Mette Dencker Johansen

Toward big data analytics: review of predictive models in management of diabetes and its complications

Diabetes is one of the top priorities in medical science and health care management, and an abundance of data and information is available on these patients. Whether data stem from statistical models or complex pattern recognition models, they may be fused into predictive models that combine patient information and prognostic outcome results. Such knowledge could be used in clinical decision support, disease surveillance, and public health management to improve patient care. Our aim was to review the literature and give an introduction to predictive models in screening for and the management of prevalent short- and long-term complications in diabetes. Predictive models have been developed for management of diabetes and its complications, and the number of publications on such models has been growing over the past decade. Often multiple logistic or a similar linear regression is used for prediction model development, possibly owing to its transparent functionality. Ultimately, for prediction models to prove useful, they must demonstrate impact, namely, their use must generate better patient outcomes. Although extensive effort has been put in to building these predictive models, there is a remarkable scarcity of impact studies.

Professional continuous glucose monitoring in subjects with type 1 diabetes: retrospective hypoglycemia detection.

Background: An important task in diabetes management is detection of hypoglycemia. Professional continuous glucose monitoring (CGM), which produces a glucose reading every 5 min, is a powerful tool for retrospective identification of unrecognized hypoglycemia. Unfortunately, CGM devices tend to be inaccurate, especially in the hypoglycemic range, which limits their applicability for hypoglycemia detection. The objective of this study was to develop an automated pattern recognition algorithm to detect hypoglycemic events in retrospective, professional CGM. Method: Continuous glucose monitoring and plasma glucose (PG) readings were obtained from 17 data sets of 10 type 1 diabetes patients undergoing insulin-induced hypoglycemia. The CGM readings were automatically classified into a hypoglycemic group and a nonhypoglycemic group on the basis of different features from CGM readings and insulin injection. The classification was evaluated by comparing the automated classification with PG using sample-based and event-based sensitivity and specificity measures. Results: With an event-based sensitivity of 100%, the algorithm produced only one false hypoglycemia detection. The sample-based sensitivity and specificity levels were 78% and 96%, respectively. Conclusions: The automated pattern recognition algorithm provides a new approach for detecting unrecognized hypoglycemic events in professional CGM data. The tool may assist physicians and diabetologists in conducting a more thorough evaluation of the diabetes patients glycemic control and in initiating necessary measures for improving glycemic control.

A multistep algorithm for processing and calibration of microdialysis continuous glucose monitoring data

BACKGROUND The deviation of continuous subcutaneous glucose monitoring (CGM) data from reference blood glucose measurements is substantial, and adequate signal processing is required to reduce the discrepancy between subcutaneous glucose and blood glucose values. The purpose of this study was to develop a multistep algorithm for the processing and calibration of continuous subcutaneous glucose monitoring data with high accuracy and short delay. Algorithm PRESENTATION The algorithm comprises three steps: rate-limiting filtering, selective smoothing, and robust calibration. Initially, the algorithm detects nonphysiological glucose rate-of-change and corrects it with a weighted local polynomial. Noisy signal parts that require smoothing are then detected based on zero crossing count of the sensor signal first-order differences, and an exponentially weighted moving average smooths the noisy parts of the signal afterward. Finally, calibration is performed using a first-order polynomial as the conversion function, with coefficients being estimated using robust regression with a bi-square weight function. ALGORITHM PERFORMANCE: The performance of the algorithm was evaluated on 16 patients with type 1 diabetes mellitus. To compare the algorithm with state-of-the-art CGM data denoising and calibration, the rate-limiting filter and selective smoothing were replaced with an adaptive Kalman filter, and the calibration method was replaced with the calibration algorithm presented in one of the Medtronic (Northridge, CA) CGM patents. The median (mean) of the absolute relative deviation (ARD) of the sensor glucose values processed by the newly developed algorithm from capillary reference blood glucose measurements was 14.8% (22.6%), 10.6% (14.6%), and 8.9% (11.7%) in hypoglycemia, euglycemia, and hyperglycemia, respectively, whereas for the alternative algorithm, the median (mean) was 22.2% (26.9%), 12.1% (15.9%), and 8.8 (11.3%), respectively. The median (mean) ARD in all ranges was 10.3% (14.7%) for the new algorithm and 11.5% (15.8%) for the alternative algorithm. The new algorithm had an average delay of 2.1 min across the patients, and the alternative algorithm had an average delay of 2.9 min. CONCLUSIONS The presented algorithm may increase the accuracy of CGM data.

