<![CDATA[Patient feedback represents an important form of healthcare communication through which patients articulate experiences, evaluations, and expectations toward healthcare services. In practice, this communication is often conveyed through unstructured, subjective, and multidimensional narratives, in which a single message may simultaneously address multiple service aspects. Such characteristics complicate the systematic interpretation of patient communication, particularly when sentimental expressions are unevenly distributed and dominated by positive evaluations. This study aims to examine patient feedback as a communicative practice in healthcare by analyzing multidimensional sentiment expressions from an interdisciplinary communication perspective. Computational methods are not positioned as the primary contribution of this study, but are employed as analytical tools to support the interpretation of large-scale patient communication data. An aspect-based sentiment analysis framework with a multilabel classification scheme is used to capture how sentiments are communicated toward predefined service aspects. The dataset consists of 1,131 anonymized patient feedback texts collected from JIH Purwokerto Hospital. To reduce interpretive bias caused by imbalanced sentiment distributions that may obscure less explicit communication expressions, label-based data balancing strategies are applied. Indonesian language modeling is used to accommodate the informal and context-dependent characteristics of patient narratives. The findings indicate that this approach enables a more structured reading of patient communication across service aspects, particularly in identifying explicit positive and negative sentiments. In contrast, neutral sentiment remains more difficult to identify due to its implicit and contextual nature, reflecting the complexity of patient communication strategies. Overall, this study demonstrates that computational analysis can function as a supportive instrument in healthcare communication research for systematically mapping multidimensional patient feedback, provided that the results are interpreted contextually rather than mechanically.]]>
