Abstract
This cross-sectional study investigates how menopause-related symptoms discussed in user-generated online posts compared with symptoms documented in electronic health record clinicial notes.
Introduction
The transition to menopause presents with a series of symptoms that affect women’s physical, cognitive, and emotional well-being.1 Although clinical data has documented menopausal symptoms,2 research suggests these data sources may inadequately capture the full scope of patient-reported experiences. Social listening from online forums can provide authentic narratives of health experiences, especially for women,3 often shared in unaltered and lived contexts4 and can play a role in identifying silent needs, underreported concerns, and unnoticed symptoms,5 particularly in stigmatized or poorly understood conditions.6 With this in mind, this cross-sectional study asked how do menopause related symptoms discussed in user-generated online posts compare with symptoms documented in electronic health record (EHR) clinical notes?
Methods
To examine discrepancies between clinical documentation and online patient experiences, we conducted an analysis of menopause symptoms using 2 data sources: (1) a randomly selected sample of the 2 053 715 records in the University of California, San Francisco’s (UCSF) EHR clinical notes, filtered for the substring menopause, and (2) 999 all-time top posts from the Reddit menopause forum, retrieved using the site’s application programming interface (API) and ranked by upvotes.7 The site limits data scraping to 999 posts. We analyzed post titles and self-text only, excluding comments. We selected the top all-time posts to capture themes that were most visible and resonant within the online community. Data were first explored using latent Dirichlet allocation. The resulting unsupervised latent Dirichlet allocation topics were thematically adjacent and not distinct enough to serve as final categories. Therefore, the extracted topic keywords were used as guidance to inform a prespecified taxonomy of medically recognized symptom domains.8 To confirm that the taxonomy captured the range of content represented in both datasets, we reviewed a random sample of posts and clinical notes. We further prompted UCSF’s large language model Versa’s application programing interface gateway, a Health Insurance Portability and Accountability Act–compliant artificial intelligence chat tool that can be securely used with restricted and sensitive UCSF data, to assign each clinical note and online post to the relevant domain(s) based on the representative keywords.7 Records were classified as excluded by the large language model and removed before analysis if they did not reflect menopause-related personal experiences, questions, or discussions. This yielded analytic samples of 577 posts and 646 EHR clinical notes. A separate manual annotation of the final large language model labels was not conducted, but we instead focused on iteratively refining the taxonomy and prompt to support consistent domain assignments. Percentage distributions by domain were calculated for posts and EHR notes. Differences in domain proportions between data sources were assessed using χ2 tests, with statistical significance was 2-tailed and defined as P < .05. Comparison domains included sleep disturbances, hormonal skin and hair changes, and bone and joint health, as these commonly reported symptoms are expected to appear at similar frequencies in both traditional clinical datasets and online forums. Statistical analysis was performed from July to November 2025 using Python, version 3.9.21 (Python Software Foundation). This research is considered nonhuman participant research due to deidentified nature of the data and used the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies.9
Results
The Table provides the differences in distribution of discussion between online forum and EHR notes. Cognitive impairment, emotional well-being, and weight change were more frequent in online forum posts than in EHR clinical notes. Hot flashes and night sweats, hormone replacement therapy treatments, lifestyle changes, sexual health, non–hormone replacement therapy menopause treatments, herbal remedies, and uncategorized content were more documented in EHR notes than posts. No statistically significant differences were observed for the comparison domains of sleep disturbances, hormonal skin and hair changes, and bone and joint health.
Table. Topic Distribution in Online Forum Posts vs Electronic Health Record Clinical Notes About Menopause.
| Topica | No. (%) | χ2b | P valueb | |
|---|---|---|---|---|
| EHR Clinical Notes | ||||
| Emotional well-being | 446 (77.30) | 168 (26.01) | 318.67 | <.001 |
| Hot flashes and night sweats | 107 (18.54) | 301 (46.59) | 106.61 | <.001 |
| Lifestyle changes | 123 (21.32) | 249 (38.54) | 41.93 | <.001 |
| Sexual health | 116 (20.10) | 231 (35.76) | 35.99 | <.001 |
| Cognitive impairment | 113 (19.58) | 31 (4.80) | 62.72 | <.001 |
| HRT treatments | 211 (36.57) | 309 (47.83) | 15.37 | <.001 |
| Non-HRT menopause treatments | 44 (7.63) | 117 (18.11) | 28.40 | <.001 |
| Weight change | 119 (20.62) | 117 (11.30) | 19.32 | <.001 |
| Uncategorized | 4 (0.69) | 45 (6.97) | 29.57 | <.001 |
| Herbal remedies | 7 (1.21) | 44 (6.81) | 22.52 | <.001 |
| Sleep disturbances | 108 (18.72) | 146 (22.60) | 2.56 | .11 |
| Hormonal skin and hair changes | 81 (14.04) | 103 (15.94) | 0.72 | .39 |
| Bone and joint health | 68 (11.79) | 73 (11.30) | 0.03 | .86 |
Abbreviations: EHR, electronic health record; HRT, hormone replacement therapy.
