{"id":15729,"date":"2022-02-21T22:43:09","date_gmt":"2022-02-21T22:43:09","guid":{"rendered":"https:\/\/mailinvest.blog\/index.php\/2022\/02\/21\/how-nlp-engines-are-fulfilling-ais-promise-in-healthcare\/"},"modified":"2022-02-21T22:43:09","modified_gmt":"2022-02-21T22:43:09","slug":"how-nlp-engines-are-fulfilling-ais-promise-in-healthcare","status":"publish","type":"post","link":"https:\/\/mailinvest.blog\/index.php\/2022\/02\/21\/how-nlp-engines-are-fulfilling-ais-promise-in-healthcare\/","title":{"rendered":"How NLP Engines are Fulfilling AI\u2019s Promise in Healthcare"},"content":{"rendered":"

<\/a>\n<\/p>\n

\n

In an ideal world, the healthcare industry would be the perfect arena for unleashing the\u00a0transformative potential<\/a>\u00a0of artificial intelligence.<\/p>\n

Sophisticated AI models require vast amounts of data to learn from, and healthcare data represents\u00a0nearly one-third of all data in existence. Nevertheless, all that information could fuel the development of algorithms capable of detecting and diagnosing disease, prescribing preventive treatments, and fulfilling myriad other functions with far greater speed and accuracy than even the most highly skilled team of doctors. With this kind of promise, that power is undoubtedly worth harnessing.<\/p>\n

AI\u2019s Uses in Healthcare<\/h2>\n

In some cases, AI\u2019s potential<\/a>\u00a0has already been realized. For instance, radiologists equipped with AI-powered tools have been able to review and translate mammograms\u00a030 times faster<\/a> than they otherwise would, with 99% accuracy. So naturally, these results have fueled tremendous excitement among stakeholders throughout the healthcare ecosystem.<\/p>\n

However, the present reality is that most healthcare organizations don\u2019t have the infrastructure or data management capabilities needed to sufficiently train algorithms that can be reliably deployed in a clinical setting.<\/h3>\n

Moreover, the complex,\u00a0fragmented nature of the industry<\/a>\u00a0and misaligned incentives that support the status quo have further\u00a0hampered technological innovation<\/a> and created barriers that simply don\u2019t exist in other sectors. Yet the potential for transformation remains.<\/p>\n

Searching for Solutions<\/h2>\n

Although there are still massive obstacles to deploying AI in clinical settings, the hurdles aren\u2019t as significant for other use cases. One of the most promising of these is in the area of medical coding, particularly when it comes to risk adjustment. However, as the quantity of healthcare data continues to grow exponentially, medical coders will remain in high demand\u2014and in short supply.<\/p>\n

Fortunately, AI tools equipped with\u00a0natural language processing capabilities can help relieve some of the coding burdens, enabling health plans to overcome the widening talent gap while also improving coding speed and accuracy.<\/p>\n

At present, many organizations relying solely on (often understaffed) human teams tend to view speed and accuracy as inversely correlated. However, a good AI\/NLP engine won\u2019t require them to sacrifice one for the other.<\/p>\n

Speed up Coding<\/h3>\n

Using AI to speed up medical coding gives human coders more time<\/a> to fix costly mistakes, increasing accuracy and potentially allowing facilities to direct more resources toward where they\u2019re needed most.<\/strong><\/p>\n

NLP has advanced significantly over the past five years. And coding tools that harness its technology increase the accuracy of codes and solve the problem of overcoding by deleting the codes that shouldn\u2019t be there.<\/p>\n

On the surface, this type of application might seem relatively mundane, but its implications could indeed be revolutionary for healthcare plans, providers, and patients.<\/p>\n

The Need for Speed<\/h2>\n

The slow pace of medical coding has long been a pain point for health plans. It can take weeks, months, or sometimes even years to manually review and process chart notes to identify pertinent information. In addition, patients with ongoing or chronic conditions need treatment while health plans and providers work to identify the most accurate hierarchical condition category (HCC) codes.<\/p>\n

Many plans use concurrent, and retrospective chart reviews to address these conditions in the same calendar year. Still, as healthcare data volume, variation, and complexity<\/a> increase, better solutions are sorely needed. This is why more organizations are turning to NLP to quickly search, analyze, and interpret massive patient data sets<\/a>.<\/p>\n

