Challenges to Dialysis Treatment during the COVID-19 Pandemic: A Qualitative Study of Patients’ and Experts’ Perspectives.

Hereby I present the content analysis of interviews of this project and the natural language processing technique results. We used Atlas.ti version 8 for analysis and coding. We supported our content analysis of interviews with a natural language processing technique called Latent Dirichlet Allocation (LDA) to support the thematic analysis. LDA is an unsupervised, generative, probabilistic topic modelling technique that extracts meanings of a pre-defined number of topics/concepts. The number of concepts resulted from the clearest differentiation in the representation of heat maps and in examination of words (topic coherence). For this purpose, we first created a semantic space by stemming the words, removing stop words (such as 'and', 'the', 'a' and similar) and converting all text to lowercase. In addition, punctuation marks, names and personal words were removed from the text. LDA analyzes texts by considering word frequency in combination with the co-occurrence of words. LDA characterizes concepts based on word frequency and the words that best distinguish one concept from others. Separate analyses was conducted for physicians, nurses, HD patients and PD patients, assuming that these four groups had different experiences during the two COVID-19 waves. R software was used to conduct the LDA.