Exploring interrelationships between structural, social, and biological determinants of vaccine impact in Kenya and Uganda: VAnguard community mixed methods study protocol

Conceptual framework

Our conceptual framework (Figure 1) illustrates the expected relationships between the structural, social and biological determinants that influence vaccine impact. It is further guided by the WHO Behavioral and Social Drivers of Vaccination (BeSD) framework²⁰, which identifies modifiable factors that can improve vaccine uptake. These drivers are grouped into four domains: thinking and feeling about vaccines, social processes that influence vaccination, motivation to seek vaccination and practical issues related to accessing vaccines. Vaccine impact in this study is conceptualized as a combination of vaccine uptake and the level of immune responses elicited by vaccination. To quantify vaccine impact, we will use multiplex serology to measure immune responses alongside vaccine uptake data as proxy indicators. This approach is necessary because direct measurement of disease incidence or prevalence is not feasible within the scope of this study.

Figure 1. Conceptual framework illustrating the interrelationships between factors influencing vaccine impact.

Study aim and objectives

The overall aim of this study is to investigate structural, social and biological determinants that hinder vaccine impact in vulnerable communities in Uganda and Kenya. The specific objectives are to

Patient and Public Involvement

A communication and engagement strategy has been developed to guide in the implementation of the study. Stakeholder mapping and engagement at district and community level will be done in the study communities, starting with visits for meetings with leaders. This will supplement the stakeholder mapping that was done during the preparatory phases of the project. To inform the development of this protocol, a consultative meeting was organised¹⁹, to which representatives of national and community stakeholders were invited, and preliminary data shared.

Inception meetings will be conducted with national and local level stakeholders to ensure that they are informed and included in study plans. These engagement meetings will continue throughout the study period at the local and national levels. We will also utilize the research partnership’s existing membership in relevant working/steering groups to engage and update stakeholders on the study.

Local stakeholders will be involved through local community meetings, which will be held at regular intervals. These meetings will comprise the host communities, the study partners, and local level leadership, and will be conducted to ensure uptake and understanding. We will also use local avenues such as the administrative meetings and any local gatherings to disseminate information about the study. We will share a newsletter regularly to keep stakeholders informed about the study's progress. We will also engage the public through the mainstream and social media channels and the VAnguard website.

Key communication output/activities are peer-reviewed publications, and knowledge translation products such as learning and policy briefs, which will summarize the study progress and findings. We will also participate in various fora such as workshops, conferences, webinars and community events related to vaccines and vulnerable communities.

Study design

This is a cross-sectional mixed-methods study that will use (i) in-depth interviews, (ii) focus group discussions (FGDs) with key community stakeholders, and (iii) a household-based quantitative survey where blood and stool samples of 1032 participants (172 in each community) will be collected. The planned study period is from July 2024 to May 2025.

Study sites

The study will be conducted in six communities, comprising three counties in Kenya and three districts in Uganda. In each country the study will aim to include urban and rural settings and to include settings with varying levels of vulnerability. Vulnerability, in this context is defined as the susceptibility or increased risk of certain communities to experiencing reduced or sub-optimal vaccine outcomes. This could be due to high prevalence of hypothesized structural, social, biological determinants that may impede optimal vaccine impact such as remote location, poverty, lack of education, undernutrition and infections.

To inform the selection of communities for inclusion in our study, we utilized the community vaccine impact vulnerability index (CVIVI), described in detail in our earlier work published Manuscript²¹. The index integrates data for 16 indicators across three domains: structural factors (e.g. distance to the health facility, access to postnatal care), social factors (e.g. household wealth, maternal education) and biological factors (e.g. nutritional status, malaria prevalence), to assess possible community vulnerability to impaired vaccine impact. The percentile rank methodology was used to compute both domain specific and overall vulnerability indices. Subsequently, the domain specific indices were aggregated into an unweighted overall vulnerability index that quantifies the relative degree and magnitude of community vulnerability to suboptimal vaccine impact. Findings were used to select study communities across vulnerability categories to allow us to capture a range of community settings, providing a holistic view of vaccine related vulnerabilities. In addition to the vulnerability indices related to vaccine impact, there were practical considerations for instance feasibility of conducting research in each community, and cost implications associated with survey activities.

The vulnerability index has been developed iteratively, evolving to reflect the availability and scope of data. In Uganda, study districts were selected based on an earlier version of the index than the published one. The earlier version incorporated a limited number of indicators, primarily focusing on malaria, stunting, socio-economic status and minimal structural factors. The initial index classified districts into five categories, ranging from least to most vulnerable. Based on this classification, three districts were selected in Uganda: Kampala (categorised as less vulnerable and urban), Namayingo (more vulnerable and rural) and Kikuube (most vulnerable and rural) as shown in Figure 2. The vulnerability classification of all Ugandan districts based on this initial index is shown in Figure 2. Subsequently, a refined index (Figure 3) was developed, which includes a more comprehensive set of indicators, particularly on structural factors, while excluding vaccine coverage in the computation of the index. A comparison of the two versions showed minimal differences in the overall vulnerability patterns across districts, indicating that the initial index provided a reliable basis for study site selection in Uganda.

