Evidence-based medicine (EBM) is a system with interconnected components. It begins with evidence production, the process of designing and carrying out clinical studies. Evidence synthesis gathers, evaluates, and combines data from these studies to answer specific clinical questions. These syntheses inform the creation of Clinical Practice Guidelines, which clinicians and patients use together to decide on the best treatment options. This application of evidence is known as evidence implementation. Finally, evidence evaluation monitors the quality of care, including how well healthcare practices align with evidence-based recommendations.  

The Evolving Challenges of Evidence-Based Medicine 

Evolution of  EBM is challenging. Despite widespread adoption, inconsistent research methods often yield poor-quality results. EBMs come across criticisms at times. Some argue that it emphasizes on randomized controlled trials (RCTs), with their often-strict inclusion criteria, can lead to biased results and limit generalizability. Furthermore, clinicians perceive EBM as undermining their professional autonomy and patient relationships. The reliance on numerous, expensive RCTs, which often addresses narrowly defined questions, is another concern. Finally, the time lag between research publication and clinical implementation as seen as rendering EBM outdated for timely decision-making.   EBM effectiveness is hampered by research gaps (unanswered clinical questions alongside wasteful studies), slow, costly evidence synthesis (leading to outdated information), difficulty applying population-level findings to individual patients, and data silos that hinder evaluation and knowledge sharing.  AI can be used in several stages of the EBM ecosystem to address potential solutions to these forthcoming solutions. 

How AI is Changing the Game? 

AI offers several distinct advantages. First, it can identify research gaps, preventing redundant studies and wasted resources. Second, AI can significantly accelerate the currently slow and expensive process of evidence synthesis, ensuring that clinical decisions are based on more timely and complete information. Thirdly, AI can be used to provide integrated treatment for patients with multiple comorbidities, as well as to engage patients and elicit values and preferences (e.g, chatbot-based decision aids). Lastly, it can support the new emerging type of research, which is living systematic review. 

• It can identify research gaps and assist in more efficient use of research funding. 

• Evidence synthesis is accelerated with the use of machine learning. 

• LSRs, or living systematic reviews, can be made possible via automation. 

EBM’s original vision involved clinicians expertly integrating the best available evidence with their own clinical judgement and patients’ values. However, implementation has proven difficult, requiring the balancing of objective, population-level data with individual patient needs. AI offers a transformative potential to address these challenges.  

Living Systematic Reviews 

Research often relies on rigorous, but costly and time-consuming, evidence synthesis methods. A single systematic literature review can be very costly and may consume much time. This slow pace means reviews are often outdated by time they are published, as new research emerges daily. Machine learning is increasingly being used to accelerate evidence synthesis, enabling the creation of Living Systematic Reviews (LSRs) that are continuously updated with new evidence. 

Shared Decision Making 

AI can also improve patient involvement in care decisions. New AI methods can enhance user experience and resolve treatment conflicts arising from multiple health conditions. Successful development in this area could lead to a new generation of CDSS (Clinical Decision Support Systems) that truly realize the potential of personalized medicine.  

The AI Dilemma 

AI has the potential to revolutionize EBM and science, but its misuse can also create problems. For Example, AI driven decision support systems that ignore patient values could lead to a resurgence of paternalistic medicine, but with a computer making the decisions. Likewise, evidence sysnthesis should remain human-led, as human judgement and contextual understanding are crucial for evaluating study quality and the relevance of evidence. 

Conclusion 

The growing volume of scientific evidence, particularly with the recent pandemic, makes AI’s potential benefits and rapid development increasingly relevant. AI offers new opportunities and is becoming essential for managing the exponential growth of research, especially the surge in studies related to global events such as the pandemic.

Read Whitepaper AI: A Smart Prescription for Evidence-Based Medicine