Real-Time Hypoglycemia Detection from Continuous Glucose Monitoring Data of Subjects with Type 1 Diabetes

BACKGROUND Hypoglycemia is a potentially fatal condition. Continuous glucose monitoring (CGM) has the potential to detect hypoglycemia in real time and thereby reduce time in hypoglycemia and avoid any further decline in blood glucose level. However, CGM is inaccurate and shows a substantial number of cases in which the hypoglycemic event is not detected by the CGM. The aim of this study was to develop a pattern classification model to optimize real-time hypoglycemia detection. MATERIALS AND METHODS Features such as time since last insulin injection and linear regression, kurtosis, and skewness of the CGM signal in different time intervals were extracted from data of 10 male subjects experiencing 17 insulin-induced hypoglycemic events in an experimental setting. Nondiscriminative features were eliminated with SEPCOR and forward selection. The feature combinations were used in a Support Vector Machine model and the performance assessed by sample-based sensitivity and specificity and event-based sensitivity and number of false-positives. RESULTS The best model was composed by using seven features and was able to detect 17 of 17 hypoglycemic events with one false-positive compared with 12 of 17 hypoglycemic events with zero false-positives for the CGM alone. Lead-time was 14 min and 0 min for the model and the CGM alone, respectively. CONCLUSIONS This optimized real-time hypoglycemia detection provides a unique approach for the diabetes patient to reduce time in hypoglycemia and learn about patterns in glucose excursions. Although these results are promising, the model needs to be validated on CGM data from patients with spontaneous hypoglycemic events.

Comparison between one-point calibration and two-point calibration approaches in a continuous glucose monitoring algorithm

Background: The purpose of this study was to investigate the effect of using a 1-point calibration approach instead of a 2-point calibration approach on the accuracy of a continuous glucose monitoring (CGM) algorithm. Method: A previously published real-time CGM algorithm was compared with its updated version, which used a 1-point calibration instead of a 2-point calibration. In addition, the contribution of the corrective intercept (CI) to the calibration performance was assessed. Finally, the sensor background current was estimated real-time and retrospectively. The study was performed on 132 type 1 diabetes patients. Results: Replacing the 2-point calibration with the 1-point calibration improved the CGM accuracy, with the greatest improvement achieved in hypoglycemia (18.4% median absolute relative differences [MARD] in hypoglycemia for the 2-point calibration, and 12.1% MARD in hypoglycemia for the 1-point calibration). Using 1-point calibration increased the percentage of sensor readings in zone A+B of the Clarke error grid analysis (EGA) in the full glycemic range, and also enhanced hypoglycemia sensitivity. Exclusion of CI from calibration reduced hypoglycemia accuracy, while slightly increased euglycemia accuracy. Both real-time and retrospective estimation of the sensor background current suggest that the background current can be considered zero in the calibration of the SCGM1 sensor. Conclusions: The sensor readings calibrated with the 1-point calibration approach indicated to have higher accuracy than those calibrated with the 2-point calibration approach.

Accuracy Evaluation of a New Real-Time Continuous Glucose Monitoring Algorithm in Hypoglycemia

BACKGROUND The purpose of this study was to evaluate the performance of a new continuous glucose monitoring (CGM) calibration algorithm and to compare it with the Guardian(®) REAL-Time (RT) (Medtronic Diabetes, Northridge, CA) calibration algorithm in hypoglycemia. SUBJECTS AND METHODS CGM data were obtained from 10 type 1 diabetes patients undergoing insulin-induced hypoglycemia. Data were obtained in two separate sessions using the Guardian RT CGM device. Data from the same CGM sensor were calibrated by two different algorithms: the Guardian RT algorithm and a new calibration algorithm. The accuracy of the two algorithms was compared using four performance metrics. RESULTS The median (mean) of absolute relative deviation in the whole range of plasma glucose was 20.2% (32.1%) for the Guardian RT calibration and 17.4% (25.9%) for the new calibration algorithm. The mean (SD) sample-based sensitivity for the hypoglycemic threshold of 70 mg/dL was 31% (33%) for the Guardian RT algorithm and 70% (33%) for the new algorithm. The mean (SD) sample-based specificity at the same hypoglycemic threshold was 95% (8%) for the Guardian RT algorithm and 90% (16%) for the new calibration algorithm. The sensitivity of the event-based hypoglycemia detection for the hypoglycemic threshold of 70 mg/dL was 61% for the Guardian RT calibration and 89% for the new calibration algorithm. Application of the new calibration caused one false-positive instance for the event-based hypoglycemia detection, whereas the Guardian RT caused no false-positive instances. The overestimation of plasma glucose by CGM was corrected from 33.2 mg/dL in the Guardian RT algorithm to 21.9 mg/dL in the new calibration algorithm. CONCLUSIONS The results suggest that the new algorithm may reduce the inaccuracy of Guardian RT CGM system within the hypoglycemic range; however, data from a larger number of patients are required to compare the clinical reliability of the two algorithms.

