Assessing Unmet Information Needs of Breast Cancer Survivors: Exploratory Study of Online Health Forums Using Text Classification and Retrieval (Preprint)

Abstract

IntroductionGoal of this work Our project aims to assess the unmet information needs of breast cancer survivors (e.g. [1][2][3][4][5]) from the patients' perspectives. Such an assessment would be valuable for health care educators selecting and developing content to be incorporated in resources for breast cancer survivors. While existing resources cover many of the topics that breast cancer survivors might need, it is unknown how well these resources address their actual needs. Existing content is largely selected and developed by healthcare providers. Existing methods for identifying patient needs, such as surveys used in [1][2][3][4][5], have limitations because they collect self-reported information, rather than observe survivors' informationseeking. These limitations should be addressed, especially before investing in new, more personalized platforms that answer questions.Most existing resources for survivors, including brochures, books, and care plans, are static paper documents or webpages. By design, static content must balance the goals of covering the most commonly needed topics, while remaining manageable in size. By contrast, interactive electronic documents and services can be made available in numerous formats and, by dynamically targeting expressed information needs, can both cover a wide range of topics and remain usable for any individual.Creating new education materials can be costly, especially if they must be revised multiple times because they were not well-planned. Poorly planned resources risk being rejected, if patients cannot find the information they need easily. Voice assistants and chatbots that support question answering will require even more specific information about the topics of interest than a document-based information retrieval system, to deliver content in manageable size chunks. Thus, assessing patients' perspectives on their information needs, in the most authentic means possible, contributes to the goal of designing of new tools. BackgroundThere have been several recent efforts to assess the unmet needs of cancer patients. Many use variants of the validated Supportive Care Needs Survey [1][2][3][4][5]. This survey covers five factors (psychological, health system and information, physical and daily living, patient care and support, and sexuality needs). Among the most commonly reported needs are the need for counseling to deal with psychological distress and the need for information, including more information about treatment, prognosis, wellness, and managing symptoms and sideeffects. These factors provide a useful starting point for an automated analysis of needs, as well as aspects that might benefit from evidence obtained from a complementary method of assessment.Having multiple methods for assessing information needs would be valuable, as relying on self-reported needs allows for many sources of bias that limit the reliability of the results. For example, bias can arise from how questions are worded, how subjects are recruited, and the beliefs and psychology of indivi...