Listening to unhappy voices: A topic modeling analysis of negative hospital reviews in Vietnam

Patient dissatisfaction is a critical indicator of healthcare service performance, yet research relies heavily on predefined survey measures such as SERVQUAL, which constrain evaluation to fixed dimensions and underrepresent the operational and administrative drivers of patient complaints. This study pairs a targeted scoping review of 40 studies, synthesized through the Theory–Context–Characteristics–Methodology (TCCM) framework, with Latent Dirichlet Allocation (LDA) topic modeling of 8,224 negative reviews of 20 hospitals in Ho Chi Minh City, Vietnam. A nine-topic solution, selected on coherence and interpretability, shows that dissatisfaction turns less on clinical outcomes than on how care is delivered: poor staff attitude and communication (15.1%), unsatisfactory clinical encounters (13.5%), and access and parking barriers (12.1%) lead, while waiting time, diagnostic delay, and overcrowding together account for 31.6%. Read against Johnston’s satisfier–dissatisfier classification, 42.1% of reviews concern predominantly dissatisfying determinants, and parking, insurance administration, and system capacity have no counterpart in the dominant instruments. The study reframes dissatisfaction as a process-based construct and gives hospital quality managers and administrators a low-cost diagnostic that locates where delivery breaks down and feeds directly into improvement cycles.