Background The Hospital Survey on Patient Safety Culture (HSOPSC), originally developed in the USA, is used worldwide to assess patient safety culture in hospitals. A limited number of studies have provided data on psychometric properties outside of the US healthcare system. Our aim was to determine if all 12 dimensions of the US HSOPSC were applicable, valid and reliable to Croatian healthcare workers.

Methods The study was conducted from September 2010 to April 2011. Questions were translated into Croatian and then translated back into English. The questionnaires (ie, the Croatian translation of the US HSOPSC) were distributed in unmarked envelopes along with a consent form to all the doctors and nurses in four Croatian hospitals. The responses were analysed using explorative factor analyses, reliability testing, and confirmatory factor analyses.

Results The study included 561 healthcare workers in four Croatian hospitals—a response rate of 32.69%. Our results are similar to the original US sample, but with some differences: 11 dimensions with acceptable reliability scores were identified by exploratory factor analysis compared with the original 12 in the US sample; five of 12 dimensions had a Cronbach's α higher than 0.7, suggesting a reasonable fit to the original US HSOPSC; the dimensions ‘Staffing’ and ‘Organisational learning—continuous improvement’ were found to have a Cronbach's α <0.6. The use of confirmatory factor analysis confirmed a good fit to the original US model.

Conclusions Results show that the Croatian translation of the US HSOPSC is compatible in 11 of the original 12 dimensions. Results suggest that for the purposes of research in Croatia, the dimensions ‘Staffing’, ‘Communication openness’, and ‘Organisational learning—continuous improvement’ should be revised. For example, the use of question A7 (‘We use more agency/temporary staff than is best for patient care’) in the context of European healthcare systems should be adapted or removed for the Croatian version of the US HSOPSC questionnaire.