Stanford uses AI to navigate health information sharing with teens and their families

The models developed by Stanford Medicine Children’s Health have been so effective that Epic has incorporated a similar guardrail into its own system, extending the impact of Stanford's work to teens worldwide.
By Bill Siwicki
11:00 AM

Photo: Stanford Medicine Children’s Health

In the rapidly evolving landscape of healthcare, the sharing of health information is a delicate balance, especially when it comes to teenagers.

THE PROBLEM

Patient portals facilitate efficient sharing of health information between patients, their families and healthcare providers. However, a unique challenge arises when it comes to teenagers.

Certain state laws may allow minors to consent to specific types of sensitive care and give them the right to keep this information confidential from their guardians. This is crucial for building trust with vulnerable youth who might otherwise avoid seeking the care they need.

Yet, it's technically challenging to configure EHRs and patient portals to respect these privacy rights while also facilitating necessary information sharing.

"Some organizations respond to this challenge by withholding all health information from guardians," said Dr. Natalie Pageler, chief health information officer and division chief, clinical informatics, at Stanford Medicine Children’s Health. "However, this approach can inadvertently create barriers to optimal care. The reality is, most health information is not sensitive and can be crucial in engaging families in a teen's healthcare journey, particularly for those with chronic medical conditions.

"Investing in the technical work to create a nuanced approach to health information sharing is not just a necessity, it's an opportunity," she continued. "It's an opportunity to revolutionize the way we handle teen health data, ensuring both teens and their guardians have appropriate access to the information they need."

This goes beyond mere compliance with laws or future-proofing systems: It's about creating a healthcare environment where teens, regardless of their background, feel safe, supported and empowered, she added.

"Moreover, this nuanced approach supports the concept of graduated responsibility in healthcare for teens," Pageler said. "As they grow older, they gradually take on more responsibility for their health decisions. By giving them control over their sensitive health information, we're facilitating this transition to adulthood and fostering a sense of responsibility and autonomy.

"This work is about acknowledging the critical role families can play in supporting a teen's health while respecting the teen's rights to privacy in certain areas," she continued. "By striking this balance, we can improve patient outcomes, build trust and set a new standard in patient-centered care."

This is where AI and language models are stepping in.

PROPOSAL

Pageler and staff saw a significant opportunity to leverage AI technology to help navigate the nuances of health information sharing with teens and their families and to increase efficiency of provider note writing, messaging and manual monitoring programs.

"The first step in achieving these aims was to use natural language processing to set up a dual monitoring system," she explained. "This system was aimed at ensuring the privacy and autonomy of teenage patients while also maintaining the necessary level of guardian involvement in their healthcare.

"Our goal is to further decrease documentation burden and enhance the efficiency and effectiveness of provider communication."

Dr. Natalie Pageler, Stanford Medicine Children’s Health

"First, the AI technology was used to monitor provider notes for inappropriate sensitive content," she continued. "This means the AI system was trained to identify and flag any notes that contained information about sensitive care that a teenager had consented to, such as contraception or STI treatment. This system was designed to ensure this information was kept confidential from the guardians, respecting the teenager's privacy and autonomy."

Second, a natural language algorithm was created to monitor messages within teen patient portal accounts. The system was designed to detect any inappropriate use of the account by the guardian, such as attempts to access sensitive information.

This ensured the guardian was involved in the teenager's healthcare to the appropriate extent, but not beyond the boundaries set by the law and the teenager's consent.

"The combination of these two AI monitoring systems was designed to provide a comprehensive solution to optimally support teen and family engagement in their healthcare," Pageler explained. "These systems are designed to ensure both the provider notes and the patient portal accounts are used appropriately, respecting the rights of the teenager and the guardian.

"This not only helped to build trust with the teenagers but also ensures the guardians are able to support their child's healthcare effectively," she added.

MEETING THE CHALLENGE

Stanford Medicine Children’s Health embarked on an innovative journey, leveraging the expertise of its clinical informaticists who specialize in pediatric and adolescent medicine. These experts, well-versed in the nuances of California’s adolescent confidentiality laws, meticulously labeled large data sets of teen progress notes and patient portal messages. This formed the foundation of the project.

