Strengthening evidence for social and behavior change

Background/Objectives: Establishing causal evidence for Social and Behavior Change (SBC) programs remains challenging when randomized trials are infeasible. This study demonstrates how advanced statistical approaches, including machine learning methods, can enhance causal inference in observational SBC evaluations while respecting community realities that preclude experimentation.
Methods: Using data from 3,167 women in northwestern Nigeria exposed to Breakthrough ACTION's SBC interventions (radio, home visits, community events), we applied four analytical approaches: standard regression, inverse probability weighting (IPWRA), entropy balancing, and double machine learning (DML). The DML approach used ensemble learning with cross-fitting to flexibly model treatment and outcome relationships, providing bias-robust estimates. We assessed impacts on five maternal/child health behaviors and robustness to unobserved confounding.
Results: Radio programming showed robust effects on modern contraception (5.7-6.1pp) and antenatal care (9.5-10.5pp). Home visits improved breastfeeding (8.9-9.7pp) and immunizations (6.9-11.8pp). The DML ensemble learner validated findings from traditional methods while revealing heterogeneous treatment effects. Convergence across methods, particularly between DML and doubly-robust estimators, strengthened causal claims.
Implications: This work challenges RCT-centric paradigms by demonstrating how machine learning and advanced observational methods generate credible causal evidence for community-based interventions, democratizing rigorous evaluation in contexts where experimentation is inappropriate or unfeasible.

The Challenge Initiative (TCI) India hub introduced High Impact Interventions (HIIs) and High Impact Practices (HIPs) to strengthen family planning (FP) and maternal-child health services by improving supply, enhancing the basket of choice, and improving quality. A major gap identified was the absence of a structured coaching tool for stakeholders including government officials, community health workers, and providers to ensure efficient implementation of these HIIs & HIIPs. To address this, TCI developed a practical model emphasizing standardization and efficiency to promote participatory, equitable, and community-centred approach to knowledge generation. 
This included user-friendly coaching cards to guide discussions and accelerate government engagement, an e-learning application "Samadhan" for Community Health Workers (CHW) to build their FP knowledge and skills flexibly via mobile phones, and an integrated approach comprising a digital Playbook, support teams (Playteam), and demonstration cities (Playground) to foster experiential learning in health systems. These tools have shown quick impact in improving implementation bottlenecks and health system responsiveness in pilot states and cities. The innovations promote local ownership especially at grassroot level and can be adapted across government levels and other TCI hubs to enhance FP outcomes and health service delivery at scale.

Despite sustained investment in Sexual and Reproductive Health and Rights (SRHR), fragile and conflict-affected countries in East, Central, and Southern Africa continue to report some of the lowest SRHR outcomes globally. While humanitarian actors increasingly recognize the influence of social and gender norms on SRHR behaviours, existing evidence remains fragmented and highly context specific. The lack of a consolidated understanding of how these norms operate in fragile settings has limited the ability to design scalable, contextual and adaptive Social and Behaviour Change (SBC) interventions.
The Women's Integrated Sexual Health (WISH2) project, funded by the UK's FCDO, operates across seven countries, four of which are fragile and conflict affected, delivering integrated SRHR services and embedding SBC approaches to strengthen equity and inclusion. At inception, the project conducted a comprehensive country needs assessment, across each of the seven countries that included an SBC landscape analysis, to identify behavioural, cultural, and systemic barriers to SRHR access. The analysis generated insights into marginalized groups that need to be prioritized and media channels to be utilized in developing, norm-transformative interventions. This underscores the importance of evidence to inform and strengthen SBC programming, contributing learning to the global SBC stage on operationalizing social norm programming in fragile and conflict-affected environments.

Health system level changes are difficult to enact yet have potential for wide and sustained impact. When organizations open their private sector data to public health decision-makers, they unlock an avenue for public stakeholders to allocate resources more effectively, respond to emerging community needs, and strengthen health systems at scale. Understanding how, when, and with whom these data should be shared requires understanding how they are used by government stakeholders. Traditional methods for tracking data utilization, such as dashboard logins or frequency of access, rarely capture whether data is shaping decisions. 
Jacaranda Health is a Kenya-based non-profit focused on improved maternal and newborn health. Jacaranda Health deploys shared dashboards that triangulate quasi real-time data from 3 million + mothers, facilities (including staffing, equipment, commodities data), and Kenya's Health Information System (KHIS), offering stakeholders a comprehensive window into health system performance. To test whether these dashboards are useful, Jacaranda Health and ideas42 employed an SBCC lens to understand pathways of data utilization: starting with systems changes mapping and working backwards to build behaviour pathways that describe what data was used and how decisions were made. This behaviour change research provides insight into the conditions necessary for data to lead to change. 
This presentation will share the methods used to capture and validate these data utilization stories, highlight examples of how maternal health data informed county-level action, and discuss why understanding the behavior of decision makers  is critical for building roadmaps for health systems change.