Predicting hypertension control using machine learning.

Area of Science:

• * Medical Informatics
• * Machine Learning in Healthcare
• * Cardiovascular Disease Research

Background:

• * Hypertension is a prevalent condition with significant adverse effects when uncontrolled.
• * Predicting blood pressure control is challenging due to diverse therapeutic regimens and patient factors.
• * Accurate prediction of hypertension control is crucial for effective patient management.

Purpose of the Study:

• * To investigate the efficacy of machine learning (ML) in predicting hypertension control within 12 months.
• * To develop and evaluate an ML model using retrospective electronic medical record data.
• * To assess the potential of ML in improving the accuracy of hypertension control predictions.

Main Methods:

• * Retrospective analysis of electronic medical records from 350,008 patients (aged ≥18) between January 2015 and June 2022.
• * Data included medication, lab values, vital signs, comorbidities, encounters, and demographics.
• * A sliding time window approach created 287 predictive models, each trained on 2 years of data and tested on 1 week, to prevent data leakage.

Main Results:

• * The ML model achieved an Area Under the Curve (AUC) of 0.76 for predicting blood pressure control within 12 months.
• * Key performance metrics included sensitivity of 61.52%, specificity of 75.69%, positive predictive value of 67.75%, and negative predictive value of 70.49%.
• * The AUC of 0.756 is considered moderately good for machine learning models in this domain.

Conclusions:

• * Machine learning shows promise in accurately predicting hypertension control using existing electronic health record data.
• * The developed model, incorporating uncertainty analysis, offers a robust solution for hypertension management.
• * Further research and clinical deployment are necessary to confirm the clinical relevance and impact on health outcomes.