Diagnostic Robotics AI Advances Predictive, Personalized Medicine

The number of research papers investigating the potential of AI in healthcare is rising rapidly, almost as fast as the launch of new versions of generative AI and chatbots. The successful deployment of AI in real-life healthcare settings, however, is evolving slowly. The good news is that a number of healthcare providers and focused startups such as Diagnostic Robotics, are making steady progress towards a more predictive, personalized, even personal practice of healthcare.

When I talked to Diagnostics’ founder and CEO Kira Radinsky two years ago, the Israeli startup was focused on helping emergency department physicians quickly diagnose incoming patients and directing them to the right healthcare resource, sometimes even before they arrive at the hospital. The right way to deploy AI in healthcare, said Radinsky at the time, is to start with “something small and tangible.”

ADVERTISEMENT

By now, Diagnostic Robotics has expanded its scope from successfully deploying its AI-powered triage system in different hospital departments such as oncology and surgery, to address “the general field of population health management and the subfield of preventive care,” says Radinsky.

Diagnostic Robotics combines the latest AI tools for the statistical analysis of large-scales data stores and natural language processing or NLP to read and summarize text input. Its analysis is based on 60 billion claims or patient visits. The knowledge gained from this analysis enables it to suggest next clinical steps and predict possible future medical conditions. With its NLP, Diagnostic generates from responses to patient questionnaires automated digital summaries that accompany the patient throughout the hospital and save precious time for the clinicians.

ADVERTISEMENT

In a recent study published in the Journal of American Emergency Medicine, the quality of these automated summaries was compared to manual notes taken for the same 148 patients. The automated notes, while reducing administrative burden, had a similar quality of patient information but more reliably documented billing components of severe conditions.

Moving beyond summaries, “we started actually developing and suggesting the clinical next steps,” says Radinsky. On the basis of its analysis of historical medical data, and assisted by generative AI, Diagnostic Robotics builds a comprehensive picture of a patient’s clinical background and provides recommendations for high-impact, personalized clinical next steps.

“We’ve managed to build personalized trajectories and clinical flows for patients, actually proactively treating them,” says Radinsky. While healthcare plans traditionally focused on the most expensive segment of their member population, those that were using their services more often, Diagnostic Robotics moved to use its AI predictive powers to help them focus on those that may become expensive in the future.

ADVERTISEMENT

A case in point is Blue Cross Blue Shield of Rhode Island (BCBSRI). Analyzing its claims data, electronic medical records, emergency department visits, and other data sources, Diagnostic Robotics identified which members were at risk of future condition exacerbation and could be most significantly impacted by personalized interventions.

The results of a costs analysis of the impact of Diagnostic Robotics’ population health management approach on the health plan’s Chronic Heart Failure (CHF) program were impressive. Members flagged by Diagnostic Robotics generated a 4x cost savings per member compared to those identified solely through the health plan’s transition of care list. Additionally, the analysis demonstrated a substantial $1,600 per member per month reduction in cost of care for engaged members.

Driven by its success in employing AI to improve predictive and personalized medicine, Diagnostic Robotics is now exploring generative AI to help medicine become more personal and physicians more compassionate. “One of our newest pilots for our population health management solution uses generative AI to suggest talking points to help persuade patients to take the clinical next steps that will help them,” says Radinsky.

ADVERTISEMENT

With this pilot, Diagnostic Robotics is joining in the recent explosion of interest in generative AI chatbots and their use in healthcare setting. A study published in April in JAMA Internal Medicine, compared the responses to patient questions by doctors with those given by ChatGPT, and concluded that “The chatbot responses were preferred over physician responses and rated significantly higher for both quality and empathy.”

In “When Doctors Use a Chatbot to Improve Their Bedside Manner,” the New York Times reported that the department of internal medicine at the University of Texas at Austin is experimenting with ChatGPT, using it to write a script—at fifth-grade level, not doctors’ speak—for doctors trying to help patients who are drinking too much alcohol but have not stopped after talking to a therapist.

Diagnostic Robotics goes further, beyond the use of prepared script and generic chatbots. Its generative AI’s persuasion is based on the knowledge it has of the context, medical and otherwise, of the specific patient who needs to be nudged towards a recommended action. In addition, it does it in real-time, providing its conversational suggestions while the physician is on the phone with the patients or engages with them with text messaging.

ADVERTISEMENT

“A lot of what’s happening in the clinical space is building relationships,” says Radinsky and the chatbot is not intended to replace physicians and their important interactions with patients. It’s reminding the physician of “the different things they need to add in the conversation to make it more persuasive in order to push the patient to actually do the clinical next step.”

Diagnostic Robotics demonstrates how to reduce costs and improve health outcomes by using AI to advance preventive and personalized medicine and making physicians more compassionate and persuasive.

ADVERTISEMENT