Technology Trends: Measuring Person-Centered Care Using Innovative Artificial Intelligence
By Michele Deppe
Social Work Today
Vol. 23 No. 1 P. 5

Researchers are probing the problem of service disengagement and creating a new measure for person-centered care with cutting-edge technology.

Service disengagement is a pervasive problem in the mental health care system. Often, people opt out of a generic care plan because it doesn’t meet their personal needs. According to research, the US mental health care system, compared with that in similar countries, typically spends more money on each person but tends to have the poorest outcomes. Victoria Stanhope, PhD, MSW, MA, associate dean of faculty affairs, a professor, and director of the PhD program at NYU Silver School of Social Work, has proven that when people don’t receive real-life solutions, their needs are expensively addressed with emergency department and hospital admissions. “It’s a very high disengagement rate, up to 60%. And many people don’t show up for outpatient appointments even though they’ve been hospitalized and referred to outpatient,” Stanhope says. “We spend time developing beautiful evidence-based practices, but it’s all moot if people don’t want to engage in services.”

One of the most effective ways to gain what Stanhope calls the triple aim of “better health, better care, and lower cost through improvement” is to practice person-centered care, providing consumers with a unique plan based on their needs, supported by meaningful choices, and the opportunity to give input regarding their health goals. But even when clinicians practice person-centered care there remains the challenge of how to measure that care.

A study is underway to address this challenge by creating a strategy to help clinicians measure the person-centered care they deliver. The NYU Constance and Martin Silver Center on Data Science and Social Equity has provided the $70,500 funds for the project.1 Previously, Stanhope and a team of researchers meticulously reviewed 800 service plans, manually examining them for data that would help clinicians measure individualized care. “It took about half an hour per service plan,” Stanhope says. “It’s an unwieldy process, and it’s not practical for agencies. They need a tool to evaluate their practices. That’s where the germ of the idea for this study came from, to use machine learning to assess clinical notes and service plans.”

Using computers to gather information from EHRs, the data can be used to create an algorithm that can assess the extent to which narrative language is reflective of person-centered care. Utilizing natural language processing—an innovative artificial intelligence that understands the complexities of everyday language—and even emotion—the computer sifts through the data, gleaning relevant insights and concepts that become a body of information that can be searched, analyzed, and interpreted. The goal is to create an algorithm to analyze clinical visit notes and measure the level of person-centered care delivered to the consumer.

Patients as Active Participants
One of the adages of the mental health recovery movement is nothing about us without us. Soon, if not already, writing clinical notes will be an open process. Charting progress changes quite a lot when the person sitting next to the clinician can read what is written about them.

Stanhope was the lead author of a study published in Psychiatric Services to test person-centered care planning in community mental health clinics. Person-centered care is generally defined as providing care that is respectful of and responsive to individual patient preferences, needs, and values, and moving toward a more integrated care delivery.2

“Some of the challenges around service disengagement is [inherent] in the community mental health system,” Stanhope says. “We’re talking about people with severe mental health problems who are in a system focused very much on symptom reduction, and it’s also a system with a history of being coercive. So, person-centered care planning is an intervention that orients care toward a person’s unique life goals.”

Collaborative Documentation
Stanhope challenges clinicians to consider how much of the treatment plan is developed collaboratively. How much does the person get to determine it themselves? “Providers have a hard time Individualizing service plans that are often driven by predetermined domains,” she says. “This leaves out personal life goals, which can be a whole myriad of things, and the format of a service plan can preclude recovery-oriented work.”

Stanhope’s study used natural language processing to see whether the practice of collaborative documentation led to clinical notes that were more person-centered. Whereas most clinicians write notes after the client leaves, collaborative documentation is a practice where clinicians write the note with the consumer, potentially transforming the “busy work” of documentation into a more meaningful activity.

Collaborative documentation is an opportunity to achieve a more positive tone and document in a more person-centered way. Gahwan Yoo, a PhD student and Stanhope’s research assistant, shares examples from the preliminary study. “The computer uses sentiment analysis utilizing the dictionary developed by a computer scientist and a psychologist to analyze emotion-related words. Certain words, like the verb ‘improved,’ appeared more frequently after collaborative documentation than before,” Yoo says. “In the computer’s sentiment analysis, there is a way to measure the documentation’s tone, positive or negative. We found that collaborative documentation is associated with the increase in the positive tone of some sections like goals and objectives, or progress toward an outcome.”

Most Providers Rate Themselves Highly
Researchers learned that while providers often rate their ability to provide person-centered care highly, self-reported competence can be unreliable, Stanhope says. “Those who didn’t know anything about person-centered care said, ‘we’re very person-centered.’ However, it’s like, you don’t know what you don’t know. And this is very common for self-reported competencies, so it’s good to have an objective measure and that is what we are trying to create with this study.”

The study is still in progress, but cultivating person-centered care in your practice doesn’t have to wait. Stanhope says the first step is building a relationship with your client. “Relationships take time,” she says, acknowledging that many agencies don’t give the providers enough time to understand the person they’re working with. “There’s this drive to diagnose and then come out with a treatment plan,” Stanhope says. “Providers feel pressure to document primarily for billing purposes and must submit notes within 24 hours for Medicaid reimbursement. I think the challenge is trying to make it a more meaningful activity,” she says. Despite the limited appointment time, there’s a possibility for greater engagement when people are encouraged to share their perspectives. Ask yourself, how much of this treatment plan is being developed collaboratively? How much does the person get to determine it themselves?

Reaching out to people who are part of the client’s network or important to them is underutilized. “We rarely ground people’s recovery in the community,” Stanhope says. “So as much as you can, bring in nontraditional and informal supports that allow people to integrate into the community rather than just being sort of in this bubble of services.”

Finally, she says people often have a hard time getting out once they’re in the system. “I think one of the most controversial parts of recovery—but very important—is that people’s recovery is not linear. We must be flexible and calibrate care on an individual basis,” Stanhope says. Part of this, she says, is allowing people to take risks. “Providers make a lot of decisions for people, like ‘you’re not ready to do this, I don’t think you can live independently,’” she says. “Not surprisingly, agencies are very risk averse. And that prevents people from being able to grow.”

Conclusion
When an algorithm exists that can measure and assess the person-centeredness of narrative data, it will be a most helpful tool for quality improvement. “Everyone is mandated to do this now, but everyone is still trying to agree on exactly what person-centered care is. So, if we can come up with a way of measuring it, and agencies can use this within their system to see how they’re doing, then that would be powerful,” Stanhope says. The hope is that these computer algorithms will allow agencies to use a program and analyze the documentation, even for a microlevel review of individual providers. This is helpful for research purposes and quality improvement, Stanhope says, but it could also help agencies when they must report to governing bodies like the Joint Commission and show that they are delivering person-centered care.

— Michele Deppe is a freelance writer based in South Carolina.

References
1. Funded projects: harnessing natural language processing to measure person-centered care in behavioral health settings. New York University Silver School of Social Work website. https://socialwork.nyu.edu/faculty-and-research/centers-and-institutes/c-m-silver-center.html. Accessed November 9, 2022.

2. Stanhope V, Choy-Brown M, Williams N, Marcus SC. Implementing person-centered care planning: a randomized controlled trial. Psychiatr Serv. 2021;72(6):641-646.