## An Epistemic Network Analysis of Patient Decision-Making Regarding Choice of Therapy ###### This OSF project is a proxy for our repository on Gitlab, which is available [here][1]. ### Research aims and questions The research project aims to explore patterns in patient cognition that contribute to choice of therapy: 1) employing only biomedicine (conventional medicine) and 2) employing complementary and alternative medicine (CAM, non-conventional medicine). Our objective was to gain an understanding of phenomenological factors in patient decision-making through investigating how therapy choice and rejecting conventional cures interact with a) push factors (negative experiences), b) pull factors (dispositions), and c) lay theory of illness causation (etiology) in narratives from users of biomedicine only (B) and users of complementary and alternative medicine (CAM) among three illness groups (diabetes, musculoskeletal diseases, and digestive illnesses). ### Sampling We had two patient subsamples: those employing CAM (group name: CAM) and those only employing biomedicine (group name: Biomed). Individuals were included with non-proportional quota sampling stratifying on: therapy choice, diagnosis, and sex. These considerations were vital to ensure cognitive variability in the sample, within the subgroups. ### Data collection We collected data with semi-structured interviews. Our data collection tool consisted of three key domains: Information (how a patient comes across healthcare-related information, how they judge its trustworthiness, etc.), Ontology (concepts of health and illness, explanatory model of illness and etiology, etc.), and Behavior (milestones of the patient journey, decisions and choices of treatment). We also employed a short survey to collect demographic and clinical variables from patients (later referred to as "patient attributes"). Both of these tools are available for scrutiny in our Operationalizations. Data collection was performed in Budapest, Hungary by four researchers all trained in qualitative methods and in using our specific data collection tools. All interviews were audio-recorded, transcribed verbatim and anonymized. ### Data coding We employed deductive coding; codes were adopted from a previous, fully qualitative study on the same topic. Our coding system contained three levels of hierarchy, with a total of 52 low-level codes, 19 of which were used in the present study. Coding was performed by three researchers; each researcher “specialized” in one of the three high-level codes. Coding was preceded by a training period during which subsets of the raw data were coded independently and then triangulated; code interpretations were refined as needed. Subsequently, we created the final version of the codebook, and coded the narrative corpus deductively with those codes. Our codebook is available in our Operationalizations. ### Data segmentation In order to attain meaningful code co-occurrences, interview transcripts were segmented. Sentences constituted the smallest meaningful unit of segmentation in our project (i.e., utterances). Each sentence received a unique utterance identifier (UID); coding occurred on this level of segmentation. Mid-level segmentation (i.e., recent temporal context) was applied in the case of push factors, but not concerning pull factors and etiology. The latter two code clusters attempted to capture the patient’s value system and explanatory model of illness, indicating that more distal co-occurrences (such as a patient judging the efficacy of a treatment by “giving it time” and “tracking how symptoms change”) are valid connections, no matter where they occur in the interview transcript. Push factor codes referred to discrete events along the patient journey, such as experiencing side-effects and expressing dissatisfaction about that, thus, in this case, tighter psychological proximity was warranted. ### Data coding and curation Coding and segmentation were performed in compliance with the [Reproducible Open Coding Kit (ROCK)][2] standard that facilitates transparent qualitative research. We employed [iROCK (a browser-based graphical interface)][3] to code transcripts manually, then, based on the UIDs, we aggregated the coded data with the [rock R package][4]. In this process, discourse coding was converted to binary representation (1 if a code occurs in a particular utterance, 0 if it does not). Thus, each line in our final dataset contained an utterance, the coded data in binary form, and patient attributes. ### Analysis ENA calculates the co-occurrence of each unique pair of codes within a given segment of discourse and aggregates this information for each patient (i.e. unit of analysis) in a cumulative adjacency matrix, which is represented as a vector in a high-dimensional space. Spherical normalization is performed on the vectors, resulting in normalized vectors that quantify the relative frequencies of co-occurrence among codes independent of discourse length. ENA then projects the networks as points into a low-dimensional space. The result is two coordinated representations of the data: 1) network graphs, where the nodes in the model correspond to the codes in the discourse and the edges represent the relative strength of connection among codes, and 2) ENA scores, or the position of each network in the low-dimensional space. To read more about ENA, please see [this resource][5]. We also performed hermenutic analysis via de- and re-contextualization using the rock R package in order to interpret code co-occurrences and describe code manifestations in depth. Additionally, we performed "code-and-count" procedures and compared our results to those modelled with ENA. ### Results There was a marked difference between our two subsamples concerning push factors: although both groups exhibited similar scaled relative code frequencies, the CAM network models were more interconnected, indicating that CAM users expressed dissatisfaction with a wider array of phenomena. Among pull factors, a preference for natural therapies accounted for differences between groups but did not retain a strong connection to rejecting conventional treatments. Etiology, particularly adherence to vitalism, was also a critical factor in both choice of therapy and rejection of biomedical treatments. ### Conclusions Push factors had a crucial influence on decision-making, not as individual entities, but as a constellation of experienced phenomena. Lay etiology is closely associated with choice of therapy; belief in vitalism affects the patient’s explanatory model of illness, changing the interpretation of other etiological factors and illness itself. Scrutinizing individual push/pull factors or etiology does not explain therapeutic choices; it is from their interplay that decisions arise. Our unified, qualitative-and-quantitative methodological approach offers novel insight into decision-making by displaying connections among codes within patient narratives. [1]: https://gitlab.com/matherion/semiotic-network-analysis-of-cam-patient-cognitions [2]: https://sci-ops.gitlab.io/rockbook/ [3]: http://sci-ops.gitlab.io/irock/ [4]: https://cran.r-project.org/web/packages/rock/index.html [5]: https://www.epistemicnetwork.org/