Patterns of discussion on neuroticism and self-management behaviors in type 2 diabetes: a scoping review using machine learning-assisted text mining

Wei Sun, Dong Xie, Menglin Zhang, Wen-hui Hou, Zhixiang Wang, Chendi Wang, Guodong Xu, Yang HeFrontiers in Public Health 2025 · DOI

Abstract: Unique Identifier: 10.17605/OSF.IO/54NJD; publicly accessible URL: https://doi.org/10.17605/OSF.IO/54NJD.

Introduction

Managing type 2 diabetes mellitus (T2DM) requires a rigorous, lifelong commitment to self-care behaviors, including dietary control, regular physical activity, and frequent blood glucose monitoring. Despite the critical nature of these tasks for preventing complications, patient adherence remains a significant clinical challenge.

Psychological research indicates that personality traits—specifically neuroticism, characterized by emotional instability, anxiety, and stress sensitivity—play a pivotal role in how individuals navigate chronic illness. However, the existing literature on the intersection of neuroticism and T2DM self-management is often fragmented and descriptive. This scoping review addresses this gap by mapping the research landscape, identifying how researchers conceptualize these links and where knowledge gaps persist.

Methodology

Following the PRISMA-ScR guidelines, the researchers conducted a systematic search across seven major databases, including PubMed and Scopus. The study distinguishes itself through a hybrid analytical approach that integrates traditional qualitative synthesis with a machine learning (ML) text-mining pipeline:

  1. Natural Language Processing (NLP): Text from included studies underwent tokenization, lemmatization, and stop-word removal.
  2. Latent Dirichlet Allocation (LDA): This topic modeling technique uncovered five latent thematic structures within the corpus.
  3. BERT (Bidirectional Encoder Representations from Transformers): Semantic embedding and clustering were used to identify contextual nuances and conceptual proximities beyond simple word-frequency analysis.
  4. Keyword Co-occurrence Networks: Relationships between psychological constructs (e.g., stress) and behavioral domains (e.g., medication adherence) were mapped using TF-IDF weighting.

Literature Selection and Dataset

The primary dataset consisted of the scientific literature itself. The initial search yielded 699 records, which were screened down to 10 original quantitative or mixed-method studies. These studies spanned six countries: China, USA, Mexico, Bosnia and Herzegovina, Iran, and Australia.

Figure 1: PRISMA flowchart illustrating the identification, screening, and inclusion stages of the scoping review.

The resulting ML corpus comprised 1,933 valid sentences and 23,738 effective tokens extracted from the final selection of papers.

Key Insights and Thematic Analysis

The "Semantic Bridge"

Analysis revealed that neuroticism serves as a conceptual "semantic bridge" in the literature, linking emotional reactivity (stress and anxiety) to behavioral regulation (adherence).

Vigilance vs. Reassurance

Neuroticism was most strongly associated with blood glucose monitoring. The authors suggest that for individuals high in neuroticism, the immediate, quantifiable feedback provided by monitoring satisfies health vigilance and a need for reassurance.

The Diet Paradox and the Exercise Gap

Dietary control was the second most discussed domain. The literature suggests a dual effect: while high neuroticism can trigger "emotional eating" under stress, it can also manifest as meticulous regulation due to heightened health awareness. Conversely, physical activity was the least discussed domain, representing a significant research gap likely due to the difficulty of measuring long-term exercise adherence compared to discrete tasks like glucose testing.

Table 1: Characteristics of included studies, detailing participant demographics, assessment tools, and main conclusions regarding neuroticism and self-management.

Quantitative Results

  • Topic Modeling: The LDA analysis identified five distinct clusters: Methodological discourse, Foundational personality constructs, Emotional/Cognitive aspects, Behavioral adherence, and Comparative personality traits.
  • Frequency Weights: Within the behavioral domain, "adherence" held the highest frequency weight (0.0506). In the personality domain, "neuroticism" (0.0240) and "conscientiousness" (0.0232) were weighted similarly, indicating they are frequently analyzed as a pair.
  • Co-mention Patterns: The most frequent co-mentioned pair was "adherence–medication" (280 occurrences), while "medication–neuroticism" appeared 131 times, highlighting a strong discursive proximity between personality and pharmacological compliance.

Conclusion and Future Directions

While this review successfully maps the patterns of discussion, it is limited by the small number of available studies (n=10) and the descriptive nature of the current literature. The wide variety of measurement tools used across studies also complicates direct quantitative comparisons.

Future research should move toward hypothesis-driven longitudinal studies to establish causal pathways between personality traits and long-term T2DM outcomes. Furthermore, the authors advocate for "personality-informed" clinical interventions, where support strategies are tailored to a patient's specific level of neuroticism to improve overall self-management adherence.

Summary updated

Abstract

Unique Identifier: 10.17605/OSF.IO/54NJD; publicly accessible URL: https://doi.org/10.17605/OSF.IO/54NJD.
BibTeX
@inproceedings{sun2025patterns,
  author = {Wei Sun and Dong Xie and Menglin Zhang and Wen-hui Hou and Zhixiang Wang and Chendi Wang and Guodong Xu and Yang He},
  title = {{Patterns of discussion on neuroticism and self-management behaviors in type 2 diabetes: a scoping review using machine learning-assisted text mining}},
  booktitle = {Frontiers in Public Health},
  year = {2025},
  doi = {10.3389/fpubh.2025.1708967},
}
Figure