Introduction

Since 1974, the Expanded Programme on Immunization (EPI) has worked to establish immunization programmes throughout the globe [1] with the aim of ensuring that all children worldwide are able to benefit from life-saving vaccines [2]. The EPI has successfully increased global vaccination coverage, leading to substantial declines in child morbidity and mortality from vaccine preventable diseases (VPD) [3,4] and saving countless lives. Every nation in the world has a national immunization programme [2,5]. In lower-income countries, 81% of children now receive routine EPI vaccines [6]. Nevertheless, extending routine vaccination coverage to all children and communities, especially in many low- and middle-income countries (LMICs), has proven to be a demanding challenge with a global estimate of 25 million children being left un- or under- vaccinated [7,8].

In simple terms, unvaccinated children are those who have not received any routine vaccine and under-immunized are those who have missed at least one recommended vaccination. To align measurement efforts, allocate resources and define success, zero-dose (ZD) children are more closely defined as children who did not receive the first dose of diphtheria-tetanus-pertussis containing vaccine (DTP1) and under-immunized (UI) children are defined as those who did not receive the third dose of DTP containing vaccine (DTP3), usually measured at age 12 months or older [9,10]. In countries like Bangladesh, where DTP has been replaced by Pentavalent vaccine, which includes antigens against DTP as well as hepatitis B and Haemophilus influenzae type b, ZD children may be defined as those who did not receive the first dose of pentavalent vaccine (PENTA-1) and those who missed the third dose of pentavalent vaccine (PENTA-3) may be considered UI. We followed these definitions of ZD and UI for our research.

It is estimated that one out of every five children in the world is ZD or UI and therefore, susceptible to virulent VPD [8]. Approximately, 50% of ZD children reside in three geographical contexts: informal or peri-urban, remote-rural, and conflict areas [6]. Despite growing efforts over the past decade to expand immunization, the number of ZD children declined by only 4.1% from 2000–2021 [8] and the COVID-19 pandemic led to severe disruptions in access to routine childhood vaccination. UNICEF estimates that between 2019 and 2021, 67 million children missed at least one routine immunization and about 48 million of these children were ZD [8].

Since its establishment in 2000, Gavi, the Vaccine Alliance, has supported low-income countries to introduce new vaccines and vaccinate over one billion children worldwide [10]. Recently, Gavi, under its Gavi 5.0 strategy and the global Immunization Agenda 2030 have focused on immunization equity or reaching ZD and UI children and their communities. Gavi set up Country Learning Hubs (CLHs) in Bangladesh, Mali, Nigeria, and Uganda with the aim of using evidence to understand the factors influencing implementation and performance of approaches to identify and reach ZD and UI children.

In Bangladesh, where the highly successful EPI program resulted in an increase in full vaccination coverage from less than 2% in 1985 to 83.9% in 2019, communities remain where as many as 15% of children are un- or under-vaccinated [11]. Identifying and addressing these pockets of ZD and UI children is critical to improving immunization coverage and ensuring population protection against VPD. The Bangladesh CLH is tasked with developing the evidence base for tools and approaches that can be used to reach these children and their communities. This paper describes the results of the preliminary activities of the Bangladesh CLH to identify ZD and UI children and detect missed communities—both the methods used and the results—as part of a rapid assessment.

Methods and approaches

Between December 2022 and May 2023, we collected and analyzed both primary and secondary data to identify potential missed communities and barriers to immunization as the first step in an implementation research study. We identified missed communities using a three-step process: 1) identification of high prevalence of ZD upazilas (i.e. sub-districts), 2) identification of potential pockets of ZD within the selected upazilas, and 3) conduct lot quality assurance sampling (LQAS) survey to confirm whether or not the pockets were high prevalence. We conducted qualitative mapping to describe the characteristics of the missed communities, quantitative analysis of secondary data to identify factors related to ZD and UI, and interviews with service providers and mothers of children to identify barriers to immunization. Each method is described below.

