Artificial Intelligence and Compliance
“An Interview With Frank Cohen”
by Sean Weiss, Partner & VP of Compliance
Artificial Intelligence (AI) in healthcare has become the sexy new term and what everyone is making so much noise about. So, what’s the big deal and how does AI work in the word of healthcare compliance? February of this year, President Trump signed an executive order titled “Maintaining American Leadership in Artificial Intelligence” which can be found here and is known as the American AI Initiative. According to Michael Kratsios who serves as deputy assistant to the president for technology policy, the executive order is designed to “prepar[e] America’s workforce for [the] jobs of today and tomorrow.”
The question I have been getting asked by clients is, “How does this impact our practice/organization and what do we need to do to prepare for the “Future””? Most working in the field of AI will tell you it specifically impacts healthcare. In healthcare, we use AI for things like improving quality of care in addition to driving down the costs associated with the delivery of patient care. Some of the key areas we see AI impacting healthcare is in Electronic Medical Records whereby companies are developing programs that analyze “unstructured” patient medical records via AI tools such as machine learning through algorithms that have an ability to learn from data without relying on rules-based programming. Other companies like EPIC are creating natural language processing, which allows computers to understand and interpret human language, which in turn allows them to deliver meaningful and searchable data. This data can be attached to things like patient diagnosis, treatment plans created by clinicians, and so much more. Other areas where we find AI is in the capabilities of smartphones that now have finger sensors that when using a specific app allow you to monitor your heartrate, saturation levels of O2, etc. We also have tons of wearable devices that offer a whole host of options for creating a healthier lifestyle. However, we see some of the greatest advances with AI in Research and Development in the field of Diagnostics.
While all of this is critical, in my humble opinion the impact of AI on compliance in how we think about it with coding, billing, auditing, sampling, extrapolation, etc. is minimal. So, I set out to learn about this and in doing so, I reached out to my very good friend Frank Cohen as I knew he would be tied into this world. The lesson I got from Frank, as with most things, was quite interesting and not exactly what I was expecting to hear. Frank Cohen uses the term AI a bit differently; he refers to it as Augmented Intelligence.
Augmented Intelligence, as he refers to it, is a paradigm that is designed to assist, or “augment” the work that healthcare administrators and staff do rather than to replace the person. If engaged properly, non-clinical AI efforts will help to reduce workload in areas that are not as productive and increase productivity in areas that are more productive. For example, it would take an individual auditor some 1,250 hours to audit 10,000 charts that mixed E/M with procedures; a daunting and expensive task, to say the least. And when you are done, you may have created a repayment that you didn’t want. Using predictive analytics, which incorporates AI technology, you could review the claims for those 10,000 charts in a matter of seconds, spitting out those that would be most likely to be billed in error or subject to an external audit, depending on the purpose of the algorithm. Take the latter case – If I could tell you which procedure codes or modifiers for which physicians were most likely to be audited, you could pull a sample and proactively audit them yourself. This approach to risk-based auditing is significantly more efficient and accurate than, say, random probe auditing and a whole lot more cost effective than the 1,250 hours required to audit all the claims.
AI also incorporates the technique of machine learning, which is a process whereby the computer algorithm “learns” or tweaks itself based on a feedback loop. Let’s say that out of those 10,000 claims, the algorithm picked 135 that, based on its calculations, are most likely to have been billed in error. The auditor reviews each of these and then enters their results into the feedback loop. The algorithm then analyzes these results to validate its accuracy and its classification process. Let’s say that of those 135, 100 were in fact billed in error but 35 were not. In the context of machine learning, the algorithm will try to figure out the difference between the 100 that were properly identified and the 35 that were not, improving its classification capabilities.
While AI has had a history of benefits on the clinical side of healthcare, its reach at this point has been mostly limited with respect to non-clinical areas of healthcare which could include administration, management, Human Resources, Revenue Cycle Management, and perhaps most beneficial in the areas of risk management and regulatory compliance. In the World according to Frank Cohen, at least for the foreseeable future, “AI simply isn’t smart enough to replace humans; particularly within the healthcare industry” and if you don’t believe us, click this link for an AI fail! Funny Dog Video – Roomba