Machine Learning Analysis of Lifestyle Modification in Hypertension Treatment

Tuesday, April 14, 2015 -
4:00pm to 5:00pm
Biotechnology 1360, 425 Henry Mall

Speaker Name: 

Kimberly Shoenbill

Speaker Institution: 

UW-Madison, Biostatistics and Medical Informatics




47% of the 78 million US adults with hypertension do not have blood pressures controlled to target levels. Uncontrolled hypertension is associated with increased risks of stroke, heart failure, heart attack and kidney disease. Multiple guidelines on the diagnosis and management of hypertension include lifestyle modification as first line treatment for all hypertensive patients (e.g., exercise, dietary sodium reduction). Despite guidelines, delays in hypertension diagnosis and medication initiation have been documented in prior studies, but not fully explained. My PhD work will help fill this knowledge gap by providing insight into clinical characteristics and lifestyle modification efforts that may be correlated with delays in diagnosis and medication management of hypertension.

In this presentation, I will discuss preliminary findings including the description and frequency of lifestyle modification counseling and patient-reported activities in adult hypertensives’ EHR records. Since much of this information does not exist in coded form, I am using natural language processing techniques to extract this information from narrative reports. In order to more effectively accomplish this task, I have created a lifestyle modification terminology that will be added to existing NLP tools. Future steps in this project will include using machine learning algorithms (e.g., decision trees, random forests) to determine characteristics of adult hypertensive patients and their hypertension interventions that correlate with delays or accelerations in hypertension diagnosis and/or medication initiation.