Results

Identifying Missed Communities using CES and Routine Service Delivery Statistics depicts the results of our secondary data analysis (Table 1). At the initial stage of identification of missed communities, we found that Jamalganj upazila of Sunamganj district at Sylhet division had the highest ZD prevalence (11.1%). However, when we verified our initial findings and updated the data as described in the above section, we found that Gaffargaon Upazila of Mymensingh at Mymensingh division, in fact, had the highest prevalence (15.7%) with Jamalganj upazila placing second in regards to ZD prevalence. The top 30 upazilas were located across 21 different districts in all eight divisions. The average estimated prevalence of ZD was lower in the top highest prevalence zones in CCs at just over 1% compared to the average of 4% found in the 21 highest prevlance districts. This may be because urban areas have more socioeconomic diversity and this mutes the effects of ZD at the population level, even though there are pockets that are higher ZD prevalence.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Top high ZD upazilas identified after updating routine EPI data.

https://doi.org/10.1371/journal.pone.0312171.t001

This is further illustrated in Fig 1 below which pictorially shows the top high ZD upazilas we identified from our secondary data analysis.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Geographic distribution of high ZD upazilas identified using Routine EPI data (2022).

https://doi.org/10.1371/journal.pone.0312171.g001

Socio-economic Determinants of ZD and UI

The determinant analysis found that mothers’ education, occupation, number of antenatal visits made, mother’s access to media and geographical location were associated with children being ZD or IU (Table 2). In contrast, type of residence and wealth quintile were observed not to have any significant association with children being ZD or UI.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Socio-economic determinants for children being ZD and UI [BDHS 2017–18].

https://doi.org/10.1371/journal.pone.0312171.t002

The possibility of a child being ZD or UI was significantly lower if the child’s mother had more than primary education. Children of mothers who made more antenatal visits or of working mothers were less likely to be ZD and UI compared to their counterparts. We also observed that geographical location played a part, children living in Sylhet are more likely to have ZD or UI children than those residing in Rangpur.

These associations were all seen across both ZD and UI children. One factor, however, was only associated with children being ZD (and not with UI), that is: mothers who reported that “their last child was wanted” were less likely to have a ZD child. The details of the determinant analysis along with crude and adjusted odds ratio (AOR) of ZD and UI by the covariates can be found in (S1 and S2 Tables).

Results from LQAS data analysis

Findings from the LQAS analysis confirmed the presence of missed communities across the districts identified in the secondary analysis (Table 3). A total of 18 clusters (16 from eight upazilas and two from one CC) were accepted as high either high ZD or UI based on the LQAS survey.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Confirmation of ZD and UI areas using LQAS.

https://doi.org/10.1371/journal.pone.0312171.t003

Table 1 also depicts the list from which we identified potential areas where we wanted to conduct the LQAS for final selection of study areas where we will conduct implementation research (IR) as the next step to bring the success of Bangladesh CLH to fruition. We prioritized areas with high ZD and UI for study area selection but we considered districts found to have at least two upazilas with ZD children in a district/CC. This was done to enhance the study area bifurcation into intervention and comparison upazilas for the IR. Table 3 shows the results of LQAS. In LQAS sample, the prevalence of ZD and UI among the accepted cluster were 7.7% and 24%, respectively.

We were unable to conduct initial LQAS in Rangamati due to security concerns, however, we were later able to gain support from governmental officials and decided to include Rangamati as an implementation research (IR) study site.

Overall, we prioritized ten upazilas from five districts and two zones from one CC for the implementation resrarch. These five districts and one CC reflect the variety of Bangladesh’s ecological settings: haor (wetlands), hilly, char (sandy/silty land surrounded by water), coastal, plain land and urban areas are shown in Table 4 below.

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Confirmed missed communities.

https://doi.org/10.1371/journal.pone.0312171.t004

Based on the LQAS survey, prime reasons for not having their child vaccinated included the child being sick, vaccination not being permitted by her husband, financial limitations and busy schedule of parents. Based on the observational mapping of the clusters, most clusters had primary schools but fewer had secondary and/or missionary educational institutions. Notably, the coastal clusters of Noakhali coastal, did not have any educational institutions. With regard to healthcare facilities, none of the clusters had an NGO clinic and only one cluster from an upazila of Gaibandha (Char area) and one cluster from upazila of Noakhali district had public/government health centers. Private clinics were only found at the urban clusters. Bus was the transport system only in the urban zones while boats were a prominent mode of transport used by individuals in the haor and char clusters whereas rickshaws, compressed natural gas (CNG), motorcycles and cars were used more in the coastal, plain and urban clusters. The LQAS also showed that the majority of households in almost all the clusters were mud-built (i.e. kacha), however in urban clusters households were mostly the slum shacks and mud-built followed by half-cemented residences.

