AI and Data in South Africa's Health Sector

Advances in data-driven technologies such as artificial intelligence (AI) are transforming the health sector at an unprecedented rate. AI is enabling significant progress in healthcare, public health research, and drug development. In addition, AI has the potential to address issues around the broader social determinants of health by increasing access to health services, wellness and lifestyle management, and enabling efficient health systems management. However, these advances also raise social, legal and ethical questions around the protection of personal information, equitable access to health care and patient safety, among others. It is therefore critical that South Africa develops and implements appropriate regulatory frameworks around the responsible use and governance of data and AI within the health sector. There is a substantial body of research and several global, multilateral and national policy frameworks that engage with critical policy issues in this field. In South Africa, the nature of these issues is complicated by high levels of poverty, a large disease burden and highly unequal resourcing and access to health services. This Topical Guide reviews the current policy and research environment, with a view to adopting a more inclusive, human rights-based approach to the use of AI and data for improving health outcomes. Specific recommendations include the development of national policies and strengthening key public interest institutions to: (1) explicitly recognise and protect the human rights of patients and practitioners, (2) address issues related to data quality and bias, (3) clarify mechanisms for ensuring safety and accountability in the context of automated decision-making, (4) explore ex-ante and ex-post approaches for strengthening transparency, (5) understand the impact of data-driven technologies on trust and patient-centred care, (6) recognise concerns about automation and the need for skills development amongst healthcare workers, (7) seek ways for aligning the governance of data whilst recognising potential risks of broad data integration, and (8) support existing work on language and translation for more user-centred AI systems.