Patient and Caregiver Insights from Social Media into the Disease Burden of Myelodysplastic Syndrome With A Sub-Cohort View of High-risk Patients (Preprint)
Bibliographic record
Abstract
BACKGROUND Social media platforms offer valuable insights into the patient’s experience, revealing organic conversations that reflect their immediate concerns and needs. Through active listening to lived experiences, we can identify unmet needs and discover real-world challenges patients and caregivers face. OBJECTIVE This study aimed to develop a reusable framework to collect and analyze evolving social media data, capturing insights into the experiences of individuals with MDS and higher-risk myelodysplastic syndromes (HR-MDS) and their caregivers. The findings can inform the development of appropriate patient support interventions. METHODS We conducted an extensive Google search to identify social posts of interest using validated URLs and keywords on English-language websites relevant to MDS. The search covered the period from 1/1/2008 to 12/31/2022. We utilized scraping algorithms to collect, clean, and standardize pertinent information. To classify the perspective of each experience as either that of a patient or caregiver, we employed classification algorithms. This involved contextualizing and summarizing all user posts, followed by decision tree tagging to assign them to the patient or caregiver category. Advanced algorithms were employed to analyze the semantic and temporal structure of the data. Patients or caregivers were categorized as HR-MDS based on contextual mentions of high-risk in their posts or specific factors aligned with NCCN guidelines (e.g., blast percentage, transplantation, use of high-intensity chemotherapy or hypomethylating agents, or disease progression). Each post was assigned major themes and sentiments using a supervised classification machine learning model. Additionally, we employed a semi-supervised machine learning approach for the identification of latent themes in the data corpus. RESULTS The data collected comprised approximately 5.5 million words from 42,000 posts across 5,500 threads, involving about 4,000 users predominantly from the US, UK, and Canada. Out of the 1,249 users classified as HR-MDS, 588 (47%) were patients and 661 (53%) were caregivers. Among the HR-MDS users, the predominant sentiments included concern (78%), anxiety (60%), frustration (58%), fear (58%), and confusion (49%). Concern was the predominant sentiment expressed by caregivers (n=971, 59%), and anxiety by patients (n=752, 55%). Common concerns were specifically related to blood counts (n=677, 54%), burden of the disease (43%), QoL (36%), available treatment options and effectiveness (31%), and disease progression and prognosis (31%). Anxiety related to health and disease (48%), treatment (26%), and the diagnostic process (20%) were also common. The most common sentiments related to fear were the potential development of health complications and the manifestation of symptoms (19%) and the progression and exacerbation of MDS (19%). Additionally, confusion was pervasive among participants, with 295 (24%) individuals finding it challenging to comprehend the nuances of MDS and its diagnosis. A systematic analysis of the principal domains for which information is being sought about HR-MDS revealed frequent mention amongst users of acquiring information on therapeutic intervention (19%), and an interest in ongoing research associated with the disease (17%) CONCLUSIONS The application of sophisticated NLP techniques demonstrates promise in effectively identifying the emerging complex themes and sentiments experienced by HR-MDS users, thereby highlighting the unmet needs, barriers, and facilitators associated with the disease. CLINICALTRIAL NA
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Direct model labels (unvalidated)
Per-model category and study-design labels from the labeling rounds. They are machine output, unvalidated, and the disagreement between models ships as data. No study design here is MEDLINE-validated yet.
| Model arm | Categories | Study design | Confidence |
|---|---|---|---|
| gemma | no category Domain: not available · Genre: Empirical About the Canadian research system: no · About a Canadian topic: no | Observational | low |
| gpt | no category Domain: not available · Genre: Empirical About the Canadian research system: no · About a Canadian topic: no | Observational | low |
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.000 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.003 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedLabeled directly by 2 models reading the full record.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".