Discussion

This study reports the methodologies and initial findings from a rapid assessment of ZD and UI children in Bangladesh conducted by the CLH. The rapid assessment is part of a larger implementation research study that will pilot targeted approaches to reaching ZD children. As well as outlining the ZD and UI challenge, these findings also help identify ZD areas that could serve as implementation research sites. Given that this was an aim of the assessment, it was not designed to find all pockets of ZD in the country, or even within the identified upazilas. However, the approach used is a potential method for identifying pockets of ZD or UI children at the local level, with some validation. The initial findings on demand- and supply- side drivers of ZD and UI are in part of specific to Bangladesh (such as ecological factors) but are also consistent with the drivers of ZD found in other settings.

Our findings demonstrate the utility of combining existing survey and administrative data for identifying missed communities, but also highlight the importance of field validation to confirm the results. This combination approach was particularly useful for ensuring that we chose the most relevant clusters of ZD children to use as intervention and comparison sites for future implementation research. That said, current surveys (e.g. CES, BDHS) are limited in their ability to provide information at the upazila and zone level needed to address ZD. However, the combination of utilizing both survey and administrative data along with application of LQAS process as monitoring tool was the innovative strategy attempted by us to identify the missed communities. Other countries also applied different methodologies to identify missed communities [18–20]. In Mali, electronic community immunization and birth registries were carried out to identify priorities and assess the burden of ZD [18]. Nigeria and Uganda conducted situational analysis and targeted surveys to identify/locate and monitor ZD children [19,20] A unique aspect of our study in Bangladesh, was the use of LQAS to confirm the presence of missed communities in the areas that secondary data analysis identified as home to ZD and UI children. The LQAS results not only confirmed the prevalence of ZD and UI in select communities in Bangladesh, but also showed that these communities exist in all eight divisions of the country and in geographically diverse areas. This supports findings from a geospatial analysis of data from 99 LMICs [7] that found ZD and UI children in remote rural, urban and peri-urban areas as well as in conflict-affected areas.

Our successful use of LQAS for confirming missed communities suggests that LQAS surveys can be used for this purpose, particularly in contexts with relatively high immunization coverage and where un- or under-coverage may be hidden in most data sets. Previously, Sandiford had suggested LQAS as an efficient method to monitor poor performance of immunization programme [21]. A recent study showed LQAS as a useful sampling technique to integrate data from different programmes in a single survey of immunization [22]. Polio Eradication & Endgame Strategic Plan 2013–2018 recommended LQAS sampling method to monitor Polio immunization method as well [23]. A study conducted in Kosovo used LQAS process to evaluate vaccination status among children aged 12–24 months [24]. We also found two studies from India that assessed full immunization coverage using LQAS sampling methodology. To the best of our knowledge, our study is the first to apply LQAS to identify the areas with ZD and UI children and successfully locate the missed communities [25,26]. It is important to note that our results may be subject to selection bias as we purposefully selected upazilas for LQAS to ensure geographic variability, which is critical for the larger IR. In addition, there is a risk of selection bias during the LQAS, although to reduce this bias, we instructed our data collectors to begin data collection from the north-east corner of each selected EPI cluster.

Our experience highlights the potential challenge in reaching ZD and UI children in conflict-affected areas. We were unable to conduct the LQAS survey in Rangamati district due to the presence of various armed groups in the areas and resulting a travel ban enacted by the local administration. While support from key stakeholders, and their understanding of the need to assess and address ZD and UI in this area based on our secondary analysis, may facilitate our access to this district for the larger implementation research, this is not guaranteed.

The results of the determinant analysis suggest that while wealth does not exert much influence over a child being ZD or UI, mother’s educational status, access to media, and practice or ability to make ANC visits were most strongly associated with children’s vaccination status. The greater the status that a mother had in regards to these factors, the less likely was it for her child to be ZD or UI. This is consistent to studies in Eritrea and Nigeria that showed mothers with primary level and above education or mothers who made more ANC visits, were more likely to have children that were fully immunized [27,28]. In addition to factors related to both ZD and UI, mothers who reported that “their last child as unwanted” were more likely to have a ZD child. This is similar to findings of a study [29] from Ethiopia that showed a positive association between mother’s contraceptive use and the children having full immunization coverage. The nature of this relationship needs further exploration to understand its implications for addressing ZD.

