Famous artificial intelligence (AI) errors such as a Google search recommendation to put glue on a pizza to keep the cheese from falling off may cause a smile. However, when AI is mentioned in conjunction with healthcare, the stakes become serious — quickly since ‘hallucinations’ are not an option in healthcare!
But make no mistake: AI is infiltrating healthcare, from both clinical and operational angles. Estimates show the global AI healthcare market growing by 1,600% during the decade, from $11.06 billion in 2021 to nearly $188 billion by 2030.
A 2023 survey indicates that two-thirds of health systems plan to launch AI pilots or projects over the next two years. Just 8% say they have no plans to explore AI.
Initial mainstream AI goals for healthcare include workflow improvements that can bring data together so clinicians can make quicker, evidence-based care decisions. But the promise of personalized healthcare remains a pipe dream in most circumstances. There are myriad challenges, but central to any discussion is the difficulty of matching the right person and knowing who’s who across care settings and disparate software systems.
When lives are on the line, clinicians cannot rely on 70% accuracy. They need absolute confidence that the information they see pertains to the correct patient — every time.
Overwhelming Support for AI Projects
Overall, people in the United States hold a positive opinion on the future of AI in healthcare. A 2023 survey showed that 51% believe AI will drive significant progress and innovation over the coming year, and over 60% have positive thoughts about the power of AI to efficiently diagnose health conditions. Further, two-thirds believe technology can reduce barriers to care, with 56% pointing to AI as the solution.
But survey participants are aware that significant challenges remain, including 80% who say there is a lack of evidence that AI improves health outcomes and 83% who point to the potential for mistakes.
Industry leaders agree about the challenges, according to a December survey of nearly 200 hospital, health system, and payer executives. Matching patient data across software systems remains a considerable concern, with 57% believing that errors in patient data matching will bring a healthcare crisis in the next five to 10 years.
Nearly half of respondents specifically point to the challenge of data residing in fragmented, siloed software systems. Only 47% say healthcare data is well-managed within their facilities, and even then, only in pockets of their organization.
The solution? More than 90% say that successful patient matching is very important or extremely important to their strategic initiatives, and about nine out of 10 agree that a master data management (MDM) or healthcare master data management (hMDM) system is required to bring the accuracy necessary for AI projects to flourish and find widespread adoption.
Accurate Patient Matching Key to AI Success
Healthcare systems possess an incredible amount of data that dates back decades. Even data from legacy systems can retain its value long after that surgery has been performed or that patient has been declared cancer-free.
Until recently, historical data was perceived as having limited use beyond medical release-of-information requests and regulatory compliance purposes. But as machine learning, natural language processing, deep learning, and other aspects of AI come to the fore, that data has gained new prominence.
To be truly useful, however, several data challenges must be solved. One challenge is the prevalence of duplicate records, a rate that commonly approaches 10% in single hospitals but can rise to over 20% in health systems with multiple facilities. A robust MDM-enabled Identity Data Management platform can take nearly all the guesswork out of patient matching, creating one-on-one relationships that bring data confidence.
Another issue is data cleansing and normalization. Over time, hospital systems likely have changed EHR vendors or have as many as 14-19 different EHRs across their healthcare system, imaging and lab systems, and other software. Data from one system type may or may not be compatible with another. Formatting standards also change, which creates silos of disparate data. The normalization process maps one set of data to another, creating relationships that make data actionable.
Data noise reduction can also be an obstacle to the success of AI projects. Even after matching patient records one-to-one, people change employers, move residences, get new phone numbers, and change permissions related to their data. For an AI-powered, automated test result notification to work, for example, which of several phone numbers should be used to call or text the patient? Reducing the amount of “noise” in data will improve its fidelity.
Solving Patient Identity Issues Can Tame the AI ‘Wild West’
With past technological advances, healthcare has understandably been cautious, waiting on mature, proven products and methodologies before adopting them. But when it comes to AI, healthcare is moving faster than ever.
The landscape can feel a bit like the AI Wild West, with so many healthcare organizations and vendors looking to add AI capabilities to their products and processes. But savvy providers know that AI isn’t a panacea. Significant issues remain to be solved — not least among them are the issues of identity resolution, data normalization, understanding relationships among data sets, and governance. AI-enabled solutions and workflows must also be developed responsibly, making data useful, safe, equitable, secure, and transparent.
Despite these challenges, the benefits of deploying AI in healthcare are numerous. These include freeing up clinician time through workflow improvements, diagnosing patients earlier, streamlining care delivery, and keeping people healthy. Healthcare leaders can establish a stronger foundation for AI success in a rapidly changing environment by addressing issues with robust and accurate patient matching.
About Avi Mukherjee
Avi Mukherjee is the Chief Product and Technology Officer (CPO) at Verato, a healthcare identity company. He was appointed to the role in April 2022 and leads the company’s product and engineering teams to develop the industry’s only healthcare master data management (hMDM) identity platform. Mukherjee’s responsibilities include product development, data science, and solutions across the Verato product portfolio.
Mukherjee has a background in building and scaling cloud companies for healthcare. Before joining Verato, he worked at Verily, a Google Life Sciences technology company, where he led healthcare customer solutions.
About Andy Dé
Andy Dé is the Chief Marketing Officer (CMO) of Verato, a healthcare identity company. He joined Verato in March 2024 and leads the company’s global go-to-market (GTM) strategy, planning, and execution for its hMDM platform and solutions. Dé’s role is to drive the growth and market expansion of Verato’s healthcare identity solutions. He is also a member of Verato’s Executive Leadership Team and reports to CEO Clay Ritchey.
Before joining Verato, Dé has worked at companies including GE Healthcare Systems, Alteryx, MedeAnalytics, SAP, and Sabre. He has also been inducted into the Forbes Communication Council as an Industry Thought Leader and was a HIMSS Social Media Ambassador in 2018 and 2019