Online consultation platforms contain vast amounts of doctor-patient information, making it difficult for patients to make efficient decisions when selecting a doctor. Existing research on doctor recommendations were primarily based on interaction data from a single platform, which often faces the cold-start problems and fails to adequately consider patients'' personalized preferences. The study presents a collaborative decision-making model for doctor selection that integrates both doctor expertise and patient preferences. First, by analyzing patient reviews across multiple platforms, an indicator system for patient decision-making is developed. Second, initial candidate doctors are identified based on the similarity of patients’ illnesses by processing consultation texts using word embedding algorithms and cosine similarity. Subsequently, the LDA topic model and weighted Jaccard similarity coefficient are used to analyze doctor expertise information and similar patient texts, expanding the pool of candidate doctors based on both doctor expertise and patient illness similarity, thereby resolving the cold-start problem. Finally, a preference disaggregation-based doctor selection decision model is constructed, extracting both group and individual preferences from review texts to identify key attributes influencing patient choice and compensatory relationships among them, thereby enabling personalized doctor recommendations. An empirical study on four online consultation platforms confirms the effectiveness of the proposed method.