Emergent climate protection strategies in German hospitals: A cluster analysis

PLOS One | https://doi.org/10.1371/journal.pone.0312661 May 16, 2025 5 / 18 A. Socio-demographic data of respondents: This section captured age, gender, and working experience in hospital administrative leadership. B. Characteristics of hospitals: This section gathered information on the hospital´s size, ownership, and level of care. C. Climate action implementation: Focused on collecting data regarding the extent to which and in which areas hospi- tals have already implemented climate actions and collected performance indicators. D. Climate protection initiatives and projects: Included questions about the hospital´s employees´ engagement in adopting climate protection initiatives. E. Organization: Explored the hospitals´ organization concerning climate protection, including strategic planning, requirements for suppliers, and the allocation of human resources for coordination and management. F. Communication: Investigated the communication practices of hospital administrative leadership regarding climate protection, including communication about carbon footprint performance indicators, opportunities to contribute ideas for implementing climate actions, and regular meetings for climate protection. Questions in sections two to five were formatted to allow binary responses (yes/no) and ratings on a Likert scale, ranging from 1 (strongly disagree) to 7 (strongly agree). A Likert scale is predicated on the assumption that personal assessments can be quantitatively measured [65] and was used to analyze the distribution of responses and to group the hospitals.

Pretests were performed with five hospital administrative leaders to evaluate the questionnaire´s practicability, com- pleteness, and comprehensibility. These tests were also instrumental in identifying and rectifying any potential errors within the online survey. As a result, some questions were restructured and rephrased. The questionnaire was designed using an online survey tool (LimeSurvey version 6.4.1). Data analysis Upon completion of the survey, the data set of 288 individual responses was screened to ensure the validity of the responses, identify missing values, and recognize outliers [66]. Eighty-three responses were excluded from the analysis due to incomplete data, ensuring that only fully completed questionnaires were considered. The data set was transferred to IBM SPSS Statistics version 22, cleaned, and further analyzed. To provide a comprehensive overview, the demographic characteristics of the respondents and the hospitals characteristics were described using descriptive statistics. Cluster analysis A cluster analysis with inferential tests was conducted. The goal was to identify a grouping wherein the characteristics within each cluster are similar yet distinct across clusters [67]. A hierarchical algorithm based on Ward’s method, which is known to be appropriate for small data sets, was applied [68] to distinguish homogeneous clusters reflective of hospital climate protec- tion strategies in organization and communication. To identify and, if necessary, exclude outliers, the single-linkage method

(nearest neighbor) was employed [69]. The number of clusters was identified using the elbow method, which provides visual support for the cluster decision [70,71]. Based on this analysis, a four-cluster solution was chosen. Subsequently, the four clusters were empirically examined to compare the differences in means regarding their strategies for climate actions. The Bonferroni post-hoc test confirmed that all variables significantly contribute to the differentiation of the four clusters. Inferential statistics Consistent with other studies [72,73], a multivariate analysis of variance (MANOVA) was used to assess whether the clus- ters differ significantly across multiple variables (see Table 3), providing statistical validation for the four-cluster solution.