JMIR Publications

ABSTRACT

Background:

Monitoring social media has been shown to be a useful mean to capture patients’ opinions and feelings about medical issues, ranging from diseases to treatments. Health-related quality of life is a useful indicator of overall patients’ health that can be captured online.

Objective:

This study aims to describe a Social Media Listening system which is able to detect any impact of diseases or treatments on