AI and machine learning could transform healthcare delivery, but caution is needed

Healthcare generates huge amounts of data, but interpreting that data can be overwhelming for people to handle on their own. This is where AI and machine learning can deliver value.

During a roundtable discussion sponsored by ECG Management Consultants at Becker's Hospital Review's 12th Annual Meeting, Nick van Terheyden, MD, principal at ECG Management Consultants, shared insights on how cutting-edge technologies can change healthcare systems. 

Five key takeaways were: 

1. Machine learning and AI hold promise, but healthcare organizations must show caution in using them. There are few prospective deep learning studies and randomized trials. Most nonrandomized trials are not prospective, have a high risk of bias and deviate from reporting standards. "In April, the FDA issued a statement that devices can be used as a prioritization and triage tool, but not for diagnosis," Dr. van Terheyden said. "They cannot rule out the presence of disease."

2. Bias is a significant concern with data sets. For example, an AI tool was developed to identify patients with asthma who needed urgent treatment. Based on self-learning using an underlying dataset, AI indicated these individuals would do best in the ICU. "Every clinician in this room would say that's not right. The data was biased because that particular hospital had all patients transferred automatically to the ICU as their standard best of care protocol for asthma patients. The ICU wasn't a predictor of better outcomes, rather it was a correlate of the good clinical protocol in use at that facility," Dr. van Terheyden explained. 

3. Future AI-based technologies will support clinicians with diagnosis and treatment. In radiology, physicians can use AI to automatically measure relevant anatomies and abnormalities. "Unlike humans, these tools create very consistent measurements. The result is reliable data for comparative studies which show whether progress is occurring," Dr. van Terheyden said. Another example of technology in action is computerized identification of brain elements through volumetric analysis. This saves time for clinicians because it produces data that leads to more accurate diagnoses. AI and machine learning can also support targeted cancer therapies. "A fine beam approach can protect organs at risk, but clinicians must first know where organs at risk are located and plot them out," Dr. van Terheyden said. 

4. Analytics are key to uncovering insights that guide clinical practice and deliver better patient care. Analytics require the "five E's:" essentials, enablement, execution, exploration and experience. According to Dr. van Terheyden, "Most organizations already have essential data from their EMR and feeder systems. We build on that through exploration of the data and understanding how the data and systems can inform better clinical decision-making." 

5. Healthcare leaders must become more proficient at understanding and using information. Given today's data-saturated healthcare environment, one attendee asked what skills entry-level managers need to succeed. Because care delivery focuses on knowledge processing, Dr. van Terheyden said statistical data-analysis skills will become more important for healthcare leaders. 

Customized service and care are hallmarks of leading healthcare organizations. "If we can turn data into knowledge and start to deliver meaningful insights, we will be in a better position to change the experience for patients," Dr. van Terheyden said. 


Copyright © 2022 Becker's Healthcare. All Rights Reserved. Privacy Policy. Cookie Policy. Linking and Reprinting Policy.


Featured Learning Opportunities

Featured Whitepapers

Featured Webinars