Robust Medical Crowdfunding Predictions: Leveraging Large Language Models and Machine Learning for Success Analysis

BackgroundIn the field of medical crowdfunding prediction, traditional statistical methods have long been the standard. Machine learning algorithms are popular because they can model complex relationships between variables, capture interactions, and provide more accurate predictions, even when input variables are highly correlated. Furthermore, previous research has largely overlooked the quantitative assessment of success levels and the selection of key predictors. To address these limitations, a novel approach is needed that leverages advanced machine learning techniques. ObjectiveThis study aimed to address these gaps by proposing a robust feature engineering approach that leverages the capabilities of large language models (LLMs). The goal was to extract the success determinants using a large language model for a cancer crowdfunding campaign. Furthermore, this study evaluated the performance of four machine learning algorithms in predicting campaign success and quantitatively assessed the level of success. MethodWe separately analyzed linguistic and social determinants of health features to understand how much each factor contributes to a crowdfunding campaigns success. These features were generated using a large language model (GPT-4o). A random forest algorithm with a permutation technique was used to rank the features. We comparatively evaluated the prediction accuracy, sensitivity, and specificity of four machine learning algorithms, random forest, gradient boosting, logistic, and elastic net, using a 10-fold cross-validation. ResultsGradient Boosting consistently outperforms the other algorithms in terms of sensitivity (consistently around 0.786 to 0.798), indicating its superior ability to identify successful crowdfunding campaigns using linguistic and social determinants of health features. The permutation importance score reveals that for severe medical conditions, income loss, chemotherapy treatment, clear and effective communication, cognitive understanding, family involvement, empathy and social behaviors play an important role in the success of campaigns. ConclusionsThis study highlights the critical role of linguistic, social, demographic, and medical features in predicting the success of cancer crowdfunding campaigns, with risk communication and medical severity emerging as key predictors. The study also suggests the need for more nuanced and optimized models and improved income protection and healthcare policies to reduce reliance on crowdfunding for cancer treatment.