A Machine Learning Model for Week-Ahead Hypoglycemia Prediction From Continuous Glucose Monitoring Data

BACKGROUND

Remote patient monitoring (RPM) programs augment type 1 diabetes (T1D) care based on retrospective continuous glucose monitoring (CGM) data. Few methods are available to estimate the likelihood of a patient experiencing clinically significant hypoglycemia within one week.

METHODS

We developed a machine learning model to estimate the probability that a patient will experience a clinically significant hypoglycemic event, defined as CGM readings below 54 mg/dL for at least 15 consecutive minutes, within one week. The model takes as input the patient's CGM time series over a given week, and outputs the predicted probability of a clinically significant hypoglycemic event the following week. We used 10-fold cross-validation and external validation (testing on cohorts different from the training cohort) to evaluate performance. We used CGM data from three different cohorts of patients with T1D: REPLACE-BG (226 patients), Juvenile Diabetes Research Foundation (JDRF; 355 patients) and Tidepool (120 patients).

RESULTS

In 10-fold cross-validation, the average area under the receiver operating characteristic curve (ROC-AUC) was 0.77 (standard deviation [SD]: 0.0233) on the REPLACE-BG cohort, 0.74 (SD: 0.0188) on the JDRF cohort, and 0.76 (SD: 0.02) on the Tidepool cohort. In external validation, the average ROC-AUC across the three cohorts was 0.74 (SD: 0.0262).

CONCLUSIONS

We developed a machine learning algorithm to estimate the probability of a clinically significant hypoglycemic event within one week. Predictive algorithms may provide diabetes care providers using RPM with additional context when prioritizing T1D patients for review.

Explainable Machine Learning for Real-Time Hypoglycemia and Hyperglycemia Prediction and Personalized Control Recommendations

BACKGROUND

The occurrences of acute complications arising from hypoglycemia and hyperglycemia peak as young adults with type 1 diabetes (T1D) take control of their own care. Continuous glucose monitoring (CGM) devices provide real-time glucose readings enabling users to manage their control proactively. Machine learning algorithms can use CGM data to make ahead-of-time risk predictions and provide insight into an individual's longer term control.

METHODS

We introduce explainable machine learning to make predictions of hypoglycemia (<70 mg/dL) and hyperglycemia (>270 mg/dL) up to 60 minutes ahead of time. We train our models using CGM data from 153 people living with T1D in the CITY (CGM Intervention in Teens and Young Adults With Type 1 Diabetes)survey totaling more than 28 000 days of usage, which we summarize into (short-term, medium-term, and long-term) glucose control features along with demographic information. We use machine learning explanations (SHAP [SHapley Additive exPlanations]) to identify which features have been most important in predicting risk per user.

RESULTS

Machine learning models (XGBoost) show excellent performance at predicting hypoglycemia (area under the receiver operating curve [AUROC]: 0.998, average precision: 0.953) and hyperglycemia (AUROC: 0.989, average precision: 0.931) in comparison with a baseline heuristic and logistic regression model.

CONCLUSIONS

Maximizing model performance for glucose risk prediction and management is crucial to reduce the burden of alarm fatigue on CGM users. Machine learning enables more precise and timely predictions in comparison with baseline models. SHAP helps identify what about a CGM user's glucose control has led to predictions of risk which can be used to reduce their long-term risk of complications.

Feature-Based Machine Learning Model for Real-Time Hypoglycemia Prediction

BACKGROUND

Hypoglycemia is a serious health concern in youth with type 1 diabetes (T1D). Real-time data from continuous glucose monitoring (CGM) can be used to predict hypoglycemic risk, allowing patients to take timely intervention measures.

METHODS

A machine learning model is developed for probabilistic prediction of hypoglycemia (<70 mg/dL) in 30- and 60-minute time horizons based on CGM datasets obtained from 112 patients over a range of 90 days consisting of over 1.6 million CGM values under normal living conditions. A comprehensive set of features relevant for hypoglycemia are developed and a parsimonious subset with most influence on predicting hypoglycemic risk is identified. Model performance is evaluated both with and without contextual information on insulin and carbohydrate intake.

