Nocturnal hypoglycemia is a common phenomenon among patients with diabetes and can lead to a broad range of adverse events and complications. Identifying factors associated with hypoglycemia can improve glucose control and patient care. We propose a repeated measures random forest (RMRF) algorithm that can handle nonlinear relationships and interactions and the correlated responses from patients evaluated over several nights. Simulation results show that our proposed algorithm captures the informative variable more often than naïvely assuming independence. RMRF also outperforms standard random forest and extremely randomized trees algorithms. We demonstrate scenarios where RMRF attains greater prediction accuracy than generalized linear models. We apply the RMRF algorithm to analyze a diabetes study with 2524 nights from 127 patients with type 1 diabetes. We find that nocturnal hypoglycemia is associated with HbA1c, bedtime blood glucose (BG), insulin on board, time system activated, exercise intensity, and daytime hypoglycemia. The RMRF can accurately classify nights at high risk of nocturnal hypoglycemia.
more » « less- Award ID(s):
- 1633130
- PAR ID:
- 10453601
- Publisher / Repository:
- Oxford University Press
- Date Published:
- Journal Name:
- Biometrics
- Volume:
- 77
- Issue:
- 1
- ISSN:
- 0006-341X
- Format(s):
- Medium: X Size: p. 343-351
- Size(s):
- p. 343-351
- Sponsoring Org:
- National Science Foundation
More Like this
-
Monitoring glucose levels is critical for effective diabetes management. Continuous glucose monitoring devices estimate interstitial glucose levels and provide alerts for glycemic excursions. However, they are expensive and invasive. Therefore, low-cost, noninvasive alternatives are useful for patients with diabetes. In this article, we explore electrocardiogram signals as a potential alternative to detecting glycemic excursions by extracting intrabeat (beat-morphology) and inter-beat (heart rate variability) information. Unlike prior methods that focused only on the standard clinical excursion thresholds (70 mg/dL for hypoglycemia, 180 mg/dL for hyperglycemia), our proposed approach trains independent machine learning models at various excursion thresholds, aggregating their outputs for a final prediction. This allows learning morphological patterns in the neighborhood of the standard excursion thresholds. Our personalized fusion models achieve an AUC of 75 % for hypoglycemia and 78% for hyperglycemia detection across patients, resulting in an average improvement of 4 % compared to the baseline models (trained using only standard clinical thresholds) for detecting glycemic excursions. We also find that combining morphology and HRV information outperforms using them individually (5 % for hypoglycemia and 6 % for hyperglycemia). The data used in this article was collected from 12 patients with type-1 diabetes, each monitored over a 14-day period at Texas Children’s Hospital, Houston. The results indicate that a combination of morphological and HRV features is essential for noninvasive detection of glycemic excursions. Also, morphological changes can happen at varying glucose levels for different patients and capturing these changes provide valuable information that leads to improved prediction performance for detecting glycemic excursions.more » « less
-
OBJECTIVE To determine the benefit of starting continuous glucose monitoring (CGM) in adult-onset type 1 diabetes (T1D) and type 2 diabetes (T2D) with regard to longer-term glucose control and serious clinical events.
RESEARCH DESIGN AND METHODS A retrospective observational cohort study within the Veterans Affairs Health Care System was used to compare glucose control and hypoglycemia- or hyperglycemia-related admission to an emergency room or hospital and all-cause hospitalization between propensity score overlap weighted initiators of CGM and nonusers over 12 months.
RESULTS CGM users receiving insulin (n = 5,015 with T1D and n = 15,706 with T2D) and similar numbers of nonusers were identified from 1 January 2015 to 31 December 2020. Declines in HbA1c were significantly greater in CGM users with T1D (−0.26%; 95% CI −0.33, −0.19%) and T2D (−0.35%; 95% CI −0.40, −0.31%) than in nonusers at 12 months. Percentages of patients achieving HbA1c <8 and <9% after 12 months were greater in CGM users. In T1D, CGM initiation was associated with significantly reduced risk of hypoglycemia (hazard ratio [HR] 0.69; 95% CI 0.48, 0.98) and all-cause hospitalization (HR 0.75; 95% CI 0.63, 0.90). In patients with T2D, there was a reduction in risk of hyperglycemia in CGM users (HR 0.87; 95% CI 0.77, 0.99) and all-cause hospitalization (HR 0.89; 95% CI 0.83, 0.97). Several subgroups (based on baseline age, HbA1c, hypoglycemic risk, or follow-up CGM use) had even greater responses.
