Article in Advances in Wound Care shows how tech can help pinpoint individual patient risk factors that delay wound healing, giving clinicians better insights into treatments and interventions
A peer-reviewed article written by clinicians and data scientists at Net Health highlights the value of machine learning and artificial intelligence (AI) in predicting wound healing time frames and risks. The article, “Predicting Chronic Wound Healing Time Using Machine Learning to Support Real-Time Clinical Decisions,” was published in the October issue of Advances in Wound Care.
Net Health is a leading provider of cloud-based software and analytics for specialty medical providers, including wound care, rehabilitation therapy, home health and hospice, and occupational medicine. The company provides several software applications and services for the wound care industry, including the leading electronic health record (EHR) system for hospitals and clinics and related services. The company’s database covers 4.7 million wound encounters collected over more than 20 years, making it one of the largest in the world.
For the study, clinicians and data scientists used information from Net Health’s Wound Care EHR to predict the likelihood of a patient’s wound healing in 4, 8, and 12 weeks from the start of treatment. The models were trained on three data sets of more than 1.2M wounds, including 187 covariates describing patient demographics, comorbidities such as weight, smoking status and other conditions, wound characteristics and time between treatments.
“Wound care costs the nation tens of billions of dollars each year,” said first author Matt Berezo, a Data Scientist with Net Health. “The impact of non-healing wounds adds unnecessary costs and negatively impacts the quality of life for patients. We wanted to find out what factors were inhibiting healing so we could help providers make faster and better decisions about how to treat patients, thereby improving outcomes and reducing cost of care.”
The Journal article is one of the first to prove the value of machine learning and AI to predict chronic wound healing and illustrate how these predictions can be used to support real-time clinical decisions. Additionally, while there have been prior studies about AI applied to wound healing, none have included as comprehensive a data set, nor evaluated such an extensive prediction horizon - from every visit until case conclusion.
“We wanted to develop models that could accurately predict wound healing time at every visit throughout the course of treatment, and also find which factors uniquely affect a patient’s likelihood to heal,” explained Berezo. “Through these insights, clinicians can make more timely decisions that will help improve treatment and reduce the costs of care.”
The algorithm developed by the company’s data scientists significantly outperforms other published models, providing a high level of confidence in the predictions being delivered to clinicians using the software. The algorithm used to analyze the data is also able to highlight comorbidities that contribute to the predicted healing trajectory of individual patients’ wounds. For example, predictors such as weight and conditions like diabetes and hypertension, especially when multiple such comorbidities are present, tend to negatively impact a wound’s likelihood of healing. This general approach aligns with the precision medicine methodologies being pursued in the healthcare industry today.
“Strategic workflow management and predictive tools are key to improving the quality and cost of care for patients,” said Cathy Thomas Hess, BSN, CWCN, Net Health’s Chief Clinical Officer. “We have built a predictive tool within the EHR to provide a valuable snapshot of clinical information based on the documentation captured within the workflow. The tool shows how select factors affect the patient’s healing trajectory so that steps can be taken by the clinical team to intervene and create a targeted and effective treatment plan.”
For additional information, visit https://www.nethealth.com/clinical-analytics-for-wound-care/. # # #