Easing Up Workflows<\/h2>\n

AI\/NLP engines turn the expertise of the medical coder into a\u00a0more useful asset<\/a>\u00a0for healthcare plans by enabling these professionals to focus on more meaningful work (such as spotting unusual trends and patterns in data).<\/p>\n

As coders gain familiarity with AI-powered tools, their coding speed will continue to increase over time, thus reducing the cost of coding a chart and making these human employees increasingly valuable to their organizations.<\/p>\n

No human being can remember all hierarchical condition category codes (which have expanded rapidly over the past few decades). Still, with the help of AI, human teams can more quickly match a code to its corresponding condition.<\/p>\n

The Cost of Complacence<\/h2>\n

In contrast, organizations that continue to rely on traditional coding tools will experience more of the same obstacles that have plagued the sector for years. Given that more than 9,500 ICD-10 codes map to roughly 80 HCCs, medical coders that can\u2019t rely on NLP-enhanced search capabilities are much more likely to miss codes.<\/p>\n

NLP medical records using machine learning algorithms can even uncover diseases that might not have been previously coded\u2014an essential feature for making HCC disease discoveries.<\/p>\n

Moreover, manually allocating charts to coders can be expensive from a management perspective and makes it challenging to scale projects to large teams.<\/p>\n

It also puts coders at a severe disadvantage: Without NLP prioritization, they must manually comb through charts, often resulting in fatigue and burnout. Besides this, given that traditional tools typically offer only limited reporting capabilities, teams using them will continue to struggle with project oversight and performance management.<\/p>\n

Harness advanced algorithms<\/h3>\n

Because they can harness advanced algorithms, machine learning and NLP engines can help teams gather relevant insights and concepts from previously considered data buried in text form.<\/p>\n

By giving voice to unstructured data, these tools can better understand data quality, continual process improvement, increased risk adjustment accuracy, and better patient outcomes.<\/p>\n

A Word of Caution<\/h2>\n

Like any emerging technology, NLP engines<\/a> aren\u2019t a panacea. Without sufficient training, they might miss valuable codes, and the reality is that many NLP coding tools are trained on a limited number of medical records. As a result, they tend to struggle with less-common conditions and electronic medical record formats.<\/p>\n

Furthermore, most NLP tools still feature a user interface that simply isn\u2019t optimized to meet coders\u2019 needs. This is no minor issue, as a poorly designed UI can cause coders to lose context around clinical notes and make additional mistakes.<\/p>\n

The effectiveness of these tools also largely depends on the workflows designed around them.<\/b><\/h3>\n

NLP coding tools built with a one-size-fits-all approach usually fail to support workflows that aren\u2019t purely associated with risk adjustment. For example, most don\u2019t provide features that allow for the auditing of vendor work. As a result, NLP tools that prioritize speed over accuracy will miss codes and expose organizations to compliance issues and the risk of lost revenue.<\/p>\n

According to a study from CMS,\u00a080% of improper payments\u00a0stem from inaccurate coding; these improper payments can lead to charges of fraud that could cost three times as much as a false claim, making accuracy a financial imperative for healthcare organizations.<\/p>\n

Choosing speed tends to result in the submission of improper codes, which has a domino effect as lower payments or risk adjustment data validation audits ultimately encumber the health plan, and thus its ability to support patient health.<\/p>\n

Powering Progress<\/h2>\n

When well-designed and properly implemented, AI\/NLP engines can drive the speed and accuracy of medical coding. They allow coders to achieve\u00a0significantly faster coding speeds while reducing instances of human error and the multitude of risks that accompany it.<\/p>\n

By making the process of extracting information from physician notes and assigning medical codes far more efficient, these tools\u00a0speed up billing<\/a>\u00a0and ensure that the organizations deploying them can continue to support patient health well into the future.<\/p>\n

Image Credit: provided by the author; Thank you!<\/strong><\/em><\/p>\n

\n
\n
\n\"\"<\/div>\n
\n

Sishir Reddy<\/a><\/h3>\n
\nco-CEO at Episource
\n<\/h5>\n

\nSishir oversees the global organization of 3,500+ people, motivating and leading teams around the company’s vision and goals. By leveraging his extensive knowledge and experience in the healthcare industry, including revenue cycle management and medical coding services, he has helped make Episource a leading provider of payer-focused solutions. Sishir received a BA in Chemistry from Claremont McKenna College and an MIS from Claremont Graduate University. <\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n