CONCLUSIONS In a large national cohort, initiation of CGM was associated with sustained improvement in HbA1c in patients with later-onset T1D and patients with T2D using insulin. This was accompanied by a clear pattern of reduced risk of admission to an emergency room or hospital for hypoglycemia or hyperglycemia and of all-cause hospitalization.
-
We present a multitask learning approach to the problem of hypoglycemia (HG) prediction in diabetes. The approach is based on a state-of-the-art time series forecasting model, N-BEATS, and extends it by adding a classification task so that the model performs both glucose forecasting (i.e., predicting future glucose values) and HG prediction (i.e., probability of future HG events sometime within the prediction horizon). We also propose an alternative loss function that penalizes forecasting errors in the HG range. We evaluate the approach on a dataset containing over 1.6M recordings from 112 patients with type 1 diabetes who wore a continuous glucose monitor (CGM) for 90 days. Our results show that the classification branch significantly outperforms the forecasting branch on the problem of HG prediction, and that the new loss function is more effective at reducing forecasting errors in the HG range than multi-task learning.more » « less
-
Background: Monitoring glucose excursions is important in diabetes management. This can be achieved using continuous glucose monitors (CGMs). However, CGMs are expensive and invasive. Thus, alternative low-cost noninvasive wearable sensors capable of predicting glycemic excursions could be a game changer to manage diabetes. Methods: In this article, we explore two noninvasive sensor modalities, electrocardiograms (ECGs) and accelerometers, collected on five healthy participants over two weeks, to predict both hypoglycemic and hyperglycemic excursions. We extract 29 features encompassing heart rate variability features from the ECG, and time- and frequency-domain features from the accelerometer. We evaluated two machine-learning approaches to predict glycemic excursions: a classification model and a regression model. Results: The best model for both hypoglycemia and hyperglycemia detection was the regression model based on ECG and accelerometer data, yielding 76% sensitivity and specificity for hypoglycemia and 79% sensitivity and specificity for hyperglycemia. This had an improvement of 5% in sensitivity and specificity for both hypoglycemia and hyperglycemia when compared with using ECG data alone. Conclusions: Electrocardiogram is a promising alternative not only to detect hypoglycemia but also to predict hyperglycemia. Supplementing ECG data with contextual information from accelerometer data can improve glucose prediction.more » « less
-
In this paper, we provide an approach to data-driven control for artificial pancreas systems by learning neural network models of human insulin-glucose physiology from available patient data and using a mixed integer optimization approach to control blood glucose levels in real-time using the inferred models. First, our approach learns neural networks to predict the future blood glucose values from given data on insulin infusion and their resulting effects on blood glucose levels. However, to provide guarantees on the resulting model, we use quantile regression to fit multiple neural networks that predict upper and lower quantiles of the future blood glucose levels, in addition to the mean. Using the inferred set of neural networks, we formulate a model-predictive control scheme that adjusts both basal and bolus insulin delivery to ensure that the risk of harmful hypoglycemia and hyperglycemia are bounded using the quantile models while the mean prediction stays as close as possible to the desired target. We discuss how this scheme can handle disturbances from large unannounced meals as well as infeasibilities that result from situations where the uncertainties in future glucose predictions are too high. We experimentally evaluate this approach on data obtained from a set of 17 patients over a course of 40 nights per patient. Furthermore, we also test our approach using neural networks obtained from virtual patient models available through the UVA-Padova simulator for type-1 diabetes.more » « less