Electronic Health Reporter

By Ray D’Onofrio, principal data architect, SPR.

Could Amazon’s personal assistant, Alexa, predict when you will be sick? Or, if Siri is the first to know you have COVID-19.  It’s not as far-fetched as it sounds. Artificial intelligence (AI) is already transforming healthcare in a number of ways and by combining today’s technology with available data from a range of sources (e.g., electronic health records (EHR), personal buying habits, etc.), we can achieve even more important breakthroughs.

There is limitless potential in the way AI and machine learning (ML) can better equip healthcare professionals for their jobs. Instead of replacing our doctors and nurses, the technology works alongside their skills and expertise to elevate their patient care overall. This pairing of human and machine can create an efficient workplace for clinicians to deliver even more quality care to patients at scale.

The expanded use of AI and ML in healthcare hinges on several factors, including data ownership concerns and the ethical implications of providing healthcare data to technology companies like Google and Amazon. But with the right approach, it’s possible to leverage AI and ML to achieve better medical outcomes.

We don’t know how to effectively use AI yet

We’re familiar with the potential applications of AI in healthcare. For example, we know that in many cases, AI is better equipped to detect skin cancer than a human doctor. In addition to improving diagnoses, AI also holds promise in the development of customized treatment plans and giving patients greater control over their conditions. When AI and clinicians work together — such as when Harvard combined analysis from human pathologists with AI to identify breast cancer cells — it can produce even more effective results.

While the potential uses for ML and AI in the healthcare space are vast, these technologies are only as effective as the data that is available to them. If we could access all of the patient’s available data, — from their electronic health record, to their data stored by Google, Amazon and other technology providers — we would have a more comprehensive view of the patient’s health and significantly improve their experience using today’s technology.

In other words, the potential for ML and AI in healthcare is not limited by the technology; it’s limited by a lack of centralization and standardization of the patient’s data. For AI and ML to succeed as catalysts for an even better patient experience, we’ll need to take a different approach — one that unlocks AI’s true potential by re-envisioning how we source, store and use all of a patient’s healthcare data.

Using machine learning to improve the patient experience

To optimize ML and AI in healthcare, several factors must operate in unison — a difficult, but achievable goal if patients, clinicians, healthcare organizations and technology providers can coalesce to overcome key challenges. These elements include:

• Technology: All of the technological infrastructure required to expand our use of ML, AI and personal health information (PHI) is already in place, such as patient records, personal health devices, mobile applications and more. Additionally, provisions of the 21st Century Cures Act sets standards for interoperability and data sharing for health information network such as FHIR and the Trusted Exchange Framework and Common Agreement (TEFCA), so the framework is there already.

• Ethical concerns: Ethical considerations present a potentially significant roadblock to the expanded use of PHI. Although there are clear benefits to enabling PHI access to technology providers like Amazon or Google, important questions remain about how these organizations will use individuals’ medical information. For example, will Amazon use your health information to improve the patient experience, or will the tech giant just use your data to sell you more products? If your wearable detects you may have COVID-19 during the pandemic, should health officials, your physician be contacted?  Before we can unlock the total potential of ML and AI in healthcare, the industry itself will need to contend with and come to a final conclusion about how to deal with these gray ethical areas.

• Ownership of data: Serious debate exists about the ownership of individuals’ PHI. And although the technology framework exists already, there is legal uncertainty about interoperability and data sharing issues, with major players like Epic pushing back on HHS interoperability rules due to privacy concerns. To make the next stage of development a reality, stakeholders must adopt a stance similar to Apple’s philosophy on healthcare data management — healthcare data belongs to the individual and the individual controls its use.

In addition to addressing ethical and data ownership concerns, centralizing and standardizing the use of ML and AI in healthcare will require a secure technical framework for sharing data, resources, analyses and more. Although blockchain technology is a great decentralized framework for solving technical problems, it’s not designed for healthcare and only deals with a few pieces of data compared to the myriad data points associated with PHI.

But that’s not to say that blockchain or another emerging technology couldn’t be the solution to optimizing AI and ML uses in healthcare. Like a puzzle, we have all the “pieces,” or technology, at our fingertips — we just need to figure out how to fit them together for a clear picture of the patient’s overall health that can be analyzed with AI and ML.

Ultimately, none of these challenges are insurmountable. ML and AI hold nearly unlimited potential to improve the patient experience and create a more efficient, more effective healthcare system for clinicians to operate in — the only questions are when and how we can give the technology access to the data it requires to transform healthcare as we know it.