ARI        acute respiratory infection

IPD        invasive pneumococcal disease

IQR        inter-quartile range

LCB        lower 1-sided 95% confidence bound

LQAS     lot quality assurance sampling

PCV       pneumococcal conjugate vaccine

PCV13   13-valent pneumococcal conjugate vaccine

WHO      World Health Organisation

Acute respiratory infection (ARI) is a leading cause of death in children aged 0–59 months old worldwide. Pneumonia, a severe form of acute lower respiratory infection that affects the lungs, claims around 0.9 million lives of infants and young children annually ¹ . Pneumonia is more common in developing countries, specifically those in Africa and South Asia ^{2, 3} . Streptococcus pneumoniae and Haemophilus influenzae are the most common causative agents of bacterial pneumonia in children and are targeted by the Pneumococcal conjugate vaccine (PCV) and the H. influenzae B vaccine, respectively ^{2, 3} . PCV also prevents invasive pneumococcal disease (IPD) ⁴ and pneumococcal meningitis ⁵ . In 2007, the World Health Organisation (WHO) recommended PCV be added to all national immunization programmes, especially in countries with high child mortality ³ .

Malawi recorded around 500 ARI cases per 1000 under-5-years population between July 2009 and June 2010, of which approximately 1.5% died, according to health management information system statistics ⁶ ; the death rate is likely to be an under-estimate, as it is based on health facility data only. In November 2011, Malawi was among the first countries in Africa to introduce the 13-valent Pneumococcal Conjugate Vaccine (PCV13) ⁷ into its routine immunization program, aiming to protect millions of children from pneumococcal pneumonia and IPD. The vaccine was introduced as a three-dose schedule at 6, 10 and 14 weeks, with an initial catch-up campaign targeting all children up to the age of 1 year. A target national coverage rate of ≥90% and local district targets of ≥80% were set for each year from 2012 to 2016 ⁸ .

We used a series of household surveys to assess changes in vaccine coverage in a rural area of central Malawi between March 2012 and June 2014. The surveys were part of a wider study exploring changes in the health-system burden of pneumonia following PCV13 vaccination introduction in two areas of central Malawi ⁹ . We used the surveys to estimate local and area-wide vaccination coverage in one of our study areas to assess the burden of pneumonia at the community health worker, health centre and hospital levels in relation to the percentage of infants vaccinated. We aimed to quantify how the burden of pneumonia on the health system changed as vaccination coverage increased, and to track the roll-out of the PCV13 vaccine in Malawi in terms of the timeliness of each of the three doses in relation to the 6, 10, 14-week schedule and population coverage of each dose over time. In this paper we present the results of six surveys conducted in Lilongwe district.

Table 1 details the overall findings for age and coverage of vaccination from all surveys. A total of 8,562 infants were recruited. Most infants (78–83%) were aged 4–16 months old at each time point. Overall 96.3% of the infants had health passports. The three-dose coverage of PCV13 vaccine in age-eligible children increased from 10.0% in March 2012 to 86.0% in June 2014. Similarly, there was also an increase during the same time period of the coverage of PCV13 doses 1 and 2 in age-eligible children from 62.0% and 30.3% respectively in March 2012 to 93.9% and 85.2% by June 2014. In all six surveys older infants (aged 4–16 months) were much more likely to be vaccinated (with one, two or three PCV13 doses) than infants aged less than 4 months old. Vaccination timeliness for the first, second and third PCV13 doses increased from 11%, 9% and 5% in March 2012 to 47%, 56% and 48% in June 2013 and 49%, 61% and 62% in June 2014 ( Table 1). The median ages at first, second and third doses mirrored this improvement, dropping throughout the six surveys and rapidly so in the first three surveys; overall they were 9 weeks (interquartile range (IQR), 7–13 weeks), 15 weeks (IQR, 12–20 weeks), and 21 weeks (IQR, 18–27 weeks; Table 1).

