trust in health care workers, higher levels of science education, younger age and high information-seeking behaviour, while in some countries, belonging to a minority religious group or refusing to state religious belief was associated with lower uptake [22]. Even when mothers are willing to accept vaccination, coverage may be low due to health systems barriers such as inadequate communications about vaccination, unreliability of sessions and high transport costs to reach them [23]. These analyses suggest that current RI programmes often produce unequal levels of RI throughout a country and fail to reach children in high-risk populations. Of note, during the elimination program in the Americas, high-risk groups were identified using coverage and surveillance data and targeted efforts developed to reach them, with cross-border collaboration where needed [2].

For countries to reduce reliance on SIAs, routine services must be capable of providing high and equitable coverage with MCV1 and MCV2. Where this is not the case, measles continues to cause major morbidity and mortality. From 2013 to 2018, the highest incidence rates occurred in low- and middle-income countries (LMICs), especially those that had low or zero historical coverage of MCV2, an indicator of RI strength [6]. Highest incidence rates were in unvaccinated pre-schoolers in each region, although incidence was also high in older persons in some LMICS such as Madagascar and in high-income countries, where outbreaks followed many years of low or absent transmission.

From 2000 to 2019, AFR reported the vaccination of 1.3 billion children via SIAs, EMR 700 million and SEAR 750 million [24]. According to data from 81 post-campaign coverage surveys (PCCS) reported to WHO from these regions, 56 (69%) of SIAs reached at least 90% of the target population [24]. Although SIAs can attain higher and more equitable coverage than RI [13, 25, 26], SIA coverage in countries with weaker health systems has rarely approached the levels that would be needed for elimination [26, 27]. Furthermore, the extent to which SIAs reach children missed by RI is a key determinant of impact but has only recently begun to be evaluated. PCCS are encouraged to report these data but surveys often exclude conflict-affected areas and ascertainment of prior vaccination status is of unknown accuracy [26].

To reduce transmission, SIAs need to increase population immunity. Because SIAs target many individuals who will already have been eligible for RI or had measles infection, the effective increase in population immunity is much lower than the nominal coverage of SIAs. Estimates of SIA impact on population immunity provide more complete information. Trentini et al. [28] incorporated data from sero-surveys in dynamic transmission models to estimate that SIAs generated about 45% of the immunized fraction of the population in Ethiopia and about 25% in Kenya (which had higher RI coverage) in 2015. Thakkar et al. [29] used measles surveillance data in their models to estimate that SIAs conducted between 2012 and 2017 in Pakistan immunized 40% of the susceptible individuals reached by the campaigns. In China, similar analyses estimated that provincial-level SIAs became more effective over time. Those conducted from 1996 to 2005 resulted in estimates of 0.5% to 45% reductions in susceptible individuals in the target age classes while SIAs from 2006 to 11 resulted in 32% to 87% reductions [30]. These useful estimates, however, are available for very few LMICS.

High-quality SIAs can help to eliminate measles, but to maintain elimination—unless routine coverage of MCV1 and MCV2 is extremely high—SIAs must be repeated frequently enough to maintain the susceptible population below the herd-immunity threshold. Predicting when to conduct follow-up SIAs is challenging. As vaccination programmes improve, measles incidence varies more from year to year, highlighted by recent outbreaks in Madagascar, Mongolia and others [6]. This increase in inter-annual variation is predicted by theoretical models of the epidemic dynamics as countries approach the elimination threshold [31]. Periodic additional interventions, such as SIAs and outbreak response vaccination, may reduce mean incidence over many years but lead to larger outbreaks in any 1 year in settings where RI is not sufficient to prevent the rapid increase of susceptible cohorts following campaigns [32, 33]. This increased variation in periodicity can make it harder to decide when to conduct follow-up SIAs.

The optimum timing of SIAs often also varies within a country. The minimally sufficient SIA interval depends on both the birth rate and the RI coverage, which govern the decrease in population immunity in between campaigns [33, 34]. Sub-national variation in birth rates and vaccination coverage mean that SIA intervals based on national rates may be either insufficient or more frequent than necessary to maintain elimination in any sub-national unit [33, 35].

To compound difficulties in choosing the desired interval, delays to planned SIAs are not uncommon, e.g. due to delays in obtaining funding and logistical support, political changes, natural or man-made disasters [36, 37]. Appropriate timing of SIAs is critical to immunize susceptible individuals before an increase in transmission occurs [29]. Major outbreaks have been reported shortly before a scheduled follow-up SIA in e.g. Burkina Faso, DRC and Kenya [36, 37] while delays in implementing outbreak response vaccination campaigns may allow widespread transmission to occur [38].

Along with immunization coverage, demographic changes can drive trends in measles incidence. Merler et al. [39] and Li et al. [30] illustrated that measles incidence declined faster than expected by vaccination coverage alone due to concomitant reductions in birth rates in Italy and China, respectively. Where birth rates are high, RI coverage must be higher [40] and SIAs must be more frequent [34] to maintain a given level of population immunity. From 1980 to 2018, crude birth rates per 1000 population fell much less in AFR than in other regions (from 46.7 to 34.9 in AFR, 42.4 to 25.7 in EMR and 35.6 to 17.6 in SEAR). From 2000 to 2018, the population aged under 15 years increased by over 50% in AFR, by about 24% in EMR and remained stable in SEAR. At given RI and SIA coverage, this increase means a greater density of unvaccinated children and higher risk of measles transmission.

Regional and national demographic summaries may mask significant subnational variations in risk profiles. For example, high rates of rural to urban migration have seen urban areas grow more rapidly than rural areas, with differing demographic profiles [41]. The influx of susceptible persons, from areas with lower access to vaccination but also low measles transmission, to crowded urban areas facilitates measles transmission [42, 43], especially if recent migrants are not recognized officially and vaccinated in a timely fashion. Cyclical rural-urban migration, as illustrated in Niger [44] and Pakistan [29], may further limit access to RI and affect the performance of SIAs. Improved socio-demographic data need to be incorporated into strategic planning of measles elimination.