Artificial intelligence (AI) is revolutionising healthcare globally, and SA is at a pivotal point to harness its potential. With strained resources, staff shortages and unequal access, AI offers transformative solutions to enhance diagnostics, streamline care and support overburdened clinicians.
From virtual assistants to predictive analytics, AI can deliver safer, more efficient and accessible healthcare. Yet its integration demands rigorous ethical, legal and practical guardrails to ensure equity and accountability.
This piece explores how SA can responsibly adopt AI in healthcare, focusing on diagnostics, decision-making and sustainability, while addressing the challenges of trust, regulation and local relevance.
A 2024 study by Beth Israel Deaconess Medical Centre compared ChatGPT-4’s diagnostic performance with that of physicians. ChatGPT achieved a 90% accuracy rate on case reports, outpacing physicians using both traditional methods (74%) and those aided by ChatGPT (76%).
This highlights AI’s potential to enhance clinical decision-making. However, many physicians hesitated to adjust initial diagnoses based on AI insights, revealing trust barriers and cognitive biases. The study underscores that successful AI integration hinges on transparency, clinician trust and human judgment, not just performance.
Sustainability is another critical consideration. The World Economic Forum notes that global healthcare ranks as the fifth-largest emitter of greenhouse gases. Training large AI models such as ChatGPT consumes significant energy, but task-specific, narrow AI models offer a greener alternative, balancing enhanced patient outcomes with reduced environmental impact. For SA, adopting such models could align innovation with sustainability goals.
SA’s healthcare context demands tailored AI solutions. Existing ethical frameworks like the Health Professions Council of SA’s (HPCSA) general ethical guidelines and telehealth guidelines, emphasise patient autonomy, informed consent, confidentiality and practitioner accountability.
These principles apply to AI use, though current guidelines do not explicitly address AI tools in clinical care. Patients must be informed when AI is involved, and clinicians must remain accountable for final decisions. Compliance with the Protection of Personal Information Act (Popia) and the National Health Act (NHA) is also vital to safeguard health data.
However, the HPCSA’s telehealth guidelines, rooted in outdated telemedicine models, assume face-to-face consultations and physical examinations as ethical care standards. They do not account for AI’s growing role in diagnostics, creating regulatory ambiguity. SA law does not recognise AI systems as legal actors, so they cannot serve as “servicing practitioners” alongside clinicians.
This gap risks stifling innovation and limiting AI’s benefits in improving access and accuracy. Amending HPCSA guidelines to address AI-specific concerns such as informed consent, accountability, liability, and transparency is essential to align with national digital health policies promoting technological innovation.
Legal challenges also arise in medical negligence. The current fault-based framework may struggle to address harm from complex AI systems where human error is untraceable, exposing patients and clinicians to uncertainty. Clear regulations are needed to define liability in AI-assisted care.
Contextual relevance is paramount. Many AI health models, trained on data from high-income countries, may lack accuracy in SA’s unique healthcare landscape, marked by distinct health trends and systemic inequities. Local validation or retraining on SA data is critical to avoid bias and ensure clinical value, especially in diagnostics, public health surveillance and telemedicine.
AI-powered chatbots and triage tools, like those in the SA Doctors App, are already improving access in rural and underserved areas by offering real-time triage, symptom checking and appointment scheduling. SA institutions are also developing proprietary algorithms using local clinical trial data, signalling a proactive approach to contextually relevant AI.
To realise AI’s potential, SA must establish robust guardrails:
- Human oversight and accountability: AI must augment, not replace, clinical judgment, with clinicians retaining responsibility for decisions.
- Local validation and transparency: AI systems must be tested on local data and provide clear, actionable outputs for clinicians.
- Data protection compliance: AI applications must adhere to Popia and NHA, ensuring ethical data handling.
- Informed consent: Patients must be informed about AI’s role in their care and consent to data use.
- Equitable access: AI should bridge healthcare gaps, ensuring inclusivity across public-private and urban-rural divides.
AI’s promise in SA healthcare is vast — enhancing public health surveillance, early outbreak detection, personalised treatment and chronic disease management. The Beth Israel study affirms AI’s diagnostic value, but trust and transparency are critical for adoption.
Policymakers must create clear regulatory pathways, healthcare institutions must invest in infrastructure and training, and developers must prioritise ethical, locally relevant design. With these safeguards SA can transform AI from a promise into a practice, building a more equitable, sustainable and resilient healthcare system that serves all its communities.
• Govuza and Moti are directors, Geel an associate, and Ukena a candidate attorney at Werksmans Attorneys.
Would you like to comment on this article?
Sign up (it's quick and free) or sign in now.
Please read our Comment Policy before commenting.