Figure 2 shows how the coverage of the first two doses of PCV13 changed in each basin from the first survey in March 2012 to the sixth and last survey in June 2014 in 6-week to 16-month-old babies. In March 2012, all but one (Muyande) of the 20 basins had a two-dose coverage below 50%, as indicated by the confidence intervals for the estimated coverage being below, or overlapping 50% coverage. By June 2014 all of the basins had coverage above 50% and eight of the 20 basins had achieved ≥80% two-dose coverage, as indicated by the lower confidence interval for the estimated proportion vaccinated being above 80% ( Figure 2, basins in order of June 2014 coverage). All basins combined were also estimated to have a two-dose coverage above 80% in June 2014 ( Figure 2).

In March 2012 coverage of all three doses of PCV13 was more than 95% likely to be <40% in all 20 basins ( Figure 3). By June 2014 three-dose coverage was more than 95% likely to be ≥60% in all 20 basins, ≥70% in 16 basins, and met the ≥80% local target in 11 basins ( Figure 3).

The two-dose and three-dose PCV13 coverage was consistently lower in 6-week to 4-month-old infants than in 4–16-month-old infants ( Table 1 and Supplementary Figure 1– Supplementary Figure 4).

Figure 4 shows the trend for the overall two-dose and three-dose coverage by age, over-time. The March 2012 survey shows that the two-dose coverage was higher (≥50%) in infants aged 6 and 7 months old but lower (≤40%) in infants aged 8–16 months. The two-dose coverage increased to ≥90% in June 2014 for all infants aged 6–16 months. The three-dose coverage was lower (<20%) among all infants aged 6–16 months old in March 2012 but increased to >80% by October 2013 and >85% by June 2014. All raw data associated with this study are freely available from the UK Data Service ¹⁷ .

Our results show that three-dose PCV13 vaccination coverage in Kabudula health area met the local district target of ≥80% ⁸ by the second year of the vaccination campaign, but only in older infants. By October 2012, the end of year 1 of the PCV vaccination program in the study area, the survey found that the overall three-dose coverage in age-eligible infants was at 64% (Infants <4 months old: 8%, Infants over 4 months old: 68%), far below the national target of ≥90% and the local district target of ≥80% ⁸ . Our analysis shows that none of the basins had three-dose coverage of ≥80% after year 1 ( Supplementary Figure 5). The three-dose coverage in age-eligible infants increased to 80% (infants <4 months old, 9%; infants over 4 months old, 84%) by the end of year 2 (October 2013 survey), and 86% (infants <4 months old: 30%; infants over 4 months old: 90%) by June 2014, two-thirds through year 3 ( Table 1). In total, 14 of the 20 basins had >80% coverage in 4–16-month-olds in June 2014 ( Supplementary Figure 4).

We found timeliness of PCV13 vaccination to be low during the first year of the vaccination campaign, probably reflecting the initial catch-up campaign of vaccinating all infants under 1 year. By June 2013 we found that around half of infants were vaccinated within the standard schedule for each dose but only around one-quarter were receiving all three doses within the first 6 months of life as per schedule. This rose slightly by June 2014 to 49% for the first dose, 61% and 62% for the second and third doses and 36% for all three doses ( Table 1). This is lower than reported by a recent study in Karonga, which found ~80% of eligible infants were vaccinated with all three PCV13 doses by 26 weeks of age ¹⁸ . Virtually all (99.8%) of the infants who were not vaccinated within the schedule were vaccinated late rather than early. Our results are comparable to the findings of Babirye et al. in Uganda, which used the same definitions of timeliness for the three doses of polio and pentavalent vaccination but found greater timeliness of each dose ranging from 71% for polio 1 to 78% for polio 2 and pentavalent 2 ¹⁹ . In 2004 Malawi was found to have 52% coverage of DPT3 (diphtheria-pertussis-tetanus vaccination, third dose) by 6 months of age ¹⁴ .

The three-doses-within-the-first-6-months schedule is recommended in settings such as Malawi, where severe vaccine-preventable disease is common in younger infants ²⁰ . Existing vaccine schedules at 6, 10 and 14 weeks, such as those for polio and pentavalent vaccine also make it logistically easier ²⁰ . Delayed vaccination extends individual susceptibility to illness and reduces herd immunity ²¹ . However, whether an initial two-dose plus booster (2+1), initial three-dose (3+0) or a three-dose plus booster (3+1) schedule is better at preventing clinical pneumonia remains unresolved ²² . A recent trial in South Africa, which has similar pneumonia epidemiology to Malawi, found a 2+1 PCV schedule of 6, 14, 38 weeks to significantly reduce IPD ⁴ .

