About $3.6 trillion is spent every year in the United States on healthcare. Between $600 billion and $1.9 trillion of that is wasted. Technology will not completely solve this staggering inefficiency, but it will be a crucial part of the solution. As healthcare organizations implement new technologies, health outcomes will improve, costs will decrease, and patients will have a better overall experience of care.
The healthcare industry, ever resistant to change, slowly chugs along in its digital transformation. The most visible aspect of this transformation—and the one that patients are likely more familiar with—is the new modes of care delivery: telehealth saw a massive increase during the COVID-19 pandemic, appointments can (sometimes) be scheduled without lengthy phone calls, insurance claims can be managed through online portals and so on.
However, there is a parallel revolution happening under the hood, so to speak. Organizations within the healthcare ecosystem gingerly take their first steps toward digitizing and streamlining administration—for example, automating common workflows and deriving actionable insights from the massive amounts of healthcare data.
In particular, the increasing deployment of machine learning (ML) and especially natural language processing have drastically reduced the time it takes to accomplish tasks common in healthcare administration. Below are two typical examples of ML’s impact on healthcare.
Computers can now analyze unstructured data at a speed and scale relevant for business. The processing of unstructured text—natural language processing, or NLP—finds numerous use cases within healthcare, as much health data is unstructured. Computers can analyze documents such as PDFs or case notes in an electronic health record. They can look for common patterns or translate human-readable text into machine-readable codes.
To give a specific example: a health insurance plan sends an SBC (Summary of Benefits and Coverage) to a new patient. That SBC contains information about the patients’ benefits written in English—and so it cannot be understood (readily) by a computer. Normally, a human would read the SBC and manually translate each individual sentence into machine-readable code—in this case an EDI X12 code (see the above image). However, an ordinary payer would have accumulated hundreds of thousands of SBCs over the years. It could take years to translate them into EDI codes. And, even once the task is done, it is impossible to efficiently verify the results.
With natural language processing, SBCs can be translated into machine-readable code quickly: given a set of correctly labeled SBCs, an NLP platform learns the same rules a human knows but is able to apply them much faster. The initial training may take a few days or a week. And the verification—which still requires human input to correct mistakes made by the algorithm—can still be completed in a matter of months as opposed to years.
Machine learning algorithms also excel at automating tasks. Claims processing—a time and labor-intensive task for health insurers—is a perfect candidate for a workflow that could be improved with machine learning. While claims processing is automated to a certain extent already, the rules are extremely complex. In fact, the rules are so complex that the majority of claims are flagged as requiring human intervention. An ML-enabled claims processing platform could optimize this process.
First, it would learn the already existing rules by analyzing historical data. Through this training, it would learn which cases are easy to decide and which require intervention. For the latter group, the platform would be able to determine those cases for which human intervention would likely result in a positive outcome for the payer, i.e. a correct decision about the claim with minimal time invested by staff. By identifying the highest impact cases with high probability, the platform optimizes the time spent by staff.
An especially advanced algorithm could even suggest possible options for claims resolution because it knows how similar cases in the past have been resolved. And the system would learn from each new case, becoming more accurate over time. All told, an intelligent claims processing platform could save a significant amount of time by directing staff towards the highest impact cases and reducing the time spent per case.
These are just two areas where machine learning has benefitted healthcare. Further areas of impact include pharmacy benefits administration, prior authorization automation, payment integrity, fraud detection, chatbots, and EHR integration. Any task which can be reduced to the repetitive application of complex rules could be improved by machine learning.
With natural language processing, SBCs can be translated into machine-readable code quickly: given a set of correctly labeled SBCs, an NLP platform learns the same rules a human knows but is able to apply them much faster. The initial training may take a few days or a week. And the verification—which still requires human input to correct mistakes made by the algorithm—can still be completed in a matter of months as opposed to years.
All these innovations aim at the same ultimate goal: that more of the money spent on healthcare actually goes towards making us healthier. The age of ML-enabled healthcare is coming. Is your organization ready?
Recommendations:
Most healthcare organizations have at least already begun their attempt to leverage the power of machine learning and natural language processing. Quantum Gears encourages all organizations within the healthcare ecosystem to:
- Assess the current status of machine learning implementation and knowledge.
- Determine areas where machine learning could provide value and educate decision makers on possible solutions and vendors.
- Engage healthcare experts—such as Quantum Gears—to guide your adoption of a strategy that will help you stay competitive in the age of ML-enabled healthcare.
About Quantum Gears
Forum Systems and its subsidiary, Quantum Gears, are leading the Enterprise GenAI revolution. Patent-pending products—like SecureGPT™, ContractsGPT™, BenefitsGPT™, and Forum Sentry—mitigate the unpredictable nature of LLMs through integration with corporate APIs, ensuring LLM output is truthful and accurate. Used by some of the largest global companies for building intelligent business workflows, Forum’s suite of products provides unique, industry-leading solutions that allow enterprises to reinvent themselves with GenAI.