Data collection and analysis.

A standardized stakeholder mapping exercise will be implemented in all DeWorm3 sites. In the first part of the exercise, working groups composed of the DeWorm3 IS team, personnel from Ministries of Health and Education, and local community-based organizations (CBOs) will identify and describe an exhaustive network of stakeholders who influence or are affected by school-based MDA and, separately, community-wide MDA for STH (adapted from [13]). Working groups will identify stakeholders at each level of the health system, including: the national level; state/regional (or equivalent) level; district (or equivalent) level; local/health center level; community level, including community drug distributors (CDDs), community leaders, and community members; and groups external to the government, including CBOs, non-governmental organizations (NGOs), or international bodies such as WHO.

The stakeholder mapping tool guides the systematic identification of a stakeholder’s influence on the outcomes of the intervention, their attitude towards the intervention, and opportunities to engage the stakeholder throughout the duration of DeWorm3 trial. Additional stakeholders might be added to the network at two follow-up time points, and their attitudes towards the intervention will be updated over time to document changing acceptability of the intervention. During the second part of the exercise, DeWorm3 staff will map the relational power dynamics and dependences between the stakeholders identified by characterizing the directionality of influence (i.e. who influences who) between the stakeholders and the influence category, including: supervision, financial support, technical support, or communication only.

De-identified stakeholder maps in which stakeholder names and affiliations have been removed will be analyzed using standard network measures to understand if: there are stakeholders with extensive influence (i.e. degree centrality), there is high interconnectedness across stakeholders, there are stakeholders who link the network together (i.e. between centrality), there is density around specific stakeholders that might hinder sustainability, there is integration across government ministries and external organizations, or if there is the potential for dysfunctional relationships that might influence intervention effectiveness [12]. Over the duration of the DeWorm3 trials, stakeholder maps will also be used to explore the association between stakeholder attitudes and effective implementation in relevant geographic areas in order to understand if high intervention acceptance of influential stakeholders has a cascade effect at other levels, thereby positively influencing treatment coverage.

Data collection and analysis.

Focus group discussions (FGDs) and individual interviews will be conducted with stakeholders from all levels of the health system. FGDs with cluster-level stakeholders will take place in four randomly selected clusters per site, including community members, CDDs, and health center staff or CDD supervisors. FGD participants will be selected using random purposive sampling [19]. Interviews will take place with individuals at higher levels of the health system. These individuals will be selected using purposive quota sampling of mutually exclusive key informant groups (i.e. stakeholder levels) [20].

Trained interviewers who are fluent in local language and customs will conduct one-on-one interviews and small FGDs in private locations with consenting individuals. The interviews will be semi-structured key informant interviews with a mix of respondent and informant style questions. In each site, we use pragmatic criterion to determine the sample size, influenced by the number of levels of the site’s health system. If it is deemed that data saturation has been achieved and redundancy criterion are met within a given stakeholder group after 80% of interviews have been conducted in that group, the sample size may be reduced.

All interviews will be audio recorded and transcribed. Transcriptions will be translated to English. Two primary coders will independently code the site-specific transcripts using CFIR sub-constructs as analysis guides and coordinate their coding to create a comprehensive codebook and thematic analysis. Analyses will be based upon a deductive approach[21]. Special attention will be paid to distinguish factors related to the content and structure of the intervention, opportunities to enhance fidelity of the intervention, penetration within the health system, and opportunities to bring the intervention to scale. Similarities and differences in identified factors will be assessed within and across sites, leveraging the inherent heterogeneity present across clusters.

At midline and endline, qualitative research at the cluster level will take place in four randomly selected clusters per site: two high coverage clusters (80–100% average treatment coverage in year 1) and two low/moderate coverage clusters (<80% average treatment coverage in year 1). CFIR constructs will provide a framework for understanding why the DeWorm3 intervention may exhibit differential effectiveness across settings, correlates of high coverage, and opportunities to optimize the intervention moving forward. Following thematic coding, two independent analysts will assign scores to data at a cluster level to distinguish CFIR constructs that are missing too much data (M), do not distinguish between coverage levels (0), weakly (+1/-1), or strongly (+2/-2) distinguish coverage levels. The ratings reflect the positive or negative influence (valence) and strength of each construct on influencing implementation [18]. These relationships will be evaluated using chi-square tests. Findings will be synthesized to develop recommendations for best practices and highlight barriers and facilitators of implementation in high and low coverage sites.

Quality control will be ensured through a process of (1) random one minute spot checks of each transcribed audio file at a site level, (2) coding review and confirmation of 10% of coded transcripts from each site at the central DeWorm3 level, and (3) coding review and confirmation of 5% of coded transcripts between sites (i.e. circular quality assurance).

Data collection and analysis.

The readiness survey instrument aims to capture an individual’s perception of the structural readiness of the health system as well as the downstream phenomena of psychological readiness. Readiness indicators fall into six domains, including: the policy environment, leadership structure, financial resources, material resources, technical capacity, and community delivery infrastructure. Survey questions within each domain are intended to assess the health system’s readiness to launch the new community-wide MDA intervention, deliver community-based public health programs, and function collaboratively. Knowledge assessment questions are used to score a survey participants knowledge of a specific domain, and weight their responses to domain questions accordingly.

