Ethical AI innovation in healthcare and sustainable development in Bangladesh

Comprehensive Summary

This qualitative, multicenter study asked whether AI innovation and ethical governance influence sustainable healthcare outcomes in Bangladesh, with patient trust and perceived safety as mediators and digital and health literacy as moderators. Using a thematic qualitative analysis framework (rather than a predictive model), researchers conducted semi-structured interviews with n = 5 stakeholders (policy, technical, managerial, and community health experts) from Bangladesh’s AI-enabled healthcare ecosystem. Data was collected from institutional and field-based perspectives and analyzed using Braun and Clarke’s six-phase thematic approach, supported by NVivo 14, with a hybrid deductive–inductive coding strategy. The analysis did not involve model comparison or performance benchmarking. Instead, it evaluated conceptual relationships across domains informed by ethical governance theory, digital transformation theory, and trust-based patient engagement models. The analysis showed that AI innovation improved diagnostic efficiency, personalization, and service delivery, while ethical governance mechanisms, such as fairness audits, transparency protocols, and accountability structures, enhanced inclusivity and user confidence. Patient trust and perceived safety emerged as central mediators translating these technical and ethical features into engagement and adherence. Digital and health literacy acted as key moderators: low literacy constrained accessibility and equitable benefit, whereas targeted literacy interventions improved trust and participation. Secondary analyses included thematic clustering and cross-domain triangulation of stakeholder perspectives. Additional findings highlighted that AI-enabled systems reduced diagnostic time (reported ~30% improvement in some cases), improved patient satisfaction, and supported more efficient resource allocation, though these benefits were uneven due to infrastructural and literacy gaps. Limitations include the small, purposive sample size, potential selection bias, lack of quantitative validation, and absence of longitudinal or patient-level outcome data. External validation was not performed, and subgroup fairness analyses were limited to stakeholder perceptions rather than empirical testing. Findings reflect system-level and behavioral insights rather than direct clinical outcomes and do not establish clinical efficacy.

Outcomes and Implications

This study suggests that ethically governed, human-centered AI systems can support more inclusive and sustainable healthcare delivery by strengthening trust and engagement, particularly in low-resource settings. However, translation to bedside care remains indirect and requires further empirical validation. Clinically, these findings emphasize that successful AI implementation depends not only on technical performance but also on trust-building, transparency, and patient comprehension. In practice, healthcare systems could apply these insights by embedding ethics-by-design principles, incorporating explainable AI features tailored to local languages and literacy levels, and deploying community-based education programs to improve digital and health literacy. Policymakers and healthcare organizations may use this framework to guide regulatory development, workforce training, and patient engagement strategies, ensuring that AI adoption enhances equity and patient-centered care rather than exacerbating existing disparities.