*2.4. Data Analysis*

We used a conventional approach to qualitative content analysis [31]. In conventional content analysis, categories are derived from the data rather than from theory. It is generally applied in study designs that aim to describe a phenomenon about which little theory or literature is yet available.

Out of 217 respondents, 63 answered all open-ended questions, 143 answered at least one, and 11 did not fill in any free text field. Many of the answers were very detailed. We received only a few keyword-like responses, as is usually the case with open-ended survey responses. In sum, we received 1448 free text comments: 308 describing changes in the therapeutic relationship, 275 describing changes regarding the content of sessions, 265 describing changes in intensity, 238 describing changes in the structure of sessions, 192 focusing on experiences regarding the lack of physical presence in remote psychotherapy and 170 addressing the spatial distance. Overall, responses to Q1 were the most comprehensive. Respondents addressed various aspects of their own accord that not only had to do with the therapeutic relationship but also related to the subsequent questions. As a result, some answers were repeated in later questions. They were only coded if new aspects were addressed.

At the beginning of the coding process, two coders read through the whole data set to familiarize themselves with the material. Subsequently, data were imported into Atlas.ti for coding [32], and one coder read through the material again, inductively defining categories in the process. After coding 30% of the material, the second coder coded the same material with the list of categories and category definitions provided by the first coder. To enhance reliability [33], we assessed the agreement of how the two coders coded the data set [34]. Percentage agreement was high at 94.3%, and inter-coder agreement using Krippendorff c-α-binary = 0.985. Any citation on which the coders disagreed was discussed between the two coders, and the category definitions were expanded in this process. In addition, this step of the coding process created larger thematic clusters to which categories were assigned. Afterward, the second coder coded the entire data set, documenting the cases in which assignment to a category was not clear. These cases were coded together.

Chi-squared tests were conducted to analyze potential differences in the frequency of main categories reported by psychotherapists in terms of sociodemographic (years of age: ≤40, 41–50, 51–60, >60; gender: female, male) and professional characteristics (years in the profession: ≤5, 6–10, >10; psychotherapeutic orientation: psychodynamic, humanistic, systemic, behavioral). Differences in the frequencies of subcategories were only analyzed with respect to gender, as the number of coded text passages in the subgroups of the different orientations, age and experience groups was insufficient to make reliable inferences about the population of psychotherapists. To analyze differences in the length of the free text answers, *t*-tests (gender) and univariate ANOVAs (age group, professional experience group, psychotherapeutic orientation) were applied. Statistical analyses were

performed in SPSS version 26 (IBM Corp, Armonk, NY, USA). *p*-values of ≤0.05 were considered statistically significant (2-sided tests).
