In a new, large comprehensive analysis led by the American Cancer Society (ACS), researchers, using a form of Artificial Intelligence (AI), found that more than one-third of fundraising stories on the GoFundMe crowdfunding platform in the United States explicitly shared experiences of medical financial hardships and health-related social needs (HRSNs). The fundraising stories included hardships such as housing and food insecurities, transportation barriers, income loss, lack of sick leave, and disruptions to both work and school. The findings are published today in the Journal of the American Medical Association (JAMA) Oncology.
“Sadly, financial hardship is common among cancer survivors across the country, forcing a growing number of patients and their families to use personal crowdfunding as an alternative source to raise money. These findings show the intense difficulties in meeting basic medical and social needs, underscoring the fragility of safety nets in the U.S.”
Dr. Zhiyuan “Jason” Zheng, senior principal scientist, health services research at the American Cancer Society and lead author of the study
For the study, researchers analyzed data from all cancer-related fundraising stories from January 1, 2021, to May 31, 2023, retrieved from the publicly available crowdfunding website GoFundMe. Scientists utilized extensive natural language processing (NLP) modeling (Open AI’s ChatGPT 3.5) to examine cancer-related crowdfunding campaigns, specifically their fundraising stories about reasons for financial assistance, including medical financial hardship and HRSNs. The advances in NLP enabled researchers to transform qualitative data to quantitative data and to help perform statistical analyses.
Study results showed a total of 91,113 cancer-related crowdfunding campaigns were identified and more than 24 million words were analyzed. The proportions with NLP outputs for individual campaign characteristics were age (19.6%), sex (61.1%), marital status (5.1%), family size (12.8%), health insurance coverage (18.3%), employment status (20.6%), living with dependent children (16.4%), and school attendance (9.2%). 79% had NLP interpretations (outputs) for cancer type. 33.9% for stage at diagnosis, 43.3% for new versus recurrent cancer status, 52.6% for cancer-related treatments, and 31% for time from diagnosis to campaign initiation. Among all fundraising stories, 25.5% had NLP outputs for any medical financial hardship, and 24.1% had mentioned HRSNs. Overall, 35.9% of fundraiser stories had NLP outputs with any medical financial hardship or HRSNs.
Dr. Robin Yabroff is senior author of the study. Other ACS researchers contributing to the report include Dr. Shaojun Yu, Dr. Farhad Islami and Dr. Jingxuan Zhao.
Source:
Journal reference:
Zheng, Z., et al. (2024) Natural Language Processing–Assessed Unmet Medical and Social Needs in Cancer Crowdfunding Stories. JAMA Oncology. doi.org/10.1001/jamaoncol.2024.4412.
Source link : News-Medica