Integration of Artificial Intelligence in Medicine: Addressing Physician Preparedness

As artificial intelligence, exemplified by tools like ChatGPT, becomes increasingly woven into everyday scenarios, medical professionals are witnessing the integration of these technologies into their clinical practice. Clinical Decision Support (CDS) algorithms are emerging as valuable aids for guiding healthcare providers in making crucial decisions regarding the diagnosis and treatment of prevalent medical conditions. These Artificial Intelligence in Medicine have the potential to assist doctors in determining appropriate courses of action, such as prescribing antibiotics or advising on high-risk surgeries.

Nevertheless, the effectiveness of these novel technologies hinges on how physicians comprehend and utilize predictive insights provided by these tools. This challenge stems from the need for a unique skill set that many healthcare professionals presently lack, as highlighted in a perspective article recently published in the New England Journal of Medicine by faculty members from the University of Maryland School of Medicine (UMSOM).

Artificial Intelligence in Medicine; CDS ALGORITHMS

CDS algorithms, designed to forecast outcomes amidst clinical uncertainties, encompass a wide range of tools from basic risk calculators to advanced machine learning and artificial intelligence-based systems. They hold the capacity to predict events such as the progression of life-threatening sepsis triggered by uncontrolled infections or the most optimal therapy to prevent sudden death in patients with heart diseases.

“Though these technologies possess the potential to significantly impact patient care, doctors must first comprehend the operational mechanics of machines and algorithms before integrating them into their medical practice,” emphasized Dr. Daniel Morgan, MD, MS, Professor of Epidemiology & Public Health at UMSOM and a co-author of the perspective.

Despite the existence of some clinical decision support tools within electronic medical record systems, healthcare providers often find the current software cumbersome and challenging to navigate. Dr. Katherine Goodman, JD, PhD, Assistant Professor of Epidemiology & Public Health at UMSOM pointed out, “Doctors need not be experts in math or computer science, but they do need a foundational understanding of algorithms in terms of probability and risk adjustment. However, most have not received training in these essential skills.” Godman is co-author of the perspective.


Addressing this gap requires a re-evaluation of medical education and clinical training, with a specific focus on probabilistic reasoning tailored for CDS algorithms. Drs. Morgan, Goodman, and their co-author Dr. Adam Rodman, MD, MPH, from Beth Israel Deaconess Medical Centre in Boston, proposed the following initiatives:

  • Enhancing Probabilistic Skills: Early in medical school, students should grasp the core concepts of probability and uncertainty, employing visualization techniques to cultivate an intuitive understanding of probability. This training should encompass the interpretation of performance metrics like sensitivity and specificity, enhancing comprehension of test and algorithm efficacy.
  • Integrating Algorithmic Output into Decision Making: Physicians must be educated to critically assess and integrate CDS predictions into their clinical decision-making process. This training necessitates understanding the context within which algorithms operate, recognizing their limitations, and accounting for patient factors that may be overlooked by algorithms.
  • Practicing Interpretation of CDS Predictions in Practical Learning: Medical students and doctors should engage in application-based learning by employing algorithms in individual patient cases, observing the influence of diverse inputs on predictions. Additionally, they should learn to communicate CDS-guided decision-making with patients.


Please enter your comment!
Please enter your name here