Figure 2. Vulnerability classification of Ugandan districts based on the initial index, from least (lightest) to most vulnerable (darkest).

Figure 3. Vulnerable and less vulnerable selected study districts in Uganda as per the refined vulnerability index.

In Kenya, community selection was informed by the current version of the index Manuscript²¹. Using this updated index, the selected countries were Kisumu (considered less vulnerable and urban), and Kwale and Kilifi (categorised as more vulnerable and rural) as depicted in Figure 4. The full classification of the all the counties in Kenya is shown in another related publication²¹.

Figure 4. Vulnerable and less vulnerable selected study counties in Kenya.

Study population

The qualitative component of the study will draw on key vaccine-related stakeholders at the county/district and village levels. This will include individuals at the county/district departments of health overseeing immunization programmes, health facility managers and staff overseeing the implementation of the vaccine programmes and groups of community health volunteers/village health teams who participate in health promotion and mobilizing the community for health activities such as vaccinations. At the community level, the study will draw on community members, community and religious leaders and other stakeholders deemed important in helping us understand more about access and uptake of vaccines at the community level.

The quantitative component of the study will draw participants from randomly selected households in the chosen counties/districts. The following eligibility criteria will be followed for inclusion in the community study:

Exclusion criteria:

Qualitative component of the study

The participants will be drawn from the selected communities and will be purposively selected to maximize the range and diversity of views obtained. The selection will ensure maximum variation on key characteristics relevant to this study, including, sex, age, socio-economic status, education, people/families affected by disability, proximity to health facility, pregnancy status and vaccine uptake. Across the different categories of participants, we will conduct approximately 10–15 interviews and 4–6 focus group discussions with around 50–60 participants per county/district making an overall number of 200–240 participants per country. The proposed sample size for the qualitative component is indicative and may change based on the attainment of saturation as the research progresses²².

Quantitative component of the study

Selecting the households and participants

Using locally available maps and district-level data, we will work with district or county leadership to select a specific area (e.g., village or local council) for the study. The selection will be based on characteristics aligned with our district-level vulnerability index.

Each selected area will be sub divided into enumeration areas by overlaying a grid on the map to create enumeration areas of uniform size and shape. The midpoint of each enumeration area will be identified by its coordinates using a geographic information system (GIS) device. This will be used as the starting point for sampling households and the nearest house will be selected for inclusion. Houses will then be sequentially selected, the next house to be sampled being the nearest to the previous house. If a household is found empty or declines to participate, the next one will be approached until the sample size for the area is achieved.

In the selected households, permission will be sought from the household head or another adult in the household if the household head is absent. Household residents will be enumerated and a single participant in each household will be randomly selected for inclusion in the survey. This approach is designed to recruit a sample of participants with age and sex representative of the community, since these demographic characteristics are expected to contribute to the overall community vulnerability to reduced vaccine impact.

Sample size calculation: We propose to enrol 172 participants per community totalling to 1,032 (of appropriate ages, see above) giving 80% power to detect differences (odds ratios ≥2) in exposure prevalence between vulnerable and less vulnerable communities for exposures with prevalence ≥20% and greater power for more common exposures (Table 1)¹⁹. Assuming that the standard deviation of continuous outcome measures (e.g., antibody responses) will lie between 0.2 and 0.6 log₁₀, with responses in vulnerable settings 0.1 to 0.2 log₁₀ lower than in less vulnerable settings and allowing for ~10% of participants not giving blood samples, 172 participants from each community will give >80% power to compare antibody responses between vulnerable and less vulnerable communities in each setting.

Qualitative data

In-depth interviews and FGDs will be conducted with community stakeholders, including community members, leaders, healthcare workers, and other relevant actors. The purpose is to explore perceptions of vaccine access, uptake, and barriers to vaccination.

Quantitative data

Household level data: A structured researcher-administered questionnaire will be used to collect quantitative data on key social and structural factors from participants in the selected communities¹⁹. Social factors to be assessed include demographic information, vaccination history, household attributes (such as income and household structure), perceptions and altitudes toward vaccines and sources of vaccine-related information. Structural factors influencing access such as mode of transport, place of immunisation, distance to the nearest health facilities, will also be captured. Furthermore, blood and stool samples will be collected to assess proxy markers of vaccine impact, including immune responses.

Health facility-level data: A brief questionnaire will be administered at health facilities providing immunisation services within the selected communities. These facilities will be identified with the assistance of village or district leaders. The questionnaire will capture additional structural factors, including the frequency of the immunisation sessions, vaccine supply availability, staff capacity, and other relevant variables. Additionally, vaccine coverage data from the past six months will be extracted from the immunisation records at these facilities.