We found routine data from the national HMIS to be useful for initial identification of ZD areas because it is one of the few sources that provides district and upazila level estimates and because it is publicly available. That said, the current system has a number of limitations that, if addressed, would make it more useful for identifying ZD areas. In particular, the currently available data only include rural areas which leaves out the 31% of the population that live in CCs and municipalities. Having coverage data for these areas, including at zonal level is critical to understanding the true situation and reaching the ZD and UI children who live there. Moreover, the routine data uses denominator estimates that in some cases result in illogical coverage estimates. Innovative approaches to address this denominator issue, which is a common challenge for using routine health information [30–32], require further exploration but should be a priority for the Directorate General Health Services (DGHS). Any efforts in this regard should account for geographic differences that may relate to denominator estimates (e.g., high levels of migration, quality of data). Improved denominators would assist in identifying ZD children.

Based on our qualitative analysis, key demand -side barriers that contribute towards children being ZD and UI include population migration, irrational fear of side-effects, busy schedules, and misconceptions and hesitancy, though the importance of these varied by area to some extent. The most prominent supply-side obstacle, and one that was mentioned in all areas, is the critical shortage of healthcare workers, especially HAs. This is exacerbated by (and contributes to) the limited opportunities to provide good client communications, EPI centers located too far from communities, and lack of transport options. Identification of ZD children is hindered by inaccurate denominators and population movements. Notably, our qualitative analysis along with evidence from quantitative analysis, enabled us to explore the hindrances and enabled us to few recommendations that may help policymakers and programme managers in identifying the ZD and UI areas and in designing of programmes to reduce the ZD and UI children in Bangladesh.

Based on our qualitative analysis, the key demand -side barriers that contribute towards children being ZD and UI include population migration, irrational fear of side-effects, busy schedules, misconceptions and hesitancy. We found critical shortage of healthcare workers, especially HAs, to be the most prominent supply-side obstacle in all study areas. This factor was exacerbated by (and further contributed to) limited opportunities to provide good client communications. Other prominent factors were EPI centers located quite distant from communities, and lack of transportation. Furthermore, inaccurate denominators and population movements hinder identification of ZD children. The demand and supply side barriers found to be similar for low coverage, drop-out, left out and ZD/UI [12,33–36], need to be addressed by the DGHS on priority basis. Notably, our qualitative analysis along with evidence from quantitative analysis, enabled us to explore the hindrances and enabled us to offer few recommendations that may help policymakers and programme managers in identifying the ZD and UI areas as well as in designing a programmes to reduce the ZD and UI children in Bangladesh.

In conclusion, our study shows that pockets of ZD and UI children do exist in Bangladesh despite the stellar performance of the country’s EPI programme and high levels of full immunization coverage. Children who are ZD or UI need to be reached so that they can also benefit from life-saving vaccines. This paper provides the methods and approaches with which the ZD and UI children as well as the missed communities were identified and explores the demand- and supply-side barriers that lead to children becoming ZD or UI within the country. These methods are both effective and easily replicable identify ZD and UI children. The drivers of ZD and UI along with the barriers to immunization provide potential areas for intervention by policy-makers and can inform interventions to be tested in future implementation research.

Supporting information

Acknowledgments

icddr,b acknowledges, with gratitude, the commitment of Gavi, the Vaccine Alliance to icddr,b’s research efforts. icddr,b is also grateful to Dr. Md. Anwar Hossain Howlader, Secretary, Health Services Division (HSD), MOHFW; Dr. Md. Akhteruzzaman, Additional Secretary (Public Health Wing), HSD, MOHFW; Dr. Md Shibbir Ahmed, Joint Secretary (PH Wing), HSD, MOHFW; M. Jashim Uddin Khan, Deputy Secretary (PH-2), HSD, MOHFW; Dr. Zobayer Ibne Zaid, Technical Support Expert (PH-2), HSD, MOHFW; Dr. Mohammed Nizam Uddin, Director and Line Director (MNC&AH), DGHS; Dr. S. M. Abdullaha—Al–Murad, Program Manager (EPI), DGHS; all other officials of EPI Headquarters, DGHS. We also acknowledge all the data collection teams for their contribution in implementing field activities. Last but not the least, icddr,b expresses gratitude and appreciation to the Government of Bangladesh and Canada for providing core/unrestricted support.