Penn Medicine has introduced an AI chatbot named “Penny” to enhance cancer care. Patients exchanged nearly 4,000 medication-related text messages with Penny, with an accuracy rate of approximately 93%. Patient engagement for symptom assessment increased from 25% to 70%. Penny provides medication reminders, assesses symptoms, and improves medication adherence. The chatbot’s implementation has shown feasibility and effectiveness, leading to improved patient outcomes and increased confidence in taking medication.
Penn Medicine has implemented an AI chatbot called “Penny” to enhance cancer care. The utilization of Penny resulted in the exchange of approximately 4,000 medication-related text messages between patients and the chatbot, with an impressive accuracy rate of 93%. Additionally, patient engagement for symptom assessment experienced a substantial increase from 25% to 70%.
Over the past 25 years, the field of oncology has undergone significant changes in the delivery of cancer care. Advancements include the development of new therapeutics, a shift towards home-based care, and widespread adoption of electronic health records. One area of innovation has been the introduction of oral anti-cancer drugs, which allow patients to self-administer medication at home. These drugs offer increased autonomy and more time spent at home, but they also present challenges such as complex medication schedules, concerns regarding dosage adherence, and potential toxicity risks.
Dr. Parul Agarwal, an assistant professor of clinical medicine at the University of Pennsylvania Health System, leads the “Penny” team, focusing on gastrointestinal cancers. Dr. Agarwal acknowledges the complexity of administering oral chemotherapies at home and the need for strategies to mitigate these challenges. Intensive outpatient monitoring programs driven by providers and pharmacists have been explored, along with the assessment of medication adherence and appropriate management of toxicities. However, these approaches often prove resource-intensive and error-prone.
To address these issues and improve patient outcomes and healthcare utilization metrics, Penn Medicine proposed the use of an augmented intelligence chatbot named Penny. Penny is an algorithmically driven chatbot that engages in text-based, bi-directional conversations with patients. Its purpose is to guide patients through complex treatment regimens and reduce the likelihood of mistakes during chemotherapy courses.
The pilot program allowed patients with gastrointestinal cancer, receiving oral chemotherapy capecitabine alone or in combination with other drugs or radiation, to participate. Penny’s bidirectional messaging system enabled patients or the chatbot itself to initiate conversations. To address medication adherence and dosing concerns, Penny sent twice-daily medication reminders to patients. Initially, participants were asked to respond “TAKEN” after each dose, which was later modified to a weekly survey to assess adherence.
To manage toxicities associated with oral anti-cancer drugs, Penny initiated a weekly survey to collect patient-reported symptoms. These symptoms were evaluated using algorithmic surveys and natural language processing, an AI technique. Participants could also initiate a message with Penny at any time to report a symptom. Real-time monitoring of all interactions between participants and Penny was conducted by the study team.
The results of the pilot program demonstrated the feasibility and effectiveness of using Penny. Out of nearly 4,000 medication-related text messages, Penny accurately interpreted approximately 93%. The measured medication adherence rate among participants who responded “TAKEN” was approximately 70%, with the actual rate likely higher. Regarding symptom-related messages, Penny correctly interpreted around 98% of over 500 text messages exchanged. Through iterative improvements, patient engagement in symptom assessment increased from 25% to approximately 70%, and any misinterpreted messages were promptly addressed by the research team.
Qualitative exit interviews were conducted with over 50% of participants, revealing a high level of patient engagement and positive feedback regarding Penny’s additional support. Patients reported increased confidence in medication adherence and interactions with their care team.
Dr. Agarwal advises others to approach healthcare innovations in an iterative manner, engaging all relevant parties, including healthcare teams, research teams, and technological partners. The implementation of Penny led to the identification of inconsistencies in care delivery among gastrointestinal cancer providers, prompting initiatives to minimize unnecessary variations. A patient journey dashboard was developed to outline treatment schedules and dosing, facilitating high-quality care delivery. Moving forward, this dashboard will be integrated into the electronic health records system to become a standard tool for clinical teams to monitor patients and adjust dosages as needed.
