NLP enables Atrius Health to gain insights from unstructured data
Technology culls through critical clinical information in physician notes to support value-based care.
Atrius Health, a large multi-specialty ambulatory group practice serving eastern and central Massachusetts, is leveraging natural language processing to tap into a treasure trove of unstructured patient data used to support its value-based care initiatives.
By implementing an NLP platform from Linguamatics, Atrius is able to mine critical clinical information hidden within unstructured text to improve Medicare Advantage reimbursement, streamline accountable care organization reporting and close gaps in care delivery.
“The majority of clinically useful information is unstructured text that includes physician notes and radiology reports, which impacts our patients and our practice,” says Craig Monsen, MD, Atrius Health’s medical director of analytics. “I’d estimate that probably 50 percent to 80 percent is unstructured data.”
According to Monsen, there are several of the healthcare organization’s workflows within which NLP is being used to identify and extract critical clinical data that exists as unstructured text.
“We found for our heart failure patients that their heart pump function is not captured as structured data,” he says. “We’ve integrated natural language processing into our quality metrics reporting pipeline so that we can figure out who are these high-risk heart failure patients.”
Monsen adds that if an organization is trying to find that kind of information—such as the ejection fraction, which measures how well a heart is pumping—on a population level, they will “need to hire a small army” to manually extract it. However, he says the NLP tool enables Atrius to automate some of that process.
When it comes to closing care gaps, Atrius is also using the platform to identify patients with difficult-to-track conditions, such as pulmonary nodules, which are small growths in the lungs. “Most often, pulmonary nodules are benign, but in some cases, they’re an early sign of lung cancer,” according to Monsen. “Frequently, that information is only captured in free text, so that creates a situation where serious concerns can fall through the cracks.”
However, Atrius has invested in a series of “patient safety net” initiatives for several types of cancer designed to improve outcomes by leveraging NLP. “In the case of lung cancer, for example, the technology helps us to identify pulmonary nodules from radiology reports, making sure that we have followed up with those patients and have carried out the necessary interventions,” he notes.
Monsen also points out that NLP is ensuring the accuracy of clinical documentation workflows so that Atrius has sufficient resources to close gaps in care.
“We’re about 900 physicians across 30 practice locations, and we are heavily committed to value-based care, with something like 75 percent of our revenue coming from risk-based contracts,” says Monsen. “For us, it’s not just upside risk, but it’s also downside risk, so it’s important to make sure that we are resourced to provide the care, quality and safety nets.”
Physician burnout, a growing epidemic among provider organizations characterized by exhaustion, cynicism and feelings of ineffectiveness, has been linked to decreased quality of care and medical errors. However, Monsen believes that NLP will help Atrius alleviate the “data overload problem” and reduce physician burnout among its ranks because of data-entry tasks.
“The answer to closing information gaps can’t be to put everything into a structured field,” he observes. “Clinicians are already spending half their time in front of the electronic health record. This time is better spent with patients and having conversations that will contribute to their care rather than documenting what’s been done. NLP can help in the closure of information gaps without asking clinicians to make dramatic changes to their workflows.”
Also See: Atrius Health looks to analytics to reduce physician burnout
Going forward, Monsen says Atrius is looking to expand the use of NLP to support other care initiatives, including identifying social determinants of health and behavioral health conditions. “There is strong evidence to suggest that those are perhaps some of the greatest contributors to patient wellness, or the lack thereof,” he adds.
By implementing an NLP platform from Linguamatics, Atrius is able to mine critical clinical information hidden within unstructured text to improve Medicare Advantage reimbursement, streamline accountable care organization reporting and close gaps in care delivery.
“The majority of clinically useful information is unstructured text that includes physician notes and radiology reports, which impacts our patients and our practice,” says Craig Monsen, MD, Atrius Health’s medical director of analytics. “I’d estimate that probably 50 percent to 80 percent is unstructured data.”
According to Monsen, there are several of the healthcare organization’s workflows within which NLP is being used to identify and extract critical clinical data that exists as unstructured text.“We found for our heart failure patients that their heart pump function is not captured as structured data,” he says. “We’ve integrated natural language processing into our quality metrics reporting pipeline so that we can figure out who are these high-risk heart failure patients.”
Monsen adds that if an organization is trying to find that kind of information—such as the ejection fraction, which measures how well a heart is pumping—on a population level, they will “need to hire a small army” to manually extract it. However, he says the NLP tool enables Atrius to automate some of that process.
When it comes to closing care gaps, Atrius is also using the platform to identify patients with difficult-to-track conditions, such as pulmonary nodules, which are small growths in the lungs. “Most often, pulmonary nodules are benign, but in some cases, they’re an early sign of lung cancer,” according to Monsen. “Frequently, that information is only captured in free text, so that creates a situation where serious concerns can fall through the cracks.”
However, Atrius has invested in a series of “patient safety net” initiatives for several types of cancer designed to improve outcomes by leveraging NLP. “In the case of lung cancer, for example, the technology helps us to identify pulmonary nodules from radiology reports, making sure that we have followed up with those patients and have carried out the necessary interventions,” he notes.
Monsen also points out that NLP is ensuring the accuracy of clinical documentation workflows so that Atrius has sufficient resources to close gaps in care.
“We’re about 900 physicians across 30 practice locations, and we are heavily committed to value-based care, with something like 75 percent of our revenue coming from risk-based contracts,” says Monsen. “For us, it’s not just upside risk, but it’s also downside risk, so it’s important to make sure that we are resourced to provide the care, quality and safety nets.”
Physician burnout, a growing epidemic among provider organizations characterized by exhaustion, cynicism and feelings of ineffectiveness, has been linked to decreased quality of care and medical errors. However, Monsen believes that NLP will help Atrius alleviate the “data overload problem” and reduce physician burnout among its ranks because of data-entry tasks.
“The answer to closing information gaps can’t be to put everything into a structured field,” he observes. “Clinicians are already spending half their time in front of the electronic health record. This time is better spent with patients and having conversations that will contribute to their care rather than documenting what’s been done. NLP can help in the closure of information gaps without asking clinicians to make dramatic changes to their workflows.”
Also See: Atrius Health looks to analytics to reduce physician burnout
Going forward, Monsen says Atrius is looking to expand the use of NLP to support other care initiatives, including identifying social determinants of health and behavioral health conditions. “There is strong evidence to suggest that those are perhaps some of the greatest contributors to patient wellness, or the lack thereof,” he adds.
More for you
Loading data for hdm_tax_topic #better-outcomes...