• Politics
  • Diversity, equity and inclusion
  • Financial Decision Making
  • Telehealth
  • Patient Experience
  • Leadership
  • Point of Care Tools
  • Product Solutions
  • Management
  • Technology
  • Healthcare Transformation
  • Data + Technology
  • Safer Hospitals
  • Business
  • Providers in Practice
  • Mergers and Acquisitions
  • AI & Data Analytics
  • Cybersecurity
  • Interoperability & EHRs
  • Medical Devices
  • Pop Health Tech
  • Precision Medicine
  • Virtual Care
  • Health equity

Targeting Diabetes with Big Data, Machine Learning, Real-Time Informatics

Article

How cutting-edge research into tech and diabetes rocked ADA 2018.

diabetes ai,machine learning diabetes,ada 2018,hca news,healthcare analytics news

The odds of responding well to “intensifying” antidiabetic regimens with an additional antihyperglycemic and of avoiding episodes of severe hypoglycemia could be increased by promising approaches in big data, machine learning, and real-time informatics, according to recent presentations at the American Diabetes Association (ADA) 78th Scientific Sessions, Orlando, Florida.

>> LISTEN: Is AI Real?

The decision to add a glucagon-like peptide-1 receptor agonist (GLP-1 RA) to basal insulin and other oral antihyperglycemic agents that have failed to adequately control a patient’s type 2 diabetes (T2DM) could be better informed, for example, with analysis of a range of patient characteristics including the other medications and dosages, and the severity and duration of diabetic symptoms and of concurrent conditions.

Big-data algorithms might be used to consider these multiple parameters, and to possibly identify optimal patient characteristics for the new drug therapy, according to Esther Zimmermann, PhD, Novo Nordisk, Søborg, Denmark.

“Machine learning is a new tool used for the analysis of big data that has the potential to identify trends and predict outcomes,” Zimmermann explained, in describing her study.

“The aim of this study was to use machine learning for extensive analysis of big, complex to data to, one, characterize patients on basal insulin for whom a GLP-1 RA was additionally prescribed and, two, identify predictors of 1 percent (or greater) reduction in A1c in (those) patients.”

Zimmermann and colleagues identified 23 variables with logistic regression analysis and then applied hypothesis-free machine-learning models to a database of 80,019 patients in the US, on the IBM Explorys system, to examine possible influence of 155,000 additional parameters. The course of diabetes illness in a matched cohort of 62,291 patients with T2DM who did not add the GLP-1 RA was compared with that of 17,728 who received the agent.

“Patients with T2DM on basal insulin who added a GLP-1RA were likely to be less than 75 years old and had characteristics of progressed disease,” Zimmermann reported.

Although variable selection via machine learning confirmed the importance of a number of variables, Zimmermann reported that baseline A1c was the only strong predictor of a decline in A1c less than or equal to 1 percent. That suggests, “patients on basal insulin with high A1c would benefit from combination treatment with GLP-1 RA,” she said.

Multiple Prediction Models Offset Missing Patient Data

It’s one thing to construct a machine-learning model to predict best outcomes from a range of patient characteristics, but another when patients in the “real world” confound this approach by not having all data relevant to the model, pointed out Lisa Chow, MD, associate professor of medicine, Division of Diabetes, Endocrinology and Metabolism, University of Minnesota, Minneapolis, Minnesota

Chow and colleagues addressed the challenge of applying a model to predict risk for severe hypoglycemia in patients with T2DM to patients who don’t have all relevant data, by developing multiple, overlapping models and a method to collectively analyze them to offset gaps in the patient data.

“While risk models for severe hypoglycemia may categorize an individual’s risk, incomplete or unavailable, ‘missing’ data limit clinical application,” Chow explained. “To circumvent the ‘missing’ data issue, we used machine learning to identify multiple predictively equivalent risk models for severe hypoglycemia in patients with T2DM.”

>> READ: IBM Watson Health Releases Diabetes App, Touts Power of Data

Chow and colleagues examined 95 candidate risk factors for model construction, from data in the Action to Control Cardiovascular Risk in Diabetes (ACCORD) study, involving more than 10,000 patients with T2DM. The investigators generated 194 predictively equivalent risk models using the SurvTIE* algorithm, which is an adaptation of TIE* algorithm of survival outcome.

Each risk model contained between 18 to 23 risk factors, enabling some to apply to a particular patient, when others would require data that were not available. With all having c-index ratings of 0.79 for predictivity, a model that doesn’t factor in a patient’s glomerular filtration rate, for example, could be used if serum creatinine values were available instead.

“In patients with T2DM, multiple predictively equivalent risk models for severe hypoglycemia can be identified, potentially personalizing model selection when data are missing,” Chow indicated.

Real-time Informatics Prevents Hypoglycemia in Hospitalized Patients

A successful real-world application of real-time informatics was described by Gary Tobin, MD, director, Diabetes Center, Center for Advanced Medicine and Professor of Medicine, Division of Endocrinology, Metabolism and Lipid Research, School of Medicine, Washington University in St. Louis, Missouri.

Tobin and colleagues had previously reported on preventing inpatient hypoglycemia with real-time informatics alerts. In a five-month prospective cohort intervention study in the acute care medical floors of their hospital in St Louis, 390 of 655 inpatients with blood glucose of less than 90m/dL were identified as high risk for the alert system. They had found that the alert process, when augmented by nurse-physician collaboration, had resulted in a 68 percent decrease in episodes of severe hypoglycemia in comparison to non-alerted high-risk patients (3.1 percent vs 9.7 percent).

At this meeting, Tobin reported they had successfully rolled out the continuous monitoring and alerting technology to the 9 hospitals in the system. The occurrence of severe hypoglycemia was reduced from 2.9 percent to 1.7 percent per 1,000 at-risk patient days during all visits, and from 4.4 percent to 2.5 percent during visits where an alert occurred across all hospitals.

“System-wide electronic surveillance and alerts improve safety in patients with diabetes across a range of clinical settings,” Tobin concluded.

Get the best insights in healthcare analytics directly to your inbox.

Related

How AI Is Shaking Up Healthcare, Beyond Diagnostics

Rise of the Anti-Opioid Algorithm

AI Diabetes Tech Continues FDA’s Software Push

Related Videos
Image: Ron Southwick, Chief Healthcare Executive
George Van Antwerp, MBA
Edmondo Robinson, MD
Craig Newman
© 2024 MJH Life Sciences

All rights reserved.