Interindividual and intraindividual variations in postprandial glycemia peak time complicate precise recommendations for self-monitoring of glucose in persons with type 1 diabetes mellitus.

Background: In glycemic control, postprandial glycemia may be important to monitor and optimize as it reveals glycemic control quality and postprandial hyperglycemia partly predicts late diabetic complications. Self-monitoring of blood glucose (SMBG) may be an appropriate technology to use, but recommendations on measurement time are crucial. Method: We retrospectively analyzed interindividual and intraindividual variations in postprandial glycemic peak time. Continuous glucose monitoring (CGM) and carbohydrate intake were collected in 22 patients with type 1 diabetes mellitus. Meals were identified from carbohydrate intake data. For each meal, peak time was identified as time from meal to CGM zenith within 40–150 min after meal start. Interindividual (one-way Anova) and intraindividual (intraclass correlation coefficient) variation was calculated. Results: Nineteen patients were included with sufficient meal data quality. Mean peak time was 87 ± 29 min. Mean peak time differed significantly between patients (p = 0.02). Intraclass correlation coefficient was 0.29. Conclusions: Significant interindividual and intraindividual variations exist in postprandial glycemia peak time, thus hindering simple and general advice regarding postprandial SMBG for detection of maximum values.

A classification model for predicting eye disease in newly diagnosed people with type 2 diabetes

Diabetic retinopathy may be present at the time type 2 diabetes is diagnosed, and initial screening encompassing an eye examination performed by an ophthalmologist or optometrist is therefore recommended. However, proper screening for retinopathy may be challenging in many parts of the world. We hypothesized that simple, commonly available patient characteristics can be used to identify patients at high risk for having retinopathy. We investigated data from multiple years extracted from the National Health and Nutrition Examination Survey which holds information about blood glucose and eye examinations. Individuals with hitherto undiagnosed diabetes were classified according to the presence or absence of retinopathy. Linear classification was used to predict which patients had retinopathy at the time of diagnosis. A total of 266 individuals with undiagnosed diabetes were identified from the cohorts. Of these, 222 individuals had no sign of retinopathy, whereas 44 had mild or moderate non-proliferative retinopathy. Using information regarding HbA1c, BMI, waist circumference, age, systolic blood pressure, urinary albumin, and urinary creatinine, we were able to construct a model that predicts the presence of retinopathy with a positive predictive value of 22% and a negative predictive value of 99%. Only one true positive (1/44) with mild non-proliferative retinopathy was falsely classified. A classification model using readily available patient information and routine biochemical measures can be used to identify patients at high risk of having retinopathy at the time their diabetes is diagnosed. The model may be used to identify high-risk patients for retinopathy screening.

Evaluation of an Algorithm for Retrospective Hypoglycemia Detection Using Professional Continuous Glucose Monitoring Data

Background: People with type 1 diabetes (T1D) are unable to produce insulin and thus rely on exogenous supply to lower their blood glucose. Studies have shown that intensive insulin therapy reduces the risk of late-diabetic complications by lowering average blood glucose. However, the therapy leads to increased incidence of hypoglycemia. Although inaccurate, professional continuous glucose monitoring (PCGM) can be used to identify hypoglycemic events, which can be useful for adjusting glucose-regulating factors. New pattern classification approaches based on identifying hypoglycemic events through retrospective analysis of PCGM data have shown promising results. The aim of this study was to evaluate a new pattern classification approach by comparing the performance with a newly developed PCGM calibration algorithm. Methods: Ten male subjects with T1D were recruited and monitored with PCGM and self-monitoring blood glucose during insulin-induced hypoglycemia. A total of 19 hypoglycemic events occurred during the sessions. Results: The pattern classification algorithm detected 19/19 hypoglycemic events with 1 false positive, while the PCGM with the new calibration algorithm detected 17/19 events with 2 false positives. Conclusions: We can conclude that even after the introduction of new calibration algorithms, the pattern classification approach is still a valuable addition for improving retrospective hypoglycemia detection using PCGM.

A novel model enhances HbA1c-based diabetes screening using simple anthropometric, anamnestic, and demographic information.

The sensitivity of HbA1c is not optimal for the screening of patients with latent diabetes. We hypothesize that simple healthcare information could improve accuracy.