The rapid increase in medical data has underscored the importance of accurately and efficiently classifying medical texts. Traditional models often face challenges due to the specialized language and complex structures inherent in these texts. To tackle this challenge, this study proposes a cutting-edge blending ensemble learning framework, aimed at transforming the classification of complex medical texts. This framework integrates three specially developed hybrid learners: BERT-BiLSTM-DPCNN, ERNIE-BiLSTM-DPCNN, and eHealth-BiLSTM-DPCNN. Each learner merges the strengths of various advanced deep learning models, focusing on thorough extraction and emphasis of key features in medical texts, thereby enriching the input for meta-learners. A unique aspect of this method is the combination of insights from meta-learners through a voting process, which significantly enhancing the classification’s accuracy and reliability. Rigorous testing demonstrates that our approach markedly outperforms existing models in critical metrics, including accuracy, macro precision, macro recall, and macro F1 scores, showing notable improvements of 8.1%, 6.9%, 8.1%, and 7.7%, respectively. This study represents a significant advancement in medical text analysis.