- KMA DESK
ARTIFICIAL INTELLIGENCE AND HEALTHCARE DELIVERY IN KENYA: CLINICAL APPLICATIONS, CHALLENGES AND THE FUTURE.
The rapid emergence of artificial intelligence (AI) and its incorporation into various sectors and industries has been transformative with notable advances in healthcare making a significant impact. The Kenyan healthcare system faces several challenges due to limited resources and therefore integration of AI goes a long way into enhancing efficiency of healthcare delivery to achieve the Quadruple Aim of Care. Currently, use of AI is already transforming healthcare delivery locally and while its clinical applications are promising, it raises significant concerns related to data protection and potential for bias, understanding the advantages and challenges will be integral in shaping a sustainable future for AI driven solutions.
Clinical Applications of AI in Kenya’s Healthcare system.
In the realm of medical imaging, the use of AI algorithms in diagnostics and imaging analysis is allowing faster and more accurate diagnostics, by AI powered diagnostic tools. For example, the Ministry of Health in partnership with USAID, Centre for Health solutions and Tamatisha TB program launched a Computer Aided Detection (CAD) chest X ray screening and triage tool for pulmonary TB that has significantly improved detection by circumventing the inefficiency of inter and intra reader variability and automating and standardizing interpretation.
This is not only applicable to X-rays, but to CT, MRI and even cardiac imaging ensuring patients, even in rural and under resourced areas, can receive accurate diagnoses in a timely fashion.
AI is also being used to enhance clinical decision making by way of predictive analytics for patient risks. Machine learning models process vast data sets from electronic health records to identify trends, predict patient outcomes, suggest treatment plans and identify high risk patients in settings such as the critical care unit. This data driven approach then allows clinicians to forecast complications, intervene proactively, personalize treatment plans and have more efficient resource allocation.
Additionally, use of AI driven triage systems has brought about the automation of triage and workflow optimization in emergency care and outpatient settings. This serves to streamline patient flow, reduce waiting times and improve the efficiency of care delivery, especially in the emergency setting.
Furthermore, AI is playing a crucial role in telemedicine with AI powered telemedicine platforms enabling virtual consultations. Remote patient monitoring is another viable option where patient access is limited. It is enabled by AI monitoring devices or wearables that can be used to alert clinicians and allow timely interventions, reducing hospital admissions as clinicians can intervene remotely.
Data protection and Bias challenges
However, with access to this frontier emerges the challenge of the ethical and responsible handling of data. AI systems require access to vast amounts of sensitive health information which raises the concern of data security and privacy. Aligning and complying with the tenets of the General Data Protection Regulation (GPDR) and Health Insurance Portability and Accountability Act (HIPAA) globally, and the Data Protection Act of 2019 locally, is a step in the right direction.
Some of the practices outlined in these policies that should be enforced include:
- Transparent collection of all data with informed consent which fosters patient trust and accountability.
- Data anonymization and de-identification before being entered into AI models to protect patient identity by removing any Personally Identifying Information (PII) and application of pseudonyms where necessary. Data is therefore anonymized even when handled by third parties.
- Data minimization, where only necessary data is collected.
- Use of secure data storage and access controls where encrypted storage systems that comply with standards such as ISO 27701 for information security management are employed. This will restrict access to sensitive data highly, requiring multi factor identification and role-based access permissions, which minimizes risk of unauthorized access.
- Having a data governance framework and regular auditing to track data handling, access and security protocols.
The challenge of bias can also not be ignored. AI systems developed in different cultural demographics and healthcare ecosystems may not perform optimally locally where factors like genetics, disease prevalence and even healthcare infrastructure differ. Bias monitoring and inclusive data practices that recognize that AI models can be biased should be carried out, and data sets should be assessed for demographic diversity to ensure that they are representative of patient populations.
What does the future hold?
Looking ahead, Kenya must invest in developing robust regulatory policies with clear guidelines on how AI algorithms should be developed and tested to minimize bias. Challenges such as lack of secure data sharing protocols between health institutions impede the potential of AI in healthcare delivery and must be addressed in our data protection law. Patient privacy should be safe guarded while allowing AI systems to function effectively.
The expansion of telemedicine and mobile health applications also promises to make health care more accessible by overcoming barriers posed by geography and limitations in healthcare infrastructure.
In conclusion, the future of AI in healthcare delivery in Kenya has the potential to be revolutionary. Provided the challenges are addressed conclusively, Kenya can build a healthcare system that utilizes AI to improve health outcomes for her people.
Dr. Cynthia. N. Kamau
Member, KMA Policy, Advocacy and Communications Committee.