"Our informaticists then joined forces with our data scientists to develop two groundbreaking natural language processing algorithms," Pageler explained. "The first algorithm was designed to automate audit and feedback for our providers, enhancing the integrity of their notes and facilitating optimal health information sharing with teens and guardians.

"The second algorithm was developed to provide alerts to our patient portal team, enabling them to swiftly address potentially compromised patient portal accounts for teens and guardians," she continued.

The advent of large language models has opened up a world of possibilities.

"We're now exploring how we can capitalize on these advancements to further augment our work and better support teens and their families," Pageler noted. "We're currently evaluating the performance of GPT 4.0 in comparison with our homegrown NLP algorithms. We're also exploring how the integration of this tool with the Epic EHR system could proactively guide providers in appropriate teen and guardian communication, making their documentation more efficient.

"This project is not just about using technology to solve a problem," she continued. "It's about revolutionizing the way we support and empower teens and their families. This is the future of pediatric healthcare, and we're excited to be at the forefront of this change."

RESULTS

The journey with this technology has yielded some truly transformative results, Pageler reported.

"First, our patient message algorithm uncovered a startling reality – more than 50% of our teen portal accounts were being inappropriately accessed by guardians," she said. "This was a wake-up call, highlighting a significant breach of adolescent privacy.

"But it also presented an opportunity for change," she added. "We took immediate action, conducting further research to understand and address the issue. Our efforts culminated in the development of a custom guardrail intervention that significantly reduced inappropriate account activation errors. This solution was so effective that our vendor partner, Epic, incorporated a similar guardrail into its foundation system, extending the impact of our work to teens worldwide."

Second, the NLP algorithm for provider notes revealed more than 20% of notes contained inappropriate confidential content. The automated review process powered by the algorithm was up to six times more efficient than manual review, a significant gain in productivity. But more important, it helped staff identify opportunities to improve provider note templates, making it easier for providers to document appropriately and respect adolescent confidentiality.

"These successes are just the beginning," Pageler said. "We're now exploring ways to integrate these algorithms into the note generation process. Our goal is to further decrease documentation burden and enhance the efficiency and effectiveness of provider communication. This is not just about improving our systems; it's about revolutionizing the way we engage teens and their families to help them live healthier and happier lives."

ADVICE FOR OTHERS

As Stanford Medicine Children’s Health navigates the rapidly evolving landscape of healthcare technology, it's crucial to recognize the unique needs of different patient populations, Pageler advised.

"Children and their families, for instance, require a nuanced approach to health information sharing that respects both the rights of the child and the role of the family in their care," she said. "Implementing technology solutions like AI and natural language processing can greatly enhance our ability to meet these needs, but it requires a deep understanding of both the clinical and technical aspects of healthcare.

"Building on this, assembling the right team is absolutely imperative to successfully navigate this complex intersection of healthcare and technology," she continued. "This team should ideally be composed of individuals with diverse but complementary skills – clinical informaticists who understand the unique needs of patients and the nuances of healthcare workflows, data scientists who can harness the power of AI and natural language processing, and legal and privacy experts well-versed in healthcare laws and regulations."

Stanford has been successful in this endeavor due to its longstanding commitment to interdisciplinary education, she added.

"Our Biomedical Data Science Graduate Program, Clinical Informatics Fellowship, and Masters in Clinical Informatics Management have been instrumental in training a new generation of professionals who are adept at bridging the gap between clinical practice and technological innovation," she concluded. "This blend of clinical and technical expertise within our team has enabled us to develop and implement solutions that are tailored to the specific needs of our patients, particularly children and their families."

Follow Bill's HIT coverage on LinkedIn: Bill Siwicki
Email him: bsiwicki@himss.org
Healthcare IT News is a HIMSS Media publication.

Want to get more stories like this one? Get daily news updates from Healthcare IT News.
Your subscription has been saved.
Something went wrong. Please try again.