Getty Images
Press Ganey all-in on AI, patient satisfaction measurement
Patient experience surveying company Press Ganey bet big on AI in patient satisfaction measurement, using the tool to streamline patient insights.
For a healthcare leader faced with a less-than-impressive patient satisfaction score, their first thought will almost always be the same: What can I possibly do to fix this? According to Chrissy Daniels, AI in patient satisfaction will make it easier to answer that question.
For decades, patient satisfaction surveying has been a beacon guiding organizations' quality improvement initiatives. The quantitative and qualitative insights gleaned from these surveys have been critical for showing where hospitals and health systems perform well and where they have room to grow, all with the ultimate goal of boosting the patient experience.
But the challenge with patient satisfaction data is the same as all big data in healthcare: it's simply a tall order to make sense of just a breadth of information.
Just as AI has enhanced data analytics, it is already making a big difference in the patient experience space. Press Ganey and its peers in the patient satisfaction surveying and consulting space are betting big on AI and large language models (LLMs) to turn insights into action for its users.
LLMs guide strategic thinking in patient experience
According to Daniels, AI, LLMs and natural language processing (NLP) can supplement the strategic thinking and decision-making of the patient experience team.
"At Press Ganey, we use AI for a few things," Daniels explained. "We use it to summarize findings. How do we look at things in a large way? How do we compose responses? And then, how do we make recommendations?"
Take that patient experience leader from the top of this article. When they see low performance in a certain patient satisfaction measure, they're almost always going to consider the best path forward to improve their scores and the patient experience. That, in essence, is the entire point of surveying about the patient experience in the first place.
But with AI, patient experience leaders can get organized and streamlined insights better tailored to their current needs. In the past, these leaders were beholden to industry and personnel variation, Daniels explained, meaning they might not always get the strongest insights to make quality improvement decisions.
"Now we can eliminate that," Daniels said. "We can immediately take someone to, 'oh, we see there's variation. What is the reason? Is it that a behavior that you were doing in the past that got you great results started to drift? Is there a group of people based on a location, based on a condition, based on an identity group who is having inconsistent care?'"
"We can bring the wisest data scientists into the understanding of every leader," she added.
But it's not just enhancements to the quantitative data analysis that has Daniels eager to see AI's potential.
Rather, AI, and particularly NLP, can get to the meat of qualitative data through free-text comments.
LLMs give power to free-text patient survey comments
According to Daniels, the most powerful patient satisfaction survey insights come from the comments.
"I am a true believer in comments -- the more, the better," Daniels stated. "There has never been a week of my life in the last 30 years that I didn't spend time reading several thousand comments. And I do it because there's always wisdom in those comments. That's where I understand either what's working or not quite working or a group that might be left behind."
But in an industry increasingly defined by burnout and dwindling workforce numbers, the last thing any organization wants to do is add to workload.
That's true for every staff member, from the frontline clinician to the backend claims analyst to the patient experience expert.
Therein lies the rub. Patient experience leaders want and need the insights provided in free-text comments, but they don't always have the time to chase them down.
"That requires a lot of discipline and a lot of time," Daniels acknowledged. "Is there a way to get those insights more consistently for people who either don't have the luxury or don't have the wherewithal to spend their time with all those comments?"
AI opens this door, creating an avenue for patient experience leaders to get both broad and granular insights into qualitative comments and potentially avoid human error and cognitive bias.
"You read a bunch of comments, and it's sometimes hard to make haystacks," or groups of comments with similar sentiment, Daniels explained. "There are all these comments, and everything sounds a little bit different. And if you're not reading enough of them, you might think one thing is a more important issue than the other, or it's hard to prioritize."
Although AI is prone to its own biases, it can be better at rooting out the true sentiment of a patient comment and identifying themes across multiple patient reviews. The technology can take these comments and make haystacks, or provide summaries of how patients are feeling about the care they receive at a clinic or hospital.
But perhaps equally exciting is that LLMs can work in the reverse. They don't have to turn insights into haystacks. Rather, they can help users find the needle in those haystacks.
"Sometimes when you get that summary information, the haystacks, it's hard to find the needles," Daniels noted. "There are one or two patients who are being left behind who really need attention. Now we can find both the haystacks and the needles."
For instance, a patient experience leader knows that kindness is important in a medical setting. Patients universally value a provider who is kind.
That's a general insight, or one of the haystacks. NLP can parse through and tell the patient experience leader that top-performing providers are considered kind.
But also using NLP, the leader can identify with specificity -- find the needles -- which behaviors make a provider come across as kind. That creates action items that let the leader further coach their team.
"Because we have over a billion healthcare voices, we can create that sweet spot of specificity but also generalization, and that is really exciting to me," Daniels said.
Coaching healthcare teams to query AI
Notably, integrating AI into any system, including patient experience surveying, will require some training and change management. For example, health system leaders will need to learn the best way to query AI to churn out the insights users truly look for.
"The AI is only as smart as the query writer, and we're not universally skilled in that way," Daniels pointed out.
Press Ganey is working with its tens of thousands of health system clients to flag good queries and folks whose searches typically yield quality information. The company can use that information to better understand the most important queries and set that as a standard within the tool.
Moreover, Daniels stressed the importance of understanding how patients actually talk, especially in a healthcare context. That will influence how the AI should be trained to fully appreciate patients' free-text comments.
For instance, the term "blood" typically has a negative connotation in everyday life. However, in healthcare, it can be neutral or even positive. Daniels said Press Ganey is investing in learning and understanding who patients are when they engage in their healthcare to make the most of their tools.
That investment is going to be crucial moving forward, as healthcare providers begin to embrace AI for its ability to streamline certain tasks. The technology is effective at summarizing key findings, including on patient satisfaction assessments, priming it to refocus and enhance the role of patient experience leaders.
Sara Heath has reported news related to patient engagement and health equity since 2015.
