Healthcare Organizations Must Focus On Business Value & Not Chase the Technology Pendulum

The use of natural language processing (NLP) in the clinical domain dates as far back as the 1960s, with important early work performed in the1970s and 1980s. While great technical advances were made during this period, clinical NLP failed to live up to initial expectations as accuracy rates were too low to use for much other than research or bulk analytics.

If you’ve been in the healthcare or technology spaces for any reasonable length of time, you’ve witnessed the tremendous growth spurred by investment as healthcare organizations pursue digital transformation. According to RockHealth, funding for U.S.-based digital health startups topped $29 billion in 2021 – adding to the $23 billion invested in 2019 and 2020 – and the number of new solutions in the market has never been higher. 

Such strong investment drives technology adoption but also generates trends that influence business strategies and buying decisions. As a result, large numbers of organizations follow each other in one direction, only to find later that a technology or operating model doesn’t deliver all the benefits anticipated. The swinging pendulum is not only costly but can delay realizing the real value that specialized solutions bring to the table. By looking back over the history of this phenomenon, healthcare organizations have the opportunity to take a more reasoned, long-term approach to technology investment and shorten the time to realize the value. 

Significant swings in recent years

One key example in healthcare is outsourcing, a market expected to be worth $488 billion by 2027. For years, healthcare organizations have been outsourcing a wide range of business processes, such as IT and revenue cycle management, as fast as possible to reduce costs and manage labor shortages. So, the pendulum started swinging hard from “in-sourcing” toward outsourcing those functions – including moving operations offshore – as much as they possibly could. 

But once the pendulum swings too far, it inevitably moves back in the opposite direction. Organizations began to realize that some functions are better kept in-house than outsourced or offshored. As a result, many brought outsourced functions back in-house, seeking an optimal balance.

Another pendulum swing I’ve witnessed in healthcare involves the adoption of electronic health record (EHR) systems, driven substantially by the Centers for Medicare and Medicaid Services (CMS) and the Health Information Technology for Economic and Clinical Health (HITECH) Act of 2009, designed to spur the adoption of health information technology. As of 2021, nearly 78% of office-based physicians and 96% of non-federal, acute care hospitals had adopted certified EHRs. I’ve talked with clients who invested in EHRs and expected these systems to be the answer to a broad range of needs. And then these organizations found out they missed the mark in many areas. 

To fill EHR performance gaps, many hospitals, health systems, and large provider groups began implementing point solutions, especially for patient-facing capabilities like scheduling, access, and payment. Now they had to manage a hodgepodge of solutions to get what they needed from a business standpoint, a workflow standpoint, and a patient experience standpoint. However, these disparate point solutions quickly became cumbersome to maintain, leading to data being siloed and patient journeys being frustrating. The challenge is to navigate the myriad options available and ensure that each solution offered is adding value.  

In response, EHR vendors started filling in the functional gaps on their platforms. Now the pendulum began to swing back toward provider organizations consolidating on their single EHR platform to address all the functionality they need. Even if some of that functionality isn’t as good as that offered by point solutions, the integrated nature of the platforms became the priority for provider organizations with limited IT staff. 

Eventually, though, we can expect the pendulum to swing back to a more balanced place where EHRs are complemented by other solutions. Business imperatives will cause provider organizations to integrate their EHR platforms with specialized solutions, such as customer relationship management (CRM) or enterprise resource planning (ERP) solutions. While more consolidated, they’ll still have to manage an ecosystem of technologies.

Current pendulum: Cloud solutions

A trend I’m seeing today is that healthcare organizations are moving applications they historically have run on-premises into the cloud, typically with Amazon, Google, or Microsoft. In one recent survey, 70% of IT professionals said their organizations had adopted cloud computing solutions. As adoption increases with the large cloud providers, organizations reason that if they’re going to spend that kind of money, they’re going to leverage it to the fullest. That’s what is now pushing the pendulum in the direction of consolidation. 

However, as with EHRs, the functionality offered by the Big 3 cloud platform vendors will feature a mix of mature and emerging technologies to support business-critical functions. In the latter case, healthcare organizations may have to supplement their cloud-based deployments with something that offers more specialized capabilities. 

One specific technology where there is a need for an upgrade over the vendor solution is natural language processing (NLP), a foundational Artificial Intelligence (AI)-enabling technology. Healthcare organizations are currently starting to use Amazon Comprehend or the Microsoft NLP engine because they’re part of those suites. 

But medical language isn’t everyday language; medical language has its own terminologies, definitions, and slang (SNOMED, RadLex, MEDCIN, ICD-10, etc.). That leaves a big gap between functionality and accuracy within large commercial NLP platforms that have not been trained on large enough pools of medical data. For example, if the NLP software is analyzing a report about a patient with a history of appendicitis, does the algorithm understand that the appendicitis happened in the past? Or does it conclude the patient has appendicitis now? If the report also includes language indicating the patient’s mother has chronic kidney disease, will the algorithm misattribute this condition to the patient?

The need to understand the context to produce accurate results is just one aspect that makes it difficult for healthcare organizations to standardize on industry-agnostic technology platforms that can’t accurately process unstructured medical text. Healthcare organizations need a more prescriptive tool than the bigger tech companies are equipped to provide today. 

Focus on value to be gained from medical NLP

The landscape has grown even more complex and confusing with the emergence of generative AI platforms such as ChatGPT and Bard. With its ability to create new content, generative AI is already having a transformative impact on publishing, research, education, and entertainment. 

Not surprisingly, people envision multiple use cases for generative AI in healthcare. However, these large language models (LLMs) still have reliability problems in terms of accuracy and “hallucinations,” a phenomenon in which an AI model, based on the training data it has been fed, generates outputs that aren’t possible in the real world. To have practical application in clinical contexts, LLMs must be trained and/or fine-tuned on large sets of annotated medical data to improve accuracy. 

Nonetheless, healthcare organizations shouldn’t wait for the pendulum to swing back, nor should they lose valuable time trying out NLP integrated into large cloud platforms or generative AI algorithms if they already know these tools aren’t good enough to meet their needs. Instead, these organizations should focus on business imperatives and the ability of specific solutions to fulfill those imperatives, while being mindful of how those tools will continue to mature. Waiting will only cost time and money.

The business imperative in healthcare is data and an organization’s ability to extract and leverage that data from medical documents without humans having to read every word. Data is what enables clinicians to provide evidence-based care for patients. Data is what allows healthcare organizations – providers and payers – to operate more efficiently and cost-effectively. And real-world data is what helps pharmaceutical companies to accelerate research into drugs and medical devices. 

Anticipating – and not being hit by – the pendulum

Healthcare organizations that chase the technology trend pendulum by partnering with large cloud vendors promising comprehensive solutions must ensure these platforms can deliver business value. While this can be approached on a case-by-case basis, NLP is a technology that requires a medical-grade understanding of clinical terminologies and can extract value from unstructured data buried in EHRs. Providers, patients, payers, and pharmaceuticals all can benefit from a purpose-built engine capable of understanding medical language and efficiently delivering useful information where it’s needed most.