Work

Sentiment Analysis Applied to Digital Conversations Among Warm Autoimmune Hemolytic Anemia (wAIHA) Patients Receiving Rituximab and/or Blood Transfusion

Year: 2024

Type: conference-abstract

Abstract: Background: Warm autoimmune hemolytic anemia (wAIHA) is a rare, life-threatening, chronic and relapsing autoimmune disorder caused by autoantibodies that bind to and ultimately destroy red blood cells, leading to anemia when hemolysis is not adequately compensated. Immunosuppressive and anti-CD20 B cell depleting therapies (e.g., rituximab) are often used to treat patients with symptomatic anemia while blood transfusion is available as a rescue therapy for those with severe anemia. Given the lack of data on the humanistic burden of wAIHA, sentiment analysis is an effective tool for highlighting areas of greatest concern to patients, specifically those treated with rituximab and/or blood transfusion. Methods: Public domain digital conversations among adults who engaged in conversations about wAIHA online over a 24-month period (August 2021 to August 2023) were analyzed using artificial intelligence (AI) -powered algorithms. Sources of conversations included topical sites, message boards, social networks, comments, and blogs. The de-identified digital conversations were limited to those written in English and originating from US Internet Protocol addresses. The AI methodology extracted and sorted content from contributors who self-identified as having wAIHA into categories and themes. Segments representing wAIHA patients treated with rituximab or transfusions were further identified from the dataset of conversations. Each conversation was categorized as either positive, negative, or neutral (ex. information seeking/sharing) in tone toward wAIHA in a sentiment analysis using natural language processing (NLP) followed by breakdown of the most frequent themes within each sentiment category. Sentiments were further categorized by different stages of the wAIHA journey, including ‘pre-diagnosis’, ‘diagnosis’, ‘treatment’, and ‘treatment adjustment’. Results: Among 22,309 conversations related to wAIHA, 2,214 were from patients discussing rituximab treatment, and 1,211 were from patients discussing transfusion. Conversations were classified as expressing positive (5%), negative (38%), or neutral (57%) sentiments among rituximab-treated patients, and positive (0%), negative (57%), or neutral (43%) sentiments among transfusion-treated individuals. The most common negative sentiment themes for both rituximab and transfusion-treated patients were lack of treatment efficacy (16% rituximab; 27% transfusion) and side effects (16% rituximab; 17% transfusion). Most of the conversations among both rituximab and transfusion-treated patients were associated with being in the ‘treatment’ or ‘treatment adjustment’ stages and were highly negative (46% and 56% for rituximab; 50% and 69% transfusion). Among the neutral wAIHA patient conversations which reflected information seeking, ‘treatment/therapies’ was the most frequently discussed topic (38% overall), followed by ‘resources’ (27%), ‘symptoms’ (21%), and ‘diagnosis’ (14%). Conclusion: This study leveraged a novel approach to explore the patient perspective in terms of sentiments and common topics of interest in wAIHA patients treated with rituximab and blood transfusions. Digital conversations revealed notable concerns related to side effects and a lack of treatment efficacy while also highlighting the importance of HCP support and involvement in implementing a successful treatment strategy. The AI methodology uncovered patient sentiments and themes in an unprompted fashion, providing candid insights that could be leveraged in the implementation of patient-centric approaches to wAIHA patient care.

Source: Blood

DOI: 10.1182/blood-2024-207786