Data Analysis

In Phase A, the quotations were first identified as expressions of a particular factor and further categorized into the sub-themes. To increase inter- and intra-coder reliability, the authors performed the analysis of the transcribed interviews independently. An inductive content analysis was performed using Mayring’s principles as also exploited by others.^40,41 From the research question, a theoretical based definition of the aspects of analysis, main categories including sub categories were formulated, which leads to coding rules of those categories. The categories were collected in an open coding agenda using the software Quirkos (www.quirkos.com), and eventually revised and formative checked for reliability within the coding process in repetitive discussions until consensus was reached (investigator triangulation). Thus, individual phases and passages summarized into one code. The final working through the texts were again checked for reliability. Prototypical text passages, categorical definitions, and rules to discern the categories were formulated. The material was completed systematically step by step and revised within the process of analysis. Using this methodological approach, the authors followed quantitative inquiry approaches, which is also the cornerstone of grounded theory.⁴² However, instead of generating a formal theory, the intention of this qualitative content analysis was to use the results to develop a protocol, which was incorporated into a questionnaire of a quantitative inquiry (Phase B).

The authors developed a quotation tree during the transcription of the interviews, consisting of three main categories (Figure 1 Supplement):

Likert scale questions (ranging from 0=decline/do not know to 7=completely agree), questions with a percent scale from 0 to 100 and questions with the option of three answers (do not know, false, fully agree) were used. An item was considered a “firm perception” when the mean response was within one-third of the lowest/highest possible answer scores. Statistical analysis was performed in the quantitative study part using SPSS (V27). Only complete data sets were analyzed. Data sets being prematurely terminated by the students were omitted. Descriptive statistics were used to summarize student demographic data. Cronbach’s alpha was calculated to assess internal consistency of categories of professional behavior in the students’ choice paragraph (Table 2). Mann–Whitney-U-Tests were used to compare the responses relating to perceptions. Group comparators were curriculum type, gender and semester levels. Pearson correlation coefficient was calculated to compare students’ socio-humanistic attitudes towards their intention to a career in medicine. For the determination of work schedule preferences between student groups, the Chi-Square test was used. A p-value of less than 0.05 was considered statistically significant.

Incentive to Study Medicine and Vision