Predicting changes in systolic blood pressure using longitudinal patient records

Area of Science:

• Biomedical Informatics
• Clinical Data Analysis
• Predictive Modeling in Healthcare

Background:

• Longitudinal clinical records contain valuable information for predicting patient health trajectories.
• Systolic blood pressure (SBP) is a critical health indicator requiring accurate prediction for proactive management.
• Integrating structured and unstructured clinical data can enhance predictive accuracy.

Purpose of the Study:

• To introduce a novel model for predicting future changes in systolic blood pressure (SBP).
• To leverage both structured and unstructured data from longitudinal clinical records for SBP prediction.
• To evaluate the model's performance against a baseline prediction method.

Main Methods:

• Clinical records were chronologically sorted, and SBP measurements were extracted.
• A feature selection algorithm identified salient features from clinical notes.
• Least median squares regression was employed to predict future SBP changes based on selected features.

Main Results:

• The prediction model achieved a correlation coefficient of 0.47 on unseen test data (p<.0001).
• This performance represents an improvement over the baseline model's correlation coefficient of 0.39.
• The model demonstrates statistically significant predictive capability for SBP trends.

Conclusions:

• The developed model effectively predicts future systolic blood pressure changes using clinical record data.
• Integrating text-based clinical notes alongside structured data improves SBP prediction accuracy.
• This approach offers a promising tool for proactive patient monitoring and management in clinical settings.