Semi-automated socio-anthropologic analysis of the medical discourse on rheumatoid arthritis: Potential impact on public health

Abstract

Background The debilitating effects of noncommunicable diseases (NCDs) and the accompanying chronic inflammation represent a significant obstacle for the sustainability of our development, with efforts spreading worldwide to counteract the diffusion of NCDs, as per the United Nations Sustainable Development Goals (SDG 3). In fact, despite efforts of varied intensity in numerous directions (from innovations in biotechnology to lifestyle modifications), the incidence of NCDs remains pandemic. The present work wants to contribute to addressing this major concern, with a specific focus on the fragmentation of medical approaches, via an interdisciplinary analysis of the medical discourse, i.e. the heterogenous reporting that biomedical scientific literature uses to describe the anti-inflammatory therapeutic landscape in NCDs. The aim is to better capture the roots of this compartmentalization and the power relations existing among three segregated pharmacological, experimental and unstandardized biomedical approaches to ultimately empower collaboration beyond medical specialties and possibly tap into a more ample and effective reservoir of integrated therapeutic opportunities. Method Using rheumatoid arthritis (RA) as an exemplar disease, twenty-eight articles were manually translated into a nine-dimensional categorical variable of medical socio-anthropological relevance, relating in particular (but not only) to legitimacy, temporality and spatialization. This digitalized picture (9 x 28 table) of the medical discourse was further analyzed by simple automated learning approaches to identify differences and highlight commonalities among the biomedical categories. Results Interpretation of these results provides original insights, including suggestions to: empower scientific communication between unstandardized approaches and basic biology; promote the repurposing of non-pharmacological therapies to enhance robustness of experimental approaches; and align the spatial representation of diseases and therapies in pharmacology to effectively embrace the systemic approach promoted by modern personalized and preventive medicines. We hope this original work can expand and foster interdisciplinarity among public health stakeholders, ultimately contributing to the achievement of SDG3.

Large-scale analysis of the medical discourse on rheumatoid arthritis: complementing a socio-anthropologic analysis

“…Observations on the prevalence of such topics (Table 1, col 4) confirm once more that the above hierarchy (conventional, innovative, non-pharmacologic) is mirrored within categories, with a dominance of PHA over EXP over USTD, and topic proportion (Table 2), where we observe that Topic 6 dominates (39 documents), followed by specialization in Topics 5, 4, 9, 7 concerned mainly with innovation with 30, 26, 25 and 21 documents respectively, followed by Topic 10 (16 documents) including acupuncture and VNS, and Topic 2 (13 document, laser), finally followed by Topics 3, 1, 8 (11, 10, 9 documents, respectively). When compared to our previous work [10] we can observe that several of the subcategories we defined a priori (reported in Table 3 for convenience) have indeed been automatically identified. In particular: Dys (dysbiosis, Topic 4) and Diet (Topic 7).…”

“…The starting point to identify the materials was the one described in [10]. Briefly, the selection emerged from a literature search on PubMed performed by three-tiers queries using Medical Subject Headings (MeSH [13]).…”

“…The algorithm automatically identifies the number of topics in Semanticase, provided a range is given. Here also, human-in-the-loop served to identify the most suitable range, which was set to [10][11][12][13][14][15][16][17][18][19].…”

“…Interestingly VNS, EL and AP are all merged under Topic 10. Additional topics (that were not discussed in [10]) are concerned with nanotechnology (Topic 5) and fever (Topic 1).…”

“…To overcome this, topic (Vayansky and Kumar, 2020) and large language models (LLM, [11]) represent effective approaches to perform textual analyses on a large amount of data. Therefore, we present an application of these approaches to explore the totality of the material selected in [10], i.e. 204 articles (from now on the corpus, see Materials and Methods and Table S3).…”