Clinicians are trained to practice evidence-based medicine, yet the realities of care often reveal its limits. In many situations, the literature does not offer clear guidance for the patient in front of us. A large share of clinical decisions are made without high-quality evidence. This is not a failure of clinical judgment. It reflects a gap in how evidence is generated and delivered.
The traditional path from study design to publication takes years. By the time results are available, patient populations and treatment patterns have already changed. Clinicians are left applying evidence that may not match the clinical context at hand. This is especially true for patients with multiple conditions or those not well represented in clinical trials.
Leveraging real-world data in clinical workflows
The challenge is not only finding existing evidence, but generating it when none exists. Advances in data and analytic methods now allow us to learn from large collections of real-world patient data. When used carefully, these data can help answer questions about patients similar to the one being treated. The goal is not to replace randomized trials, but to extend what they can answer.
Bringing this capability into clinical workflows is a critical step. Embedding tools within the electronic health record (EHR) allows clinicians to ask questions and receive answers without leaving the care setting. This approach follows a familiar model in medicine, where experts help interpret complex information and translate it into clinical decisions, while being clear about limitations.
Closing representation gaps and accelerating research
The need for this shift is greatest for groups that have been underrepresented in research, including older adults, women, and racial and ethnic minorities. Relying only on trial data has left important gaps in how we care for these populations. Generating evidence from routine care offers a way to provide more precise and inclusive guidance, if done with attention to bias and data quality.
A faster cycle of evidence generation may also benefit life sciences. Earlier detection of safety signals and better understanding of which patients benefit most from therapies could improve both care and development. These gains depend on maintaining rigor in study design and clear reporting of results.
Earning trust and the role of artificial intelligence
Trust in this approach must be earned. Methods should be transparent, results should be reproducible, and uncertainty should be clearly communicated. The clinician remains central, both in asking the right question and in applying the answer.
Artificial intelligence (AI) should be viewed as a tool that reduces the effort required to generate and summarize evidence. It does not replace clinical judgment. Closing the evidence gap will require a system that can learn from care as it is delivered and return answers in time to matter. This moves precision medicine from an aspiration to a practical part of clinical care.
Saurabh Gombar is a physician executive.