The sampling frame for survey participants is defined by the network of individuals identified during baseline stakeholder mapping and updated over time. Participants will be selected using restricted random selection to ensure participation from at least 10% of individuals in each stakeholder group (i.e. national level government staff, CDDs, external partners, etc.), attain sufficient representation from each stakeholder group, and variation across stakeholder groups. Random selection of participants at the local level (health center staff and CDDs) will be restricted to five randomly selected clusters. Surveys will be either self-administered by literate participants or administered in-person by trained study personnel in the local language.

Mean readiness scores and standard deviations will be calculated by site and stakeholder group. We will assess concordance between indicator scores within sites using analysis of variance (ANOVA) models to understand variation in scores amongst and between stakeholder groups.

Data collection and analysis.

At baseline, implementation science teams in six different clusters per site will engage in guided in-depth process mapping activities. Clusters will be selected based on site-specific indications of capacity. In some sites, this may be historical targeted STH coverage, LF coverage, or immunization coverage. We will use restricted randomization to select two DeWorm3 intervention clusters with high baseline capacity, two intervention clusters with low baseline capacity, one control cluster with high baseline capacity, and one control cluster with low baseline capacity. By including clusters with heterogeneous baseline capacity indications, we augment our ability to observe how the delivery environment responds to or evolves in the context of the DeWorm3 intervention. IS teams in each selected cluster will identify all of the activities and the flow of activities necessary for delivering school-based or community-wide MDA. Ideal metrics and timelines will be identified for each activity. For example, for the activity of sensitization of village leaders, the ideal metric may be that 80% of village headmen with phones are called about an upcoming MDA campaign and the ideal timeline may be one week in advance of drug delivery.

Activity tracking will be ongoing. Once per year, following MDA in both trial arms, selected clusters will record how each mapped activity actually occurred, including the metrics achieved and the relevant time points. This will facilitate analysis of the number of activities that did not meet ideal benchmarks, and if these activities act as bottlenecks to implementing subsequent “downstream” activities. T-tests will be performed to identify if there are changes in the number of activities that do not meet targets within clusters across study years, if certain activity targets are consistently associated with high or low coverage, as well as if there are differences in the number of bottlenecks identified in clusters achieving high or low coverage.

A routine process mapping worksheet will be conducted in all intervention and control clusters during each round of MDA. These worksheets detail activities that must take place in the effective delivery of the intervention. Routine mapping provides opportunities to optimize the intervention in real-time, to systematically measure study inputs and processes, to promote the replicability of the trial, and to identify trial components that may be easily modified or adapted with minimal effects on MDA coverage. Observations from the in-depth process mapping activities will also be used to refine the routine workflow trackers in order to build a useful monitoring tool that may be applicable in other settings.

Data collection and analysis.

Inputs required to assess the financial and economic costs of community-wide and targeted MDA for STH will be collected in all DeWorm3 study sites using a standardized cost collection tool. Aligned with economic evaluation guidelines [25–27], the tool follows an ingredients-based micro-costing approach for estimating cost of health interventions; this methodology entails identification and measurement of all inputs required for the intervention, followed by their conversion into monetary value terms to produce a cost estimate. The data will be collected throughout the implementation of the DeWorm3 trial. Data collection rounds will be triggered by site completion of specific trial activities (ex. completing a cross-sectional survey). Aligning data collection with trial activities and routine financial reporting will enhance the quality of the cost data collected by minimizing recall bias and limiting duplication of tasks within the project.

The costs will be assessed from multiple perspectives, namely: individual (i.e. participant), societal, and health systems (i.e. government). Full incremental costs will be derived. Resource use will be valued both in terms of financial and economic costs, including time demanded by community members and Ministry of Health staff to participate in different aspects of the intervention. The intervention costs will be assessed by activity (i.e. community sensitization, population census, MDA, etc.) allowing for greater transparency in reporting and broader generalizability of estimates generated. Within each activity, the contribution of specific, broadly categorized, programmatic inputs will be distinguished (i.e. drugs, wages and per-diems, materials and supplies, equipment and overheads, etc.) This will highlight key cost drivers for MDA interventions. The cost estimates will be summarized annually for each trial arm with total intervention costs calculated by summing annual costs over the duration of the DeWorm3 trial.

The relative efficiency of community-wide and standard of care targeted MDA for STH as implemented in the DeWorm3 trials will be assessed by combining the cost estimates with measures of epidemiological impact of the interventions. A range of outcomes including the percent reduction from baseline to endline prevalence and the total number of prevalent infection case-years averted will be considered to both fully capture the impact of the interventions and allow for comparisons with other studies.

The ultimate goal of DeWorm3 cost-effectiveness research is to provide policy makers and implementers with information regarding the efficiency of community-wide MDA for STH elimination in endemic settings. To this end, the trial will generate key epidemiological and cost data to populate mathematical models of STH epidemiology and control that can also be used to assess the impact of elimination strategies [28, 29]. The models will produce estimates of the incremental impact of community wide MDA for STH for settings outside of the DeWorm3 trial and longer time horizons or 5, 10, 15, 25, and 50 years to appropriately capture STH elimination scenarios. Long-term costs will be discounted and sources of uncertainty will be explored in univariate and probabilistic sensitivity analyses [30, 31].