Generative AI in Healthcare
Opportunities, Challenges, and Policy
Patients' increasing reliance on the Internet and informal information sources has been a well-documented trend in healthcare. However, the emergence of generative artificial intelligence (AI) has heightened this dependence and rapidly extended it to physicians and other healthcare providers.
In the past, AI models could only analyze and interpret existing data, but now generative AI systems can create new content. This content creation capability, combined with user-friendly interfaces, has led to a surge in adoption and use by many professionals, including healthcare providers. Patients traditionally relied heavily on digital information sources to better understand their conditions. However, with generative AI, healthcare providers can rely on AI-assisted decision-making.
Potential health functions for generative AI
Using generative AI can improve healthcare efficiency by engaging with patients, resolving uncertainties, and summarizing data for providers. It can assist in collecting medical histories, accessing patient records, and verifying medication adherence to assemble a more comprehensive medical history.
Although AI has demonstrated great potential in improving diagnostic procedures, especially for conditions with large data availability, there are still challenges to overcome in achieving accurate diagnoses and mitigating biases. These challenges are mainly present in rare diseases with limited data representation. Therefore, healthcare providers should be cautious when deploying generative AI for diagnostics until the AI has been extensively trained on medical datasets. However, even after comprehensive training, AI should support physicians in the diagnostic process instead of replacing them.
Artificial intelligence (AI) in medical treatments presents significant challenges, particularly regarding accountability and liability concerns, patients' trust and acceptance, and technological and practical limitations. Healthcare providers are responsible for their treatments, and incorporating AI into medical treatment processes may be difficult.
Post-treatment monitoring and follow-up hold considerable promise for AI deployment. AI can leverage data from wearable technology, smart devices, and smartphones to provide real-time tracking and personalized recommendations and interventions. AI can alert medical providers when immediate attention is required to address patient health deterioration proactively.
Implementing AI applications for population health management may appear straightforward, but their effectiveness hinges on the availability of substantial and diverse datasets, including information beyond what is traditionally captured in electronic health records and health information exchanges, such as patients' social determinants, lifestyle choices, and daily activities.
Policy recommendations
Transparency: To maximize AI implementation in healthcare, it is crucial to establish an environment of transparency among AI developers and promote a cooperative relationship between healthcare experts and technology professionals. This partnership is vital to guarantee that AI suggestions are not only medically sound but also thoroughly reviewed for precision, reducing the risk of mistakes.
Informed Consent: It is essential to provide patients with detailed information regarding the role of AI in their healthcare and the potential privacy concerns associated with using AI-powered tools. This education is a legal requirement and a crucial aspect of building trust between patients and the healthcare system as it continues to evolve and integrate new technologies.
Break data monopolies with HIEs: The healthcare market faces a significant challenge of existing monopolies, which could lead to higher healthcare costs and disadvantage smaller, independent providers. To address this issue, industry leaders, regulatory bodies, and healthcare consortia should work together to democratize access to medical data for AI development through Health Information Exchanges (HIEs). HIEs could act as data aggregators from multiple providers, facilitating the deployment of AI systems capable of learning from diverse medical records. Offering AI as a shared service to affiliates could level the playing field and help smaller providers compete with larger ones.