Machine-Learning Technique Improves High Blood Pressure Treatment

Published in the International Journal of Epidemiology, a new UCLA study describes how a machine-learning technique known as “casual forest” could determine the hypertension patients that would benefit the most from treatment rather than assuming that the highest-risk patients require the most clinical attention.

According to the Centers for Disease Control and Prevention (CDC), over 670,000 deaths in the US in 2020 were attributed to hypertension. In addition, while about 47 percent of US adults have hypertension, only 24 percent of this population has the condition under control.

Traditionally, clinicians treating patients with high blood pressure focus on those with the highest risk of poor outcomes, as the assumption is that they will require the highest level of treatment.

However, new research provides evidence surrounding how machine-learning tools can assist researchers in gathering subjective data to determine how treatment may vary from patient to patient.

In this study, researchers described a new technique known as “the casual forest” and detailed its ability to determine the efficacy of hypertension interventions alongside patient characteristics.

The study included data from 10,672 participants, all of whom were randomized to systolic blood pressure (SBP) targets of either less than 120 mmHg or less than 140 mmHg from two randomized controlled trials.

The researchers used the casual forest technique to create a prediction model of individualized treatment effect related to the control of SBP and its correlation with reductions in adverse cardiovascular outcomes after three years.

They found that 78.9 percent of individuals with an SBP greater than 130 mmHg achieved benefits from intensive SBP control.

Researchers also found that treating patients based on benefit rather than risk improved outcomes.

“We found that a substantial number of individuals without hypertension benefited from lowering their blood pressure,” said lead author Kosuke Inoue, MD, PhD, who undertook the study while an epidemiology graduate student at the UCLA Fielding School of Public Health and is now an associate professor of social epidemiology at Kyoto University, in a press release. “By applying the causal forest method, we found that treating individuals with high estimated benefits provided better population health outcomes than the traditional high-risk approach.”

Further, researchers noted that high-benefit approaches could increase the efficacy associated with treatment, potentially being more reliable compared to high-risk approaches.

The application of machine-learning techniques to improve care and clinical operations is growing. 

A recent study from Yale University described the creation of a machine-learning model to assist researchers in determining physician turnover rates, allowing healthcare organizations to intervene and ultimately increase retention.

Researchers reviewed EHR data and used it to evaluate physician behavioral patterns to draw conclusions surrounding turnover trends. With this data, researchers sought to determine if it was possible to use a three-month period of information to assess the odds of physician departure within six months.

Following the testing of the model, researchers concluded that it was 97 percent accurate in predicting physician departure.