Abstract Background Open-ended comments in Patient Safety Culture (PSC) instruments can provide valuable, context-specific insights into staff perceptions of safety culture and help identify salient organizational issues. However, few studies have systematically analyzed these comments at scale, and existing evidence is often limited by small sample sizes and limited geographic and cultural diversity, as analyses are typically restricted to individual hospitals or single countries. The goal of this study is to assess the feasibility of large-scale automated topic modeling and sentiment analysis of open-ended PSC comments, and to identify key recurring topics and sentiments across hospitals, countries, and survey years. Methods We study 13,752 open-ended comments from a PSC survey administered to more than 112 hospitals in Peru, Chile, Colombia, Argentina, and Honduras between 2018 and 2022. An automated topic classification was implemented using the Bidirectional Encoder Representations from Transformers (BERT) model. The RoBERTuito language model was also used for sentiment analysis in Spanish. Finally, to analyze the variability of topics and sentiments, the probabilities predicted by the marginal effects of a logit model were used. Results A total of 27 topics were identified, with Patient Safety (23.97%) being the most prominent. Sentiment analysis indicates that many healthcare personnel made positive comments on key topics such as Survey (45.85%) and Patient Safety (28.21%). On the other hand, negative and neutral comments were mainly associated with Nursing (100%) and Infrastructure (99.33%). Regarding the variability of topics and sentiments, no significant differences were observed between versions, years and countries in most topics. Conclusion This study suggests that open-ended questions can provide valuable and meaningful qualitative insights into PSC. Our study represents a useful tool to administrators, not only because of its large sample size, but also for its multicultural richness, reflecting different organizational cultures of hospitals from five different Latin American countries.
Arrieta et al. (Mon,) studied this question.