Best practices for implementing AI in value-based healthcare

The benefits of artificial intelligence (AI) include predictive analytics, personalized treatment plans, operational efficiency, and remote patient monitoring, says Liz Kwo, MD, MBA, MPH, chief commercial officer at Everly Health, lecturer at Harvard Medical School, and author of the book “ DigitalMD.

This transcript has been lightly edited.

Transcript

How can AI effectively contribute to value-based care by improving patient outcomes and reducing healthcare costs?

AI can significantly improve value-based care by optimizing various aspects of patient management and healthcare. First, predictive analytics. Algorithms can analyze massive amounts of data to predict disease outbreaks, deterioration in patient conditions, and even readmissions. For example, DeepMind worked with the VA to predict acute kidney failure and demonstrated its ability to not only look at outcomes, but also suggest personalized treatment plans, improve patient care, and reduce unnecessary interventions.

There are also personalized treatment plans that AI can customize to help patients based on their patient data, genetic, occupational, lifestyle, and environmental factors. The customization can lead to more effective treatments and improve patient adherence. One example is even AI-driven precision medicine in terms of recommending the most effective treatments for cancer patients, even during drug breaks. So, if at times when a very expensive cancer drug needs to be continued, one could potentially look at when patients can take a drug break, stop the drug, still be fine, and not have to take it again until the cancer returns.

Operational efficiency is also increased. AI can streamline administrative tasks such as scheduling, billing, and patient record management, reducing the burden on providers and lowering operational costs.

Lastly, I would like to mention remote patient monitoring. There are several ways in which AI can track patient health data, which can enable early intervention, reduce hospital admissions, and be integrated into the work of healthcare providers.

What are the main limitations and challenges in integrating AI into clinical settings, particularly with regard to data quality and the need for clearly defined clinical questions?

Integrating AI into clinical settings presents several challenges. The first is data quality. AI systems require high-quality, standardized data to function effectively. Sometimes it is unstructured; it could also be structured, but it must be accurate and not have inconsistent data formats, incomplete records, or data errors – they can affect the accuracy of error predictions. Ensuring data interoperability and standardization across different health information systems is critical.

The second point is clinical relevance. AI algorithms generally need clearly defined clinical questions to produce meaningful results. Without precise goals, AI can produce irrelevant or misleading results or information. I definitely recommend working with clinicians to identify and refine essential questions.

The third aspect is ethical and regulatory concerns. The use of AI in healthcare raises ethical questions, such as patient privacy, data security and the possibility of bias in AI algorithms. Therefore, the regulatory framework needs to be further developed to address these concerns and ensure that AI applications comply with ethical standards and legal requirements.

Finally, acceptance and trust are important. Clinicians are sometimes hesitant to adopt AI because they doubt its reliability, so building trust and demonstrating concrete benefits through rigorous validation studies is crucial.