Predicting Inpatient Glucose Levels and Insulin Dosing by Machine Learning on Electronic Health Records

Liu, Xiran; Janković, Ivana; Chen, Jonathan

doi:10.1101/2020.03.02.20029017

Cited by 4 publications

(3 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“… 14 , 15 , 16 , 17 Predictive models using statistical inference and machine learning are finding opportunities to enhance diagnosis. 16 , 17 , 18 , 19 There is prior research identifying risk factors for STEMI as a disease. 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 However, developing a predictive model to enhance timely diagnosis requires an understanding of risk factors for STEMI diagnostic delay.…”

mentioning

confidence: 99%

Fallacy of Median Door‐to‐ECG Time: Hidden Opportunities for STEMI Screening Improvement

Yiadom

Gong

Patterson

et al. 2022

JAHA

View full text Add to dashboard Cite

Background ST‐segment elevation myocardial infarction (STEMI) guidelines recommend screening arriving emergency department (ED) patients for an early ECG in those with symptoms concerning for myocardial ischemia. Process measures target median door‐to‐ECG (D2E) time of 10 minutes. Methods and Results This 3‐year descriptive retrospective cohort study, including 676 ED‐diagnosed patients with STEMI from 10 geographically diverse facilities across the United States, examines an alternative approach to quantifying performance: proportion of patients meeting the goal of D2E≤10 minutes. We also identified characteristics associated with D2E>10 minutes and estimated the proportion of patients with screening ECG occurring during intake, triage, and main ED care periods. We found overall median D2E was 7 minutes (IQR:4–16; range: 0–1407 minutes; range of ED medians: 5–11 minutes). Proportion of patients with D2E>10 minutes was 37.9% (ED range: 21.5%–57.1%). Patients with D2E>10 minutes, compared to those with D2E≤10 minutes, were more likely female (32.8% versus 22.6%, P =0.005), Black (23.4% versus 12.4%, P =0.005), non‐English speaking (24.6% versus 19.5%, P =0.032), diabetic (40.2% versus 30.2%, P =0.010), and less frequently reported chest pain (63.3% versus 87.4%, P <0.001). ECGs were performed during ED intake in 62.1% of visits, ED triage in 25.3%, and main ED care in 12.6%. Conclusions Examining D2E>10 minutes can identify opportunities to improve care for more ED patients with STEMI. Our findings suggest sex, race, language, and diabetes are associated with STEMI diagnostic delays. Moving the acquisition of ECGs completed during triage to intake could achieve the D2E≤10 minutes goal for 87.4% of ED patients with STEMI. Sophisticated screening, accounting for differential risk and diversity in STEMI presentations, may further improve timely detection.

show abstract

mentioning

confidence: 99%

Fallacy of Median Door‐to‐ECG Time: Hidden Opportunities for STEMI Screening Improvement

Yiadom

Gong

Patterson

et al. 2022

JAHA

View full text Add to dashboard Cite

show abstract

“…It is a regression model that takes into account all of the inputs to forecast the bolus and basal insulin doses. The association between scalar dependent and independent variables may be discovered using the regression model [25]. Random forest consists of a collection of different regression trees.…”

Section: Random Forest Regression Model For Insulin Dosage Predictionmentioning

confidence: 99%

Time series prediction of personalized insulin dosage for type 2 diabetics

G.,

Bhuvan,

Jerrard

2023

IJEECS

View full text Add to dashboard Cite

Careful blood glucose monitoring and consistent insulin administration are necessary for managing diabetes. People with demanding schedules or little access to medical personnel may find this difficult. Fortunately, without having to visit a doctor every day, daily insulin dosage may now be customized to a person’s unique needs using technology and customised algorithms based on their food intake, exercise routines, and blood glucose levels. This information can be entered into a diabetes management app or device, where an algorithm will determine the proper insulin dosage and offer real-time feedback to assist maintain ideal blood glucose levels. A patient's dietary preferences, degree of physical activity, and blood sugar are taken into account for determining the proper bolus and basal insulin dosages in this study. According to the tracked body data, a patient’s appropriate insulin dosage is predicted using artificial neural network (ANN)-based models. Based on patient activity, food intake, exercise, and past insulin administration, insulin projections are created. To forecast an individual’s basal and bolus insulin requirements, long short-term memory (LSTM) and random forest regression models are employed. Accuracy of both models are tested and random forest regression shows better accuracy which is used in the prediction system.

show abstract

“…To our knowledge, there have been no prior studies predicting actual insulin doses in inpatients, although one study predicting what dose of insulin clinicians would order yielded an error of approximately 73%. 24 …”

Section: Introductionmentioning

confidence: 99%

Machine learning for initial insulin estimation in hospitalized patients

Nguyen

Janković

Kalesinskas

et al. 2021

Journal of the American Medical Informatics Association

Self Cite

View full text Add to dashboard Cite

Objective The study sought to determine whether machine learning can predict initial inpatient total daily dose (TDD) of insulin from electronic health records more accurately than existing guideline-based dosing recommendations. Materials and Methods Using electronic health records from a tertiary academic center between 2008 and 2020 of 16 848 inpatients receiving subcutaneous insulin who achieved target blood glucose control of 100-180 mg/dL on a calendar day, we trained an ensemble machine learning algorithm consisting of regularized regression, random forest, and gradient boosted tree models for 2-stage TDD prediction. We evaluated the ability to predict patients requiring more than 6 units TDD and their point-value TDDs to achieve target glucose control. Results The method achieves an area under the receiver-operating characteristic curve of 0.85 (95% confidence interval [CI], 0.84-0.87) and area under the precision-recall curve of 0.65 (95% CI, 0.64-0.67) for classifying patients who require more than 6 units TDD. For patients requiring more than 6 units TDD, the mean absolute percent error in dose prediction based on standard clinical calculators using patient weight is in the range of 136%-329%, while the regression model based on weight improves to 60% (95% CI, 57%-63%), and the full ensemble model further improves to 51% (95% CI, 48%-54%). Discussion Owing to the narrow therapeutic window and wide individual variability, insulin dosing requires adaptive and predictive approaches that can be supported through data-driven analytic tools. Conclusions Machine learning approaches based on readily available electronic medical records can discriminate which inpatients will require more than 6 units TDD and estimate individual doses more accurately than standard guidelines and practices.

show abstract

Predicting Inpatient Glucose Levels and Insulin Dosing by Machine Learning on Electronic Health Records

Cited by 4 publications

References 16 publications

Fallacy of Median Door‐to‐ECG Time: Hidden Opportunities for STEMI Screening Improvement

Fallacy of Median Door‐to‐ECG Time: Hidden Opportunities for STEMI Screening Improvement

Time series prediction of personalized insulin dosage for type 2 diabetics

Machine learning for initial insulin estimation in hospitalized patients

Contact Info

Product

Resources

About