RESULTS

The model predicted hypoglycemia with >91% sensitivity for 30- and 60-minute prediction horizons while maintaining specificity >90%. Inclusion of insulin and carbohydrate data yielded performance improvement for 60-minute but not for 30-minute predictions. Model performance was highest for nocturnal hypoglycemia (~95% sensitivity). Shortterm (less than one hour) and medium-term (one to four hours) features for good prediction performance are identified.

CONCLUSIONS

Innovative feature identification facilitated high performance for hypoglycemia risk prediction in pediatric youth with T1D. Timely alerts of impending hypoglycemia may enable proactive measures to avoid severe hypoglycemia and achieve optimal glycemic control. The model will be deployed on a patient-facing smartphone application in an upcoming pilot study.

Inpatient Hypoglycemia: The Challenge Remains

Hypoglycemia in inpatients with diabetes remains the most common complication of diabetes therapies. Hypoglycemia is independently associated with increased morbidity and mortality, increased length of stay, increased readmission rate, and increased cost. This review describes the importance of reporting and addressing inpatient hypoglycemia; it further summarizes eight strategies that aid clinicians in the prevention of inpatient hypoglycemia: auditing the electronic medical record, formulary restrictions and dose-limiting strategies, hyperkalemia order sets, electronic glucose management systems, prediction tools, diabetes self-management, remote surveillance, and noninsulin medications.

Hypoglycemia prediction using machine learning models for patients with type 2 diabetes

Minimizing the occurrence of hypoglycemia in patients with type 2 diabetes is a challenging task since these patients typically check only 1 to 2 self-monitored blood glucose (SMBG) readings per day. We trained a probabilistic model using machine learning algorithms and SMBG values from real patients. Hypoglycemia was defined as a SMBG value < 70 mg/dL. We validated our model using multiple data sets. In addition, we trained a second model, which used patient SMBG values and information about patient medication administration. The optimal number of SMBG values needed by the model was approximately 10 per week. The sensitivity of the model for predicting a hypoglycemia event in the next 24 hours was 92% and the specificity was 70%. In the model that incorporated medication information, the prediction window was for the hour of hypoglycemia, and the specificity improved to 90%. Our machine learning models can predict hypoglycemia events with a high degree of sensitivity and specificity. These models-which have been validated retrospectively and if implemented in real time-could be useful tools for reducing hypoglycemia in vulnerable patients.

Hypoglycemia detection and prediction using continuous glucose monitoring-a study on hypoglycemic clamp data

MOTIVATION

The fear of hypoglycemia remains an important limiting factor in the ability of an individual with type 1 diabetes to tightly regulate glycemia. Continuous glucose monitors provide important feedback to improve glycemic control, but there remains a need for these devices to better alarm of possible impending hypoglycemia, particularly overnight or other periods when the individual is engaged in activities that take their focus away from glucose monitoring.

METHODS

We have previously proposed an algorithm, based on the use of real-time glucose sensor signals and optimal estimation theory (Kalman filtering), to predict hypoglycemia; the algorithm was validated in simulation-based studies. In this article we further refine and validate the prediction algorithm based on the analysis of clinical hypoglycemic clamp data from 13 subjects. The sensitivity and specificity of the predictions are calculated with respect to reference blood glucose values obtained at the same sampling rate of the sensor.

RESULTS

For a 30-minute prediction horizon and alarm threshold of 70 mg/dl, the sensitivity and specificity were 90 and 79%, respectively, indicating that a 21% false alarm rate must be tolerated to predict 90% of the hypoglycemic events 30 minutes ahead of time. Shorter prediction horizons yield a significant improvement in sensitivity and specificity.

DISCUSSION

Sensitivity and specificity data as a function of prediction horizon and alarm threshold enable an individual to adjust the alarm to best meet their needs. Such decisions can be made depending on the subject's risk for hypoglycemia, for example.