Background <p>Information about the reasons why patients return after a dental extraction could be used in quality-of-care indicators in dentistry, but these reasons are typically not recorded in a structured form in dental records. Our aim was to determine whether information in dental clinical notes could be used to help identify which patient returns were unplanned and due to complications.</p> Materials and methods <p>We manually annotated the clinical notes for 4,756 return visits occurring within 14 days of a dental extraction, separating unplanned returns due to complications from unplanned returns for other reasons and planned return visits. Four pre-trained language models (PLMs) and a bag-of-words logistic regression model were tested and performance was measured by F1-score. SHapley Additive exPlanations (SHAP) values were used as a post-hoc explainability analysis.</p> Results <p>When classifying unplanned returns due to complications versus all reasons for return, the best performing PLM produced an F1-score of 0.82. No significant difference in performance was found across the four PLMs. The performance of the bag-of-words logistic regression model was significantly lower (F1-score of 0.71). The SHAP analysis suggested that notes related to alveolar osteitis and post-operative infection were most predictive of unplanned returns following a dental extraction.</p> Conclusions <p>NLP methods can be used to distinguish between unplanned returns due to complications from dental clinical notes. The results suggest that dental clinical notes could be used to improve measures of unplanned returns as a quality-of-care indicator.</p>