Artificial intelligence (AI) is set to revolutionize the handling of clinical data by all parties involved, from physicians to patients, according to a presentation delivered during the 4th Annual Heart in Diabetes Conference, held virtually August 21 to 24, 2020.1

The handling of healthcare data is complicated by several factors including its location in multiple source systems (eg, electronic medical records and human resources software) and departments (eg, pharmacy or radiology), the heterogeneity in formats (eg, numeric, paper, digital), discrepancies in data definitions (eg, due to emerging research), the complexity of the data, and evolving reporting and regulatory requirements (eg, for readmission).

In addition, with the results of 75 clinical trials and 11 systematic reviews on average published every day in 2010, clinicians may experience “cognitive overload.”2


Continue Reading

AI may be used in health care to classify patients (eg, risk categorization, diagnosis), to predict outcomes and select appropriate interventions, and to facilitate interaction with and management of patients (eg, with the use of natural language processing, and for the selection of pertinent health information).

The technology can also help alleviate the burden related to electronic health records, for example with the use of voice recognition or natural language processing that provide more seamless interfaces, or with virtual AI assistants to handle tasks ranging from scheduling to preoperative validation, medication refills, and cognitive assessment.

More specifically, in cardiology, AI can be leveraged to identify potential cardiac events (eg, myocardial infarction, exacerbation of heart failure) earlier than a clinician could, using predictive analytics. The technology may also be used for precision phenotyping, the improvement of patient care and disease management (through the selection of focused interventions).

Medical-grade and US Food and Drug Administration-certified wearables (in the form of eg, watches, wristbands, sensors, patches) are used to track an array of parameters ranging from vital signs (eg, heart rate, blood pressure, and temperature), to sleep cycles, rhythms, and stress levels. Wearables can also help track diseases that include heart failure and diabetes.

For the management of diabetes, AI permits automated retinal screening (eg, for the detection of maculopathy or retinopathy), and the identification of subgroups of patients at higher risk for hospitalization or complications (eg, nephropathy and neuropathy) through predictive analytics. The technology also provides an array of self-management tools (eg, glucose sensing, activity and diet tracking) and supports clinical decision (eg, medication customization and adherence).

AI has also shown high sensitivity and specificity in grading diabetic retinopathy based on photographs of the retina, and in achieving a better control of blood sugar. The use of AI in clinical decision support is permitted by the technology’s ability to parse and analyze healthcare information and data and suggest “next steps” (eg, testing, treatments), as well as alert clinicians of potential issues (eg, drug-drug interactions), thus allowing to improve patient outcomes and physician efficiency, and reduce the incidence of medical errors.

For the management of diabetes which is complicated by factors that include incremental dosing and the use of sequential agents, AI allows to personalize treatment based on the patient’s body mass index, insulin resistance, and pancreatic β-cell function.

The Comprehensive Type 2 Diabetes Management Algorithm endorsed by the American Association of Clinical Endocrinologists and the American College of Endocrinology, “provide[s] clinicians with a practical guide that considers the whole patient, his or her spectrum of risks and complications, and evidence-based approaches to treatment.”3

“The promise of AI remains intriguing,” concluded the presenter, Irving Kent Loh, MD FACC FAHA FCCP FACP, chief medical officer and cofounder of Infermedica and adjunct lecturer in the Department of Medicine at Stanford University, California. Several improvements are still needed to “control costs, increase convenience, improve quality, and improve outcomes,” Dr Loh added. In addition, there is a “need to resolve data privacy, interoperability and regulatory uncertainty and to resolve concerns of mismanaged care due to suboptimal training data sets, inadequate human oversight, or conversely medico-legal implications of not using AI if it is validated to be better.”

References

  1. Loh IK. Artificial intelligence in the future management of cardiovascular disease and diabetes. Presented at: Heart in Diabetes 4th Annual Conference Live Virtual.  August 21-24, 2020.
  2. Bastian H, Glasziou P, Chalmers I. Seventy-five trials and eleven systematic reviews a day: how will we ever keep up?. PLoS Med. 2010;7(9):e1000326.
  3. Garber AJ, Handelsman Y, Grunberger G, et al. Consensus statement by the American Association of Clinical Endocrinologists and American College of Endocrinology on the Comprehensive Type 2 Diabetes Management Algorithm – 2020 Executive Summary. Endocr Pract. 2020;26(1):107-139.