Electronic Health Reporter

By Chetan Parikh, CEO, ezDI.

The financial implications of the COVID-19 pandemic are generally considered to be secondary to the health crisis, but there’s more overlap between them than you might think. A viewpoint published in the Journal of the American Medical Association cites the challenge of running a financially solvent hospital in the face of staffing shortages, overwhelming demand, and a pause in the outpatient and elective procedures that often make up the bulk of a hospital’s revenue.

While the situation looked dire when the article was published in May 2020, the health crisis across most of the country has only intensified. For health systems themselves to survive the pandemic, they’ll need to take a hard look at revenue cycle management best practices and implement new efficiencies whenever possible.

A proven way to inject much-needed efficiency is to rely on technologies such as artificial intelligence to automate tedious manual processes and improve the speed and accuracy of humans as they perform complicated tasks. One of those complex areas is clinical documentation.

Clinical documentation has always been a bit of a black box, with every provider documenting patients’ clinical issues differently and using shorthand that only they understand. This information is critically important, but it’s all stored in the form of unstructured data, and 56% of healthcare professionals believe it’s obstructing clinical workflow optimization.

Now that almost all clinical documentation lives in an electronic health record, the goal of standardization is pushing facilities to evaluate the documentation that accompanies each encounter. Payers are adding to the pressure, denying reimbursement to providers that fail to adequately prove the necessity of services. New payment models that incorporate risk scoring have also made it critical for providers to document underlying issues during visits, creating an additional layer of complexity — and an additional opportunity for solutions that can facilitate the process.

The AI approach

Artificial intelligence is a broad category that encompasses machine learning, natural language processing, computer vision and, to some extent, robotic process automation. Each of these subcategories requires its own implementation road map, including reference content to teach an algorithm how to process raw data and turn it into useful information. NLP, for example, can enable computers to read and process written information, such as that contained in a clinical knowledge graph.

When properly trained, an NLP solution can accurately determine the relationship between diagnosis codes, charge codes, prescription drug names, body parts and syntax. Once baseline capabilities are established, the content can be updated with the usage of abbreviations, reference language and other nuances that might be specific to a facility’s EHR template or a physician’s notation preferences.

AI has all the raw capabilities described above, but additional refinement and integration are key. There’s a big gap between being able to read something and being able to understand it. Clinical NLP powered by knowledge graphs is very good at reading, but there’s a long way to go before the technology can truly understand documentation as well as a physician, nurse or coder would. Adding layers of analysis could be the answer, and we’re within a few years of AI being able to evaluate the patient and the case in context to determine whether certain diagnosis codes really make sense.

Augmenting existing efforts

Revenue cycle departments are fighting to protect and grow revenue, but the personnel needed to staff clinical documentation improvement programs are hard to come by. By implementing AI-powered CDI tools, charts can be reviewed by a machine before being coded and billed, enforcing the consistency needed to capture risk scores, prevent denials and survive audits. CDI leaders can also get the job done with fewer people: By reducing the time physicians must spend looking at screens instead of interacting with patients, they can help reduce burnout and depression currently experienced by more than 40% of healthcare workers.

So far, the data soundly backs up CDI software implementation, and Black Book Market Research found that 90% of hospitals were able to boost revenue by $1.5 million or more after putting CDI solutions in place. They accomplish this feat by driving efficiency in the documentation process, reducing coding errors to minimize revenue loss and identifying missed opportunities to maximize reimbursement.

Even before the coronavirus crisis, revenue cycle management was critical to allowing providers to fund the services necessary to provide the best patient care. Now, as physicians struggle to keep pace with increased case volumes and complex reimbursement schemes from payers, the capacity of CDI programs to preserve documentation quality while minimizing manual processes is even more important.

“Normal” might still be a long way off, but for hospitals hoping to make it to the other side of the pandemic, AI-enabled CDI might be the most readily available vaccine.