Collection of biological samples. The survey will include the collection of samples (blood, and stool)¹⁹. The blood samples will be used tomeasure immune responses to vaccines and exposure to infectionsand stored as plasma, serum and cells, The blood volume to be collected will vary with age. The blood volume will follow guidelines from Harvard Medical School and the Massachusetts General Hospital, which recommend that a maximum of 3ml/kg body weight is taken at a time point (http://www.drgreene.com/21_1616.html). The maximum volume of blood at any given study visit will be 20ml for any one participant. Blood sample collection will be done by well trained staff and all samples will be sealed, labelled and transported in proper sample courier boxes, accompanied by their respective chain of custody forms. Results from the samples identified to have clinical relevance to the participants will be returned and they will be advised and referred appropriately.

Multiplex immunoassays will be conducted using plasma or serum samples to assess infection-specific and vaccine-specific antibody response to the routinely administered Expanded Programme on Immunisations (EPI) vaccines. These vaccines include BCG, polio, hepatitis B, yellow fever, measles, rubella, diphtheria, pertussis, tetanus, HPV, and SARS-CoV-2. The analysis will be conducted on samples collected from participants in both Uganda and Kenya. These immunoassays will measure proxy markers indicative of potential impairments in vaccine immunogenicity and efficacy such as antibodies. Furthermore, the results will contribute to establishing a proxy biological profile of vaccine impact in the study populations. Peripheral blood mononuclear cells (PBMCs) will be stored for a subset of participants and used for further exploration of cellular immune responses such as cytokine responses. Part of the plasma or serum samples will be used for micronutrient assays.

The stool sample collected will be examined for helminth infections such as Schistosoma mansoni, Strongyloides stercoralis and Necator americanus using multiplex real-time PCR. All assays will be conducted in laboratories in Uganda and Kenya where possible.

Data analysis

Qualitative component of the study

All interview and FGD data will be audio recorded, transcribed and translated into English, if conducted in a different language, to facilitate cross-site analysis. Observational data will be documented as field notes and then digitized. The qualitative data will be managed using NVIVO 14 software (QSR International, www.qsrinternational.com/nvivo) and analysed using a framework analysis approach, a type of thematic analysis recommended for multidisciplinary research teams²³.

Codes will be iteratively developed from the data, initially based on our conceptual framework and informed by findings from the case studies and scoping review, using the structural violence framework. This framework will help analyse how institutionalized inequities and power imbalances embedded within social, economic and political systems perpetuate vulnerabilities, limit access to essential health services and undermine vaccine impact²⁴. The analysis will combine inductive and deductive approaches, with codes developed and reviewed by team members.

The initial codes will be organized into categories and further refined into themes, creating a codebook that captures the analytical framework. Data will be charted to identify patterns for each country, followed by a comparative analysis between Uganda and Kenya. While recognising contextual differences, we expect to find both similarity and variations in the identified themes and their specific content.

The findings will include structural, social and biological themes that shape decision-making, and limit access, uptake, and responses to vaccines. These themes will then be further examined to identify and explain relevant intersections using the intersectionality lens²⁵.

All qualitative data will be stored in password protected computers at the KWTRP in Kenya and at the MRC/UVRI & LSHTM Uganda Research Unit head offices in Entebbe, Uganda. Access to the data will be restricted to the study investigators and authorized personnel such as field workers and data entry clerks. To ensure participant confidentiality, all personal identifiers will be removed from the data and replaced with anonymized codes. Following transcription and quality assurance checks, the original audio recordings will be securely deleted.

Quantitative component of the study

The study data collecting tools will be designed and uploaded into a Research Electronic Data Capture system (REDCap), and data collection will be done using encrypted mobile devices. Quantitative information will include that from the main survey questionnaire and biological sample results. Data will be transmitted daily to an online secure server for storage. All data collected will be reviewed for completeness. Consistency checks will be routinely run by the data manager. Descriptive analysis will be conducted by analysing means and percentages for variables such as demographic indicators e.g., sex, age, income, marital status, and educational status, and other variables of interest such as parasite exposure. Regarding proxy measures of vaccine impact, we will first look at antibody responses to individual vaccines and compare them between the communities. Then we shall investigate how antibody responses individually relate to specific exposures. Further exploration will be done on how the antibody responses correlate with each other and whether we can use data reduction approaches such as Principal Component Analysis (PCA) to characterise the overall vaccine responsiveness in different communities. Vaccine coverage will be compared across study groups using binomial regression with an identity link to estimate differences in proportions. To assess vaccine impact, as defined in our conceptual framework, linear regression will be used to compare outcomes between groups of participants.

Clustering by compound will be adjusted for using generalised estimating equations with robust standard errors. In planned subgroup analysis, antibody responses will be disaggregated by age and sex.

Equality, diversity and inclusion

The survey has been designed to ensure representation across diverse geographical locations, including urban, peri-urban, and rural settings, to address regional disparities in vaccine delivery and uptake. Efforts will be made to include participants of different genders, age groups, and individuals with disabilities, recognizing the importance of diverse perspectives in identifying barriers to effective vaccine delivery. This approach aims to achieve equitable representation of experiences and challenges within the study communities. By fostering inclusivity, the study will strengthen the validity of its findings and inform the development of targeted strategies to improve vaccine access and equity in low- and middle-income countries.