Topics were not mutually exclusive; each post or EHR clinical note could be assigned to multiple domains. A Bonferroni correction across the 13 topic comparisons would not change the statistical significance of the major findings.
χ2 tests comparing proportions; statistical significance was assessed at P < .05.
Discussion
Together, these findings show that patient-generated discussions surface psychosocial and cognitive experiences at several-fold higher rates than clinical documentation, while EHR records emphasize vasomotor symptoms and treatments, underscoring a systematic gap in how menopause-related symptoms are captured. Findings should be interpreted in light of differences in both population representation and data generation, as online forums capture self-selected communities while clinical notes reflect healthcare-seeking individuals whose experiences are filtered through clinical documentation. Therefore, online forums can complement clinical data by illuminating symptoms that patients may not disclose in clinical encounters due to social desirability bias, embarrassment, fear, or stigma.5,10 A multisource approach that combines clinical and digital narratives can inform more responsive, patient-centered care and guide future research into the understudied domains of women’s aging.
Data Sharing Statement
References
- 1.Dennerstein L. Well-being, symptoms and the menopausal transition. Maturitas. 1996;23(2):147-157. doi: 10.1016/0378-5122(95)00970-1 [DOI] [PubMed] [Google Scholar]
- 2.Andrews R, Lacey A, Bache K, Kidd EJ. The role of menopausal symptoms on future health and longevity: A systematic scoping review of longitudinal evidence. Maturitas. 2024;190:108130. doi: 10.1016/j.maturitas.2024.108130 [DOI] [PubMed] [Google Scholar]
- 3.Dhankar A, Katz A. Tracking pregnant women’s mental health through social media: an analysis of reddit posts. JAMIA Open. 2023;6(4):ooad094. doi: 10.1093/jamiaopen/ooad094 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Stair SL, Palmer CJ, Lee UJ. Wealth of knowledge and passion: patient perspectives on vaginal estrogen as expressed on Reddit. Urology. 2023;182:79-83. doi: 10.1016/j.urology.2023.08.040 [DOI] [PubMed] [Google Scholar]
- 5.Im EO, Liu Y, Dormire S, Chee W. Menopausal symptom experience: an online forum study. J Adv Nurs. 2008;62(5):541-550. doi: 10.1111/j.1365-2648.2008.04624.x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Proferes N, Jones N, Gilbert S, Fiesler C, Zimmer M. Studying reddit: A systematic overview of disciplines, approaches, methods, and ethics. Soc Media Soc. 2021;7(2):1-14. doi: 10.1177/20563051211019004 [DOI] [Google Scholar]
- 7.Williams CYK, Zack T, Miao BY, et al. Use of a large language model to assess clinical acuity of adults in the emergency department. JAMA Netw Open. 2024;7(5):e248895. doi: 10.1001/jamanetworkopen.2024.8895 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Martinez AM, Kak AC. Pca versus lda. IEEE Trans Pattern Anal Mach Intell. 2001;23(2):228-233. doi: 10.1109/34.908974 [DOI] [Google Scholar]
- 9.Norgeot B, Muenzen K, Peterson TA, et al. Protected Health Information filter (Philter): accurately and securely de-identifying free-text clinical notes. NPJ Digit Med. 2020;3(1):57. doi: 10.1038/s41746-020-0258-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Phillips E, Hanson S. “She hadn’t told anyone as she was so embarrassed”: embarrassment as a barrier to accessing women’s health care. Open Science Framework . 2025. Accessed May 22, 2026. https://ueaeprints.uea.ac.uk/id/eprint/101535
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Sharing Statement
