How natural language processing will help achieve a payback on EHRs
With almost all healthcare organizations now using electronic records, computer algorithms will increasingly mine unstructured data to help improve care.
One of the most talked-about buzzwords in healthcare is natural language processing. While NLP has carried some industry weight for a while, many think the practical applications and impact of NLP haven’t lived up to the hype.
So why are all of the organizations we are talking with about risk adjustment asking about natural language processing and wanting this capability? The answer: If done correctly, NLP can offload burden from physicians, enhance accurate risk capture and impact care delivery in a big way.
Natural language processing is the process by which computer algorithms pull out key elements and mine meaning from large amounts of unstructured, hand-typed or dictated notes within an electronic health record. Unlike text dictation platforms, NLP processes unstructured text and converts it to structured data that can be used to drive analytics and clinical action.
Although it looks and sounds great on paper, natural language processing has fallen short of expectations. Many organizations currently use voice or text translators, while some are starting to use computer-assisted documentation. However, this technology can be slow, and providers often find the workflows clunky.
There are several reasons for NLP’s slow adoption. On one hand, NLP technology was subpar for a long time, and clinical technologies weren’t designed to effectively integrate with NLP. On the other, the initial exuberance of large-scale EHR implementations driven by federal funding in the early to mid-2000s led many people to believe EHRs would be able to capture all of the discrete data and more of the unstructured data than they do.
Currently, many organizations still have so much unstructured data in notes in their EHRs that it is difficult to find and use the data effectively. The irony is that providers often don’t even look at the notes. Research has found that the majority of data in the EHR is unstructured but many notes are never read by care team members, according to a study published in the Journal of the American Medical Informatics Association in 2011.
There are several benefits to adopting natural language processing, most notably the ability to deliver better—and, at times, life-saving—care to patients. When used correctly and appropriately, natural language processing can help automate processes and make workflows more efficient and encouraging and prompt physicians and other care team members to actually use unstructured data to provide better care to patients.
In decision support, NLP can be leveraged to capture potential care gaps. For example, a physician is given a radiology report that notes a lung nodule or thoracic-aortic aneurysm supported in the text of the report. These two life-threatening conditions may go unnoticed and undocumented in the non-discrete data in the EHR; however, NLP analytics convert these flagged conditions to suspect diagnosis codes, which trigger decision support to alert for appropriate care interventions.
For perspective, NLP flags the specific radiology reports, out of the thousands that a hospital produces, that need extra attention and follow-up. There is so much unstructured data within these thousands of reports that an organization cannot capture all of it without NLP unless it hired clinical staff to review annually, a prohibitively expensive proposition.
NLP also helps organizations capture risk scores more accurately. Our clients increased their HCC (Hierarchical Condition Categories) capture by 15 percent on average when using risk adjustment point-of-care technology powered by natural language processing.
Some organizations are also using NLP to automate compliance processes. NLP analytics review provider documentation and flags those that need clinical coder review, identifying charts that may have compliance concerns in need of human review. It is not financially feasible for human staff to review every provider note to make sure HCC condition codes are justified and supported by the documentation.
With risk-score coding compliance being a major focus of the Centers for Medicare & Medicaid Services (CMS) and Congress, using NLP effectively can automate processes to keep your doctors out of hot water by protecting them from fraud and abuse claims, compliance investigations, etc.
As we increase the burden of quality scoring, risk adjustment payment mechanisms and compliance reviews on providers, we’re asking them to do an increasing amount of work. To best support providers, we need to pay attention to their hearts, wallets, watches and minds. Natural language processing helps providers deliver better care to patients, fulfilling the calling in their hearts; helps get them reimbursed appropriately; and makes their workflows more efficient and scientifically automated, respecting their time and honoring their skills.
Increasingly, the industry will continue to see NLP evolve from a capability that was once overrated and somewhat “magical thinking” to what many organizations now see as an essential tool to support providers in their pursuit of quality care and appropriate risk score capture.
So why are all of the organizations we are talking with about risk adjustment asking about natural language processing and wanting this capability? The answer: If done correctly, NLP can offload burden from physicians, enhance accurate risk capture and impact care delivery in a big way.
Natural language processing is the process by which computer algorithms pull out key elements and mine meaning from large amounts of unstructured, hand-typed or dictated notes within an electronic health record. Unlike text dictation platforms, NLP processes unstructured text and converts it to structured data that can be used to drive analytics and clinical action.
Although it looks and sounds great on paper, natural language processing has fallen short of expectations. Many organizations currently use voice or text translators, while some are starting to use computer-assisted documentation. However, this technology can be slow, and providers often find the workflows clunky.
There are several reasons for NLP’s slow adoption. On one hand, NLP technology was subpar for a long time, and clinical technologies weren’t designed to effectively integrate with NLP. On the other, the initial exuberance of large-scale EHR implementations driven by federal funding in the early to mid-2000s led many people to believe EHRs would be able to capture all of the discrete data and more of the unstructured data than they do.Currently, many organizations still have so much unstructured data in notes in their EHRs that it is difficult to find and use the data effectively. The irony is that providers often don’t even look at the notes. Research has found that the majority of data in the EHR is unstructured but many notes are never read by care team members, according to a study published in the Journal of the American Medical Informatics Association in 2011.
There are several benefits to adopting natural language processing, most notably the ability to deliver better—and, at times, life-saving—care to patients. When used correctly and appropriately, natural language processing can help automate processes and make workflows more efficient and encouraging and prompt physicians and other care team members to actually use unstructured data to provide better care to patients.
In decision support, NLP can be leveraged to capture potential care gaps. For example, a physician is given a radiology report that notes a lung nodule or thoracic-aortic aneurysm supported in the text of the report. These two life-threatening conditions may go unnoticed and undocumented in the non-discrete data in the EHR; however, NLP analytics convert these flagged conditions to suspect diagnosis codes, which trigger decision support to alert for appropriate care interventions.
For perspective, NLP flags the specific radiology reports, out of the thousands that a hospital produces, that need extra attention and follow-up. There is so much unstructured data within these thousands of reports that an organization cannot capture all of it without NLP unless it hired clinical staff to review annually, a prohibitively expensive proposition.
NLP also helps organizations capture risk scores more accurately. Our clients increased their HCC (Hierarchical Condition Categories) capture by 15 percent on average when using risk adjustment point-of-care technology powered by natural language processing.
Some organizations are also using NLP to automate compliance processes. NLP analytics review provider documentation and flags those that need clinical coder review, identifying charts that may have compliance concerns in need of human review. It is not financially feasible for human staff to review every provider note to make sure HCC condition codes are justified and supported by the documentation.
With risk-score coding compliance being a major focus of the Centers for Medicare & Medicaid Services (CMS) and Congress, using NLP effectively can automate processes to keep your doctors out of hot water by protecting them from fraud and abuse claims, compliance investigations, etc.
As we increase the burden of quality scoring, risk adjustment payment mechanisms and compliance reviews on providers, we’re asking them to do an increasing amount of work. To best support providers, we need to pay attention to their hearts, wallets, watches and minds. Natural language processing helps providers deliver better care to patients, fulfilling the calling in their hearts; helps get them reimbursed appropriately; and makes their workflows more efficient and scientifically automated, respecting their time and honoring their skills.
Increasingly, the industry will continue to see NLP evolve from a capability that was once overrated and somewhat “magical thinking” to what many organizations now see as an essential tool to support providers in their pursuit of quality care and appropriate risk score capture.
More for you
Loading data for hdm_tax_topic #better-outcomes...