Kaiser Permanente has been experimenting with AI in its patient portal, increasing patient engagement and experience in the process.
“Care is local, but at the same time it’s virtual and it’s become a global commodity,” Khang Nguyen, MD, assistant executive medical director for care transformation at Southern California Permanente Medical Group and chief medical officer of care navigation for the Permanente Federation, told Becker’s. “So patients are really expecting artificial intelligence to support healthcare in a way that is supporting other industries, in the sense that people are able to describe what they want versus being given choices.”
A new study, published in npj Digital Medicine in partnership with Seoul National University Bundang Hospital, focused on the deployment of the Kaiser Permanente Intelligent Integration (KPIN) system by the Southern California Permanente Medical Group (SCPMG), which serves 3.9 million patients. Implemented in October 2024, KPIN is an advanced patient portal system designed to enhance care navigation and patient experience.
KPIN replaced older systems (Microsoft’s LUIS and VSN Query Report) and integrated natural language processing (NLP) to generate clinical alerts for high-acuity cases and recommend appropriate care pathways. It supports multiple channels, including a web-based portal, mobile app, and self-service interactive voice response (IVR) system, ensuring consistent care navigation. The system verifies patient identity and guides them through booking processes, redirecting emergencies to 911 or hospitals.
October 1, 2024, to March 1, 2025, KPIN facilitated 2,960,945 digital visits, with 1,046,504 unique patients interacting with the system. Demographic data showed 36.6% of users were aged 30–49, 47.8% were female, and 38.9% were of Hispanic origin.
The study evaluated KPIN’s Clinical Alert System (CAS) and Virtual Safety Net (VSN) models for detecting high-acuity symptoms (e.g., chest pain, dyspnea) and guiding care navigation. Metrics included accuracy, precision, recall, F1-score, and area under the curve (AUC).
KPIN processed an average of 193,134 encounters daily, with a peak of 19,364. The adjusted successful booking rate was 53.8%, reflecting effective appointment scheduling. The abandonment rate was low at 0.94% (IQR: 2.7–3.1%), indicating high user engagement, attributed to a streamlined interface limiting interactions to nine per page.
Patient surveys showed an 8.6 percentage point increase in positive sentiment, suggesting improved user satisfaction. The system struggles with vague “Reason for Visit” entries, which can affect care pathway accuracy.
The authors of the study concluded that “KPIN’s integration into the patient portal significantly improved care navigation, clinical alert accuracy, and patient satisfaction within an integrated value-based care model, demonstrating the effectiveness of advanced digital tools in healthcare delivery.”
Dr. Nguyen called that the “biggest surprise” from the findings. “We all know with websites, if you want to bury anything on a website, just make it two clicks away, and then no one can ever find it,” he said. And when patients abandon the app because they can’t locate what they need, they typically contact the call center.