Background: Achieving glycemic control in critically ill patients is of paramount importance, and has been linked to reductions in mortality, ICU length of stay (LOS), and morbidities such as infection, regardless whether or not the patient has diabetes. The myriad of illnesses and patient conditions render maintenance of glycemic control very challenging in this setting. Methods: This study involved collection of continuous glucose monitoring (CGM) data, and other associated measures, from the electronic medical records of 127 patients for the first 72 hours of ICU admission at The Ohio State University Wexner Medical Center (OSWMC). Data collected was used to develop a neural network-based predictive models. Findings: The model was shown to be accurate in predicting glucose using patient data not leveraged for model training. Model error, expressed as mean absolute difference percent, was 10.6% and 15.9% with respect to interstitial and serum blood glucose values, respectively. Clarke Error Grid Analysis (CEGA) of model predictions with respect to the reference CGM and blood glucose measurements revealed that >99% of model predictions could be regarded as clinically acceptable. Interpretation: The noted clinical acceptability of these models illustrates their potential utility within a clinical decision support system to assist healthcare providers in the optimization of glycemic management in critical care patients. Funding Statement: This study was supported by a $98,366 grant provided by the State of Ohio's Third Frontier Program (Grant #TECG20140125) and $20,000 provided by the Department of Anesthesiology at OSUWMC. Declaration of Interests: Dr. Jaume has nothing to disclose. Mr. Owais has nothing to disclose. Dr. Papadimos has nothing to disclose. Dr. Tripathi has nothing to disclose. Dr. Mavarez Martinez has nothing to disclose. Dr. Cameron has received one initial payment from the University of Toledo that was due to an initial licensing fee paid by Analytic Diabetic Systems, Inc. as part of licensing the patents referenced above as part of potential future commercial product development of a clinical decision support system. Dr. Pappada reports grants and personal fees from Analytic Diabetic Systems, Inc., personal fees from University of Toledo, during the conduct of the study; In addition, Drs. Pappada and Cameron hold patents (U.S. Patent No. 8,762,306 B2 and U.S. Patent No. 9,076,107 B2) related to the GlyCU CDSS with royalties paid to the University of Toledo and Co-Inventors (Drs. Pappada and Cameron) from Analytic Diabetic Systems, Inc. Dr. Pappada is on the Board of Directors for Analytic Diabetic Systems, Inc. and is a Co-Founder of the company, and holds an equity position in the corporation. Ethics Approval Statement: The study protocol was approved by The Ohio State University Institutional Review Board. All patients or their next of kin (if the patient had lost decision-making capacity) provided written informed consent.
Read full abstract