Abstract
This study investigates common health problems and online healthcare seeking behavioral patterns of individuals using a large-scale online consultation dataset extracted from well-known healthcare platforms. By using text mining techniques, including tokenization, part-of-speech tagging, named entity recognition, sentiment analysis, and topic modeling, we analyzed the textual content of these consultations to identify key themes and patterns. Our findings suggest that there are significant gender differences in online medical counseling, with women more likely to seek medical advice online than men. Additionally, the analysis identified age-related patterns of health concerns, with younger people primarily reporting minor illnesses, while older people expressed concerns about chronic conditions. These results highlight the importance of understanding the unique healthcare needs of different populations and the potential role of online platforms in meeting those needs.
Introduction
Electronic Health Records (EHRs) are fundamental to modern medicine but often fail to capture the "lived experience" of patients—the nuances of how individuals describe symptoms and concerns before clinical intervention. This patient-authored text (ePAT) contains vital public health data, yet its unstructured nature makes it difficult to analyze. This research addresses the challenge of extracting structured, actionable insights from large-scale online medical consultations to map a "disease spectrum" and understand demographic-specific health trends.
Methods
The study employed a multi-stage Natural Language Processing (NLP) pipeline:
- Data Acquisition: A custom web crawler collected public Q&A data from major online healthcare platforms.
- Preprocessing: Data was cleaned to remove noise (ads, system notifications) and standardize formats.
- NLP Analysis:
- Word Segmentation: Used the Jieba library for efficient tokenization of complex characters.
- POS Tagging: Filtered for medically relevant nouns and adjectives to improve precision.
- Named Entity Recognition (NER): Categorized terms into "diseases," "medications," and "body parts."
- Stratified Analysis: Frequency analysis and word clouds were used to compare health concerns across age and gender demographics.
Results and Discussion
The Gender Gap in Digital Health
Analysis revealed a significant gender imbalance in online health-seeking behavior. A majority of users (57.43%) were female, compared to 42.57% male. This suggests that women are more proactive in perceiving and reporting health concerns, a behavior that correlates with higher life expectancy.
Age-Related Disease Spectrum
The study identified a clear transition in health priorities across the lifespan. Younger populations (under 20) primarily sought advice for acute minor illnesses like colds and coughs. In the 20–30 age bracket, concerns shifted toward reproductive health and stress-related issues, alongside a notable peak in hair loss inquiries, likely driven by aesthetic concerns.
As users aged into the 30–60 range, chronic conditions such as hypertension and diabetes began to emerge. For those over 60, cardiovascular diseases became the dominant concern.
Digital Triage and Longevity
Online platforms appear to serve as a "minor ailment" filter for younger users, potentially alleviating pressure on physical emergency services. Furthermore, the higher engagement of women in these digital spaces aligns with the observed life expectancy advantage women hold (approximately 5.4 years in the US context).
Limitations and Future Work
While providing valuable insights, the study has limitations. Healthcare-seeking behavior is culturally and geographically dependent, meaning findings may not generalize globally. Additionally, the research tracks inquiries rather than clinical outcomes. Future work should investigate whether online consultations lead to improved health metrics and explore cross-cultural datasets to understand how regional norms influence digital health engagement.
Summary updated