AI-enabled physicians: rethinking care delivery

Physicians have always looked at artificial intelligence with some amount of skepticism. Most believe that AI and machine learning technologies are overhyped, and cannot solve clinical problems in real life.

Physicians tend to not like machines dictating their decisions. They prefer to rely on their clinical acumen and judgment to diagnose and make clinical decisions.

But in today’s changing care delivery landscape and consumers demanding better engagement and care experience, physicians are rethinking how they can improve care delivery.

Empowered decision-making, improved health outcomes 

The question has never been about AI in healthcare versus the clinical decision-making of the physician. As Atul Gawande says in his best-selling book, Complications, “No matter what measures are taken, doctors will sometimes falter, and it isn’t reasonable to ask that we achieve perfection. What is reasonable is to ask that we never cease to aim for it.”

Gawande, in this book, provides real-life anecdotes of the errors that surgeons and physicians have made.

When supported by a layer of AI and ML-enabled assistant that sifts through historical data and draws parallels and relevant insights about a case, the process of decision making can go a long way toward expediting the diagnosis and decision-making process.

Consider the case of sepsis diagnosis. AI algorithms are widely used in critical care units to diagnose sepsis. A sepsis sniffer algorithm alerts the physician at least 3 to 4 hours in advance of an escalating event leading to severe sepsis. This can result in reduced mortality rates.

The early indicators are given by an algorithm that works in the background collecting all the data generated from the bedside, the patient’s labs, and produces intermittent results to alert the physician of an impending crisis.

Hospitals have seen an average 39.5% reduction in in-hospital mortality, a 32.3% reduction in hospital length of stay, and a 22.7% reduction in 30-day readmission rate for sepsis-related patient stays.

AI is one of the levers that could be used as a second opinion to clinch a diagnosis in complicated cases and aim towards perfection.

Patient engagement and care experience 

In today’s scenario of virtual care integrating with in-person care, the AI-enabled physician can delegate routine and mundane tasks like sending educational materials, ordering prescription refills, and responding to patient queries with the support of AI algorithms.

In larger facilities, by using AI-powered tools like symptom checkers to triage patients, a physician can further optimize his clinic or department’s function. The use of AI-powered chatbots to answer routine questions, book appointments are other uses of AI which help to improve the patient experience.

AI algorithms are helping to identify patients with chronic disease, send them medication reminders, educational materials, and alert physicians of any changes in their vitals or labs when using connected devices. Overall, leading to patients being more engaged and accountable for their wellness.

Selecting use cases where AI can be successfully implemented

It is important to select use cases where AI algorithms can make a measurable impact in clinical areas. Some of the areas where AI has been implemented successfully in healthcare are radiology, internal medicine, neurology, and cardiology.

In all these areas, the algorithms work quietly in the background and help physicians make a difference, sometimes by providing a second opinion or simply alerting for any impending crisis. Nowhere has AI overshadowed physician presence.

Patients always prefer to hear their diagnosis from their physicians. In imaging, today, AI models are helping to automate the contouring of healthy tissue and organs from tumors, develop adaptive dosage and treatment plans for radiation therapy, diagnose cancers in early stages, diagnose large vessel blockages in stroke, and identify disease patterns for images. This is subsequently reviewed by the physician and the radiologist, who is aware of a patient’s overall clinical, social and psychological picture.

Machine learning has an algorithmic bias and will always be appended with a tagline or disclaimer: “clinical correlation necessary.” However, AI superseded by clinical intervention from the specialist who is aware of the human aspects of the patient is a good solution for an integrated machine and human model of care.

Many other use cases have been implemented or are under development to help diagnosis at the bedside. In recent times, technologies such as natural language programming to read unstructured information in physician notes and voice-enabled assistants to predict emotional and behavioral traits are undergoing research.

Artificial intelligence has made noticeable inroads in healthcare’s administrative and operational areas and is making a measurable mark in increasing the revenues of large health systems.

But AI has also had a string of failures in clinical areas, leading to a lack of real-world deployments of machine learning algorithms in mainstream clinical practice. The IBM Watson failure in cancer diagnosis and treatment, and Google’s failure to detect diabetic retinopathy with deep learning models from images of patients’ eyes are recent examples.

The potential of AI in healthcare has not been realized to date. There are a limited number of reports available on the clinical and cost benefits arising from the real-world use of AI algorithms in clinical practices.

Though slow, AI in clinical areas is steadily picking up but needs to deliver on its promise of making a difference at the point of care.

As health systems and hospitals transform digitally to improve care delivery and patient experience, physicians cannot be left behind. They must change too and contribute to making this transformation into a more positive experience for themselves and their patients.