Identifying Inquiries for an Artificial Intelligence-Based Informatic Chatbot Among Young Adult Survivors of Childhood Cancer Using a Topic Modeling Approach

Background and Purpose: Young adult survivors of childhood cancer encounter various challenges throughout their survivorship. The enduring impact of cancer diagnosis and treatment on these individuals from childhood through young adulthood gives rise to unique informational needs. However, they often struggle to access reliable information, which can lead to diminished quality of life during cancer survivorship. An artificial intelligence-based informatic chatbot may be a promising alternative for meeting their informational needs. However, topics that pique the curiosity of childhood cancer survivors remain elusive. This study adopted a topic modeling approach to elucidate the primary concerns and questions that survivors express when engaging with a chatbot tailored to assist with their needs.

Methods: In 2024, an online survey was administered to a convenience sample of 119 young adult survivors of childhood cancer with the aim of collecting inquiries for a tailored chatbot. Participants were 64 women and 55 men aged between 19 and 24 years old. A sizeable proportion of participants were college students (38.7%) and diagnosed of hematologic tumors (61.3%). Using Python 3.10.12 and OpenAI’s ChatGPT-4, a comprehensive three-step approach to topic modeling was employed. Initially, the hierarchical density-based spatial clustering of applications with noise clustering method identified a broad spectrum of topics. Subsequently, OpenAI’s ChatGPT further categorized the questions using these refined topics. Finally, the topics were refined through iterative reviews facilitated by a research team with specialized expertise in childhood cancer survivorship. This methodology ensured a systematic and rigorous analysis of the inquiries.

Results: The survey collected 1,237 questions from participants for the chatbot. After excluding 4 questions classified as outliers, these inquiries spanned 11 topics, with varying frequencies: health and daily life management during or after treatment (245 questions, 19.9%); psychological distress resulting from cancer diagnosis and treatment (181 questions, 14.7%); disclosing cancer history and its impact on life (170 questions, 13.8%); coping with side effects and late effects of cancer treatment (128 questions, 104%); understanding causes, diagnosis, and treatment of cancer (107 questions, 8.7%); social policy and services for childhood cancer survivors (80 questions, 6.5%); academic continuity and school reentry (78 questions, 6.3%); challenges in social relationship and social reintegration (73 questions, 5.9%), risks and concerns related to cancer recurrence (61 questions, 4.9%); peer survivors’ life and communication with (58 questions, 4.7%); and social stigma experiences and coping (52 questions, 4.2%).

Conclusion and implications: Our study identified key inquiries that young adult survivors of childhood cancer had for a chatbot. These findings underscore the importance of addressing a diverse array of health and psychosocial concerns that cancer survivors may face throughout their survivorship, particularly as they reintegrate into society. Developing informatic chatbots can significantly benefit this population by providing a readily accessible and personalized platform to address their unique informational needs. Furthermore, findings highlight the critical role of oncology social work professionals in delivering tailored support to meet the concerns of childhood cancer survivors as they navigate various challenges and opportunities in young adulthood, ultimately promoting their overall health and well-being.