By Jaimes Blunt

Healthcare leaders have high expectations that AI will pay off in improved patient outcomes and customer experiences relatively soon – even as most respondents to a recent survey on AI in healthcare say their existing AI tools are flawed.

The survey, conducted online in late 2024 by emtelligent and including feedback from 250 healthcare leaders using or considering using AI tools to extract insights from unstructured clinical data, shows that nearly half (45%) of healthcare organizations have been using AI or natural language processing (NLP) technologies for three or more years. However, fewer than four in 10 respondents (38%) said they currently have a “perfect” solution in place for extracting insights from unstructured data (such as clinician notes, PDFs, and faxes), which comprises 80% of all electronic health records (EHR) data. This suggests general dissatisfaction with existing tools for extracting unstructured data.

Nonetheless, nearly all respondents (95%) said they expect to see “measurable outcomes or improvements from AI or NLP implementations,” while slightly more than half (52%) anticipate a similar return on investment (ROI) in less than a year.

Such widespread optimism reflects the justifiable belief that AI can have a transformative effect across the healthcare continuum based on the technology’s ability to convert unstructured data into actionable clinical insights, medical research breakthroughs, and greater operational efficiency. In particular, respondents see significant opportunities in using unstructured data for predictive modeling and to drive innovation.

Survey respondents who are familiar with or already deploying AI to extract unstructured data cited multiple top use cases, depending on their organizations’ role in the healthcare ecosystem. For health systems, patient summaries (79%), care management (64%), quality improvement (64%) and data analysis (64%) are top priorities. Other top use cases include:

• Data abstraction (59%)
• Thorough clinical data access (54%)
• Complex data analytics (49%)

Organizational barriers and the path to AI maturity
The barriers to implementing AI solutions most frequently mentioned by survey respondents are data privacy and security concerns (39%) in addition to skepticism regarding outcomes and value (33%). The latter percentage indicates AI champions have their work cut out for them in selling internal decision-makers on how to leverage AI and unstructured data to reduce operating costs, improve care delivery, and drive new innovation.

Overcoming these organizational concerns and recognizing the opportunity for value is the essential first stage of AI maturity for any healthcare organization. Therefore, a critical initial step in gaining internal buy-in for AI healthcare deployments is ensuring that decision-makers are aware of the untapped value of unstructured EHR data. According to survey respondents, the top benefits sought from leveraging unstructured data are:

• Improved patient outcomes (39%)
• Improved customer service (33%)
• Improved customer experience (31%)
• Increased efficiency/productivity (28%)
• Lower operating costs (28%)

While those are compelling and tangible benefits, healthcare organizations need the right tools to extract value from unstructured data. Medical AI and large language models (LLM), purposely trained on clinical data and used in conjunction with human experts, are able to understand medical shorthand, clinicians’ notes, PDFs, and other forms of unstructured data. These technologies extract and translate this unstructured data into easily digestible information far more effectively than tools previously available on the market.

Willingness to work with partners
Healthcare organizations know they aren’t in the IT business. Consequently, it’s no surprise that 83% of respondents to the survey say they would consider working in some capacity with a technology vendor or other third party to extract insights from clinical unstructured data. Only 14% are using or planning to use internally developed tools to extract insights from unstructured data, with these respondents most frequently citing security and privacy concerns (43%) as their top reasons for going it alone.

Despite these concerns, the vast majority (94%) of survey respondents said they would be interested or extremely interested in beta testing with potential partners, placing a priority on accuracy, reliability, and healthcare industry expertise.

Conclusion
Healthcare organizations who have done the work to educate themselves on the business and clinical value hidden in unstructured data expect their investments in AI to yield returns within a couple of years. With the right tools and partners in place, these organizations will be able to scale their deployment of medical AI to extract unstructured data across their entire enterprise and leverage it to meet their revenue, cost takeout, and innovation goals.

About the author: Jaimes Blunt is the executive vice president of emtelligent.