Evidence and AI: People at the Centre
Authors: Prof Girish Thunga1, Prathiksha M Shet1, Srikara K R1
1. Manipal Academy of Higher Education
A few months ago, while reviewing adverse drug reaction reports at our drug safety centre, an AI-assisted signal detection system highlighted a potential safety concern involving a commonly prescribed medication. The signal appeared convincing, supported by patterns identified across multiple reports. At first glance, it seemed to warrant immediate attention. However, as my colleagues and I examined the individual cases more closely, we found that several reports were influenced by confounding factors, including the underlying disease severity and concomitant medications. The algorithm had identified an association, but whether that association represented a true causal relationship remained uncertain.
What struck me most was that the technology had fulfilled exactly the objective for which it was designed: detecting patterns rapidly from large volumes of data. The more difficult task was determining what those patterns actually meant and how they should be communicated to clinicians. After reviewing the evidence and discussing the findings with colleagues, we concluded that further investigation was needed before recommending any clinical action. That experience reminded me that in the age of AI, trust depends not only on finding evidence quickly but also on interpreting it responsibly and communicating uncertainty honestly.
As a clinical pharmacist and drug safety researcher, I rarely encounter situations where evidence alone provides a clear answer. My role often involves balancing published evidence with clinical realities and patient circumstances. When evaluating treatment options or medication-related risks, I begin with the best available evidence, whether from clinical guidelines, research studies, or safety surveillance data. Yet the final decision also depends on factors that cannot always be captured by algorithms: the patient's comorbidities, previous treatment experiences, risk tolerance, personal goals, and preferences.
This act of recognising the context become increasingly relevant as AI tools are being integrated into healthcare. In my own practice, AI-assisted systems have helped identify potential adverse drug reactions, supported medication safety monitoring, and rapidly summarised emerging evidence. These capabilities are valuable because they allow healthcare professionals to process information at a scale that would otherwise be impossible. However, I have learned that generating evidence is only the first step. The real challenge is deciding whether the evidence is reliable, whether it applies to a particular patient, and what uncertainty remains.
One reality I have repeatedly encountered is that patients do not make decisions based solely on evidence. Two patients with similar clinical conditions may choose different treatment options because they value outcomes differently. Some prioritise effectiveness above all else, while others place greater importance on minimising adverse effects, preserving quality of life, or reducing the treatment burden. The conversations held with patients as they make these decisions cannot be replaced by algorithms. They require listening, empathy, and a willingness to explain evidence in ways that patients can understand and use.
My experience in medication safety makes me optimistic about the potential of AI, but also cautious about its limitations. AI can help identify risks earlier, highlight important trends, and improve access to evidence. What it cannot do is determine what matters most to an individual patient or communicate uncertainty in a meaningful way. Those responsibilities remain with healthcare professionals.
As AI is increasingly integrated into healthcare, I believe clinical pharmacists will evolve from being primarily consumers of evidence to becoming evidence interpreters and communicators. Patients and colleagues need more than recommendations: they need context, explanation, and honest discussions about benefits, risks, and uncertainties. In my experience, those conversations are the true instances of trustworthy evidence-based healthcare.
To link to this article - DOI: https://doi.org/10.70253/XOWL4781
Disclaimer
The views expressed in this World EBHC Day Blog, as well as any errors or omissions, are the sole responsibility of the author and do not represent the views of the World EBHC Day Steering Committee, Official Partners or Sponsors; nor does it imply endorsement by the aforementioned parties.
