September 1, 2021 | Net Health

3 Minute Read

Net Health Expands Use of Analytics to Predict Wound Healing Time

Net Health is continuing to push the boundaries of using analytics to improve wound care with its latest advance: software that can predict the probability of wound healing over time. Net Health recently announced the addition of two new predictive analytics capabilities to its Net Health® Wound Care electronic health record software solution.

The software was previewed in a recent webinar, Using Analytics to Predict Wound Healing and Improve Outcomes, by Josh Budman, Vice President of Analytics, and Data Scientist Matt Berezo. It represents a major step toward the company’s goal of providing clinicians with tools that can not only predict outcomes, but also the impact of treatments and interventions.

From Predictive to Prescriptive

The healthcare industry is embracing AI and analytics at an ever-increasing rate. As Budman explained, analytics fall into one of four buckets:

1. Descriptive analytics show what happened. 

2. Diagnostic analytics tell why it happened. 

3. Predictive analytics show what is going to happen; they improve clinical and financial decision-making by predicting outcomes. 

4. Prescriptive analytics take it to the next rung, showing that if you do this, that will happen.

In this case, Net Health is using predictive analysis to calculate the probability that a wound will heal during classic clinical benchmarks for healing timeframes of four, eight, twelve or sixteen weeks from the start of treatment. Since the predictions and wound conditions change from visit to visit, clinicians can use the information to adjust treatment accordingly.

Net Health’s models consider 177 data points, or variables, covering patient demographic data, historical comorbidities and wound type, location, description, and size as measured in previous and current visits.

In the webinar, Berezo detailed how variable importance analysis was done using Shapley Additive Explanations (SHAP), which considers the marginal contribution a variable has on a single prediction. This yields information about how a clinical variable generally affected predictions for the population and for individual patients. Importantly, the SHAP values can also be mined and analyzed for an individual patient and wound-level predictions.

Use in Clinical Settings

The new algorithms will provide rich, patient-specific insights into the positive and negative factors impacting healing— across all care settings and with limited manual data entry. Results are extremely accurate.

“The goal of these algorithms is not to replace the clinician or end-user,” stressed Berezo. “We’re giving you the important data; it’s up to you to derive the data-driven insights and determine the appropriate treatment.”

“We like to think of the algorithm as a guardian angel for identifying patient risk,” he continued. “This is a tool that can drive data-driven decision-making that may improve clinical outcomes.”

The algorithm may identify patients that are high-risk of a wound not healing but may appear low-risk or vice versa. The results also may help clinicians set expectations and drive discussion with patients.

Find the prospect of predicting wound healing exciting? Listen to the webinar to learn more.

Using Analytics to Predict Wound Healing and Improve Clinical Outcomes

See how machine learning algorithms can empower clinicians to provide better care for their patients

Share this post

Subscribe and See More