There are some limitations to this study. Firstly, the sample size for the infants aged 6 weeks to 4 months was less than half of the intended sample of 600 infants (30 per basin) for all the surveys conducted. This was due to there being insufficient infants of this age group in each basin at the time of survey. However, given the low numbers who were vaccinated and the high probability of low coverage thresholds not being met given the estimated confidence intervals, we can be fairly sure that the coverage was low in this age group. Secondly, the self-weighted analysis here implicitly assumes that every eligible respondent had the same chance of being selected, but this is not so in this case, as the interviewers always started at the centre of the village. Furthermore, for confidence intervals to be meaningful, the sample should be a probability sample and the analysis should incorporate appropriate sampling weights. This simple self-weighted analysis of a somewhat biased sample may not meet the strict requirements for confidence intervals. However, the six surveys used the same method of data collection each time, so the biases over time are likely to be constant. The coverage results also improved so dramatically over time that our broad-stroke conclusions are likely to be robust. The villages are small and a large proportion of the eligible children were sampled in each village so the improvements in coverage and the persistent shortfalls in timeliness described here are likely to be real, although subject to bias. Thirdly, due to a large proportion of eligible children being surveyed in some basins, some of the same children in the older 4–16-months age group may have been sampled in more than one survey, introducing some correlation to the repeated cross-sectional samples, meaning our assessment of trends in vaccine coverage may be less representative of the overall trend in the population. This is likely to be a minor issue given that the surveys were 4–8 months apart. Fourthly, the studies were only conducted in one of the six health areas of Lilongwe district; therefore the results may not be generalizable, unless compared with other studies carried out during the same period in other parts of the country. Kabudula is however, fairly representative of rural Lilongwe district and Malawi in general, being found to have vaccination rates around the average for the whole country in 2007, although variation in local vaccine stocks and logistics is an important consideration ²³ . Fifthly, although 96% of infants had vaccination data photographed from health passports, meaning the potential of recall bias to affect our results is low, there were a few anecdotal reports of vaccination records being filled before vaccinations were administered. Vaccine stock-outs at clinics or health facilities mean it was possible that a health passport could indicate vaccination when the infant was not vaccinated. From discussions with health facility and immunization staff, we were assured that this was a relatively rare occurrence. Mothers also verbally verified their infants were vaccinated in almost all cases where vaccination was indicated on the infant’s health passport. Finally, tracking coverage in each basin can be difficult—the same basins did not always have lower coverage ( Supplementary Figure 5), likely due to changing community and health-system dynamics, as well as due to artefacts of the relatively small sample sizes in each basin.

Before the advent of PCV13, it was estimated that PCV13 would protect against 63% of all circulating invasive pneumococci and 78% of those found in children under 5 years in Malawi ²⁴ . We previously published that at 76% three-dose PCV13 coverage, compared to 0% three-dose coverage, and among children <59 months of age, cases of pneumonia associated with low oxygen levels (i.e., hypoxemia) decreased by nearly 50% and hospital pneumonia fatalities declined by more than one-third in central Malawi ⁹ . Furthermore, the proportion of pneumonia cases with clinical danger signs associated with high risk of mortality fell by nearly two-thirds in Malawi after the introduction of PCV13 ⁹ .

Meeting and maintaining the 90% PCV13 coverage target in all areas of Malawi in the coming months and years, and improving the timeliness of vaccination in young infants, has great potential to reduce the burden of PCV13-preventable disease. Monitoring the progress towards this goal via household surveys of PCV13 coverage is important.

All aggregate data is reported in the manuscript and supplementary files. Individual anonymized data from all six waves of the survey and the Stata software code used to produce all tables and figures is freely available via the UK Data Service: https://doi.org/10.5255/UKDA-SN-853265 ¹⁷ . This UK Data Service deposit also contains the data collection tool.

Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC-BY 4.0).