ABSTRACT
OBJECTIVE: This research explored the experiences and perspectives of patients declining hospital transportation after receiving prehospital emergency care using advanced computational techniques. METHOD: Between 15th June and 1st August 2023, 210 patients in Qatar, treated by Hamad Medical Corporation Ambulance Service (HMCAS) but refusing transportation to hospital, were interviewed. Key outcome variables stratified by demographics included "reasons for refusing transport," "satisfaction with HMCAS service," and "postrefusal actions." Responses underwent sentiment analysis and topic modeling using latent Dirichlet allocation. Machine learning models, such as Naïve Bayes, K-nearest neighboring, random forest, and support vector machine, were used to predict patients' subsequent actions. RESULTS: Participants had an average age of 38.61 ± 19.91 years. The chief complaints were primarily chest and abdominal pains (18.49%; n = 39). Sentiment Analysis revealed a generally favorable perception of HMCAS-provided service. Latent Dirichlet allocation identified two main topics pertaining to refusal reasons and service satisfaction. Naïve Bayes and support vector machine algorithms were most effective in predicting postrefusal actions with an accuracy rate of 81.58%. CONCLUSIONS: This study highlighted the utility of Natural Language Processing and ML in enhancing our understanding of patient behaviors and sentiments in prehospital settings. These advanced computational methodologies allowed for a nuanced exploration of patient demographics and sentiments, providing insights for Quality Improvement initiatives. The study also advocates for continuously integrating automated feedback mechanisms to improve patient-centered care in the prehospital context. Continuous integration of automated feedback systems is recommended to improve prehospital patient-centered care.
