By Dr. Moira Schieke
As a radiologist, I understand how today’s data deluge is toxic to patient care.
Radiologists are diagnosticians who provide a core value that exploded into use only weeks after the first x-ray was taken by Roentgen over 125 years ago. Radiologists turned a variety of medical images into value through visual insights, diagnostic decisions, and actions for patient care. The volume of medical imaging data exploded 10-fold over the last 20 years. Meanwhile, radiologists have remained strapped to legacy viewers requiring independent review of every image, now at the dangerous pace of every 3-4 seconds. A “win-win” relationship with data flipped into a “win-lose” relationship. Professional quality of work-life has dropped, with record-level burnout rates. Resulting physician shortages were further fueled by the recent narrative that radiologists will be replaced by “AI.” Most concerning, patient care outcomes may be declining; for example, misdiagnosis rates for cancer appear to be increasing.
In response to the data deluge, we have seen a rise in new machine-centric visionary markets. “Artificial intelligence” for medical imaging (coined “AI Automation”) was touted as the solution, received massive investment with over $2 billion poured into start-ups alone, yet now sits in the trough of disillusionment on the Gartner hype cycle. Research now shows that enormous AI failure rates in “open” clinical patient care environments (~93%) and almost all FDA approvals have recently been called into question. AI practitioners took direct aim at disrupting the value proposition of human experts, tried “to develop computer systems that are able to simulate human-like intelligence” with machine autonomy in decision-making and actions, and failed.
Numed, a well established company in business since 1975 provides a wide range of service options including time & material service, PM only contracts, full service contracts, labor only contracts & system relocation. Call 800 96 Numed for more info.
We can choose a smarter path. We can be more strategic, and instead align with a patient-centric visionary market: decision intelligence, an identified top Gartner trend
. Cubismi defines decision intelligence as a formalized approach to augment professional decision-making and actions using established decision pathways and modern tools under expert control. Delineating how it is different from “artificial intelligence” rests in understanding the limitations of “AI'' and the vast and critical abilities of a trained human medical professional (figure 1).
According to Turing award winner Dr. Yann LeCun, AI is a categorizer that does not possess the human abilities that help us navigate the world
: the contextualization of the relationships between things, predictions to understand cause and effect, and planning and decision-making based on external and internal factors. As a trained radiologist, I understand that “frames” for clinical decision-making are formed by many years of education, residency training, and direct clinical experience. These “frames'' allow us to make good decisions from small datasets, and are essential for safe and accurate medical decisions for high quality patient care. They allow professional diagnosticians to turn data into value. The cloud provides the needed cybersecure core technology (figure 2). It supports highly accurate categorizations of health data using modern algorithmic tools leveraging big data. Yet, it also supports advanced human-computer interaction to leverage the essential value of the trained professional for optimal decision-making and actions.
I argue that decision intelligence (DI) under expert control provides a patient-centric strategy that creates a lever for decreasing costs, diminishes short-term and long-term risks, and increases potential benefits for patient care (figure 3). We can predict DI will flatten the costs of machines as more and more data is ingested into the cloud, with lower cost and higher security data storage, and as new combinations of existing data drive new insights. A DI strategy also diminishes the high risks that “AI” won't be able to achieve the level of autonomy that AI practitioners predict. It also diminishes the risks of losing critical human capital trained on generations of validated clinical science as a result of hyped claims. Most importantly, it diminishes the risks of unexpected patient harms. Maintaining the well-established independent value of patient-professional fiduciary relationships and expert communication also implies higher patient care benefits. Some will argue I’m wrong, and proclaim that all “AI” needs is more big data to deliver superhuman patient care benefits. It is thus important to re-emphasize that DI human expert-computer systems will leverage big data, but also small data through expert cognitive “frames”
. Expert control will allow us to turn a wider variety of “open” clinical data into valuable medical insights, high quality decisions, and better patient care actions.
Making predictions about the future is hard, and we face enormous risks if we make the wrong choices. I do not believe we can afford the risk of losing the essential human capital needed to turn exploding healthcare data into value. Professionals must be empowered to direct the best new disruptive business models for healthcare’s technology supply chains that deliver business revenues, yet also drive precision era patient care at lower costs.
Dr. Moira Schieke is the founder and CEO of Cubismi, a medtech company pioneering individualized precision medicine with 5D medical imaging technology.