2.6. Quality of Measurements

Professionals were trained in four training sessions to be able to facilitate a nature experience for families in their care and to be able to collect data according to the research protocol. After the first training session, the professionals conducted a tryout of data collection in which they got feedback regarding the consistency with the research protocol. After the second and third training session, data collection occurred. The fourth training session was a closing session with reflection on the results of the study.

To allow the participation of parents with diverse backgrounds and safety concerns, it was necessary that all shelters had safe natural environments on their own property. Each shelter received funding for developing a natural environment, varying from EUR 10,000 (approximately USD 11,080) to EUR 65,000 (approximately USD 72,000). Shelters developed a restorative garden, a natural play area, a children’s farm or a vegetable garden. Shelters were assisted in the development of the natural environments by spatial planners, animal experts, gardeners and construction workers. Data collection started when all shelters had the possibility to use a natural environment. The natural environments that were used in this study are specified in Table A1.

The associations between experiencing nature and parental need frustration and need satisfaction were analyzed using linear mixed model analyses in SPSS Statistics for Windows, version 27 (SPSS Inc., Chicago, IL, USA), accounting for the clustering of the two measurements (level 1) within participants (level 2) within shelter locations (level 3). Although we were not interested in the higher-order effects, we chose to incorporate the shelter location as level 3 to be able to produce more accurate standard errors [32]. The unstandardized coefficients were converted to standardized mean differences (Cohen’s d) [33]. Analyses of effect modification using two-way interaction terms were conducted for each of the potential moderators. Interaction terms with a p-value lower than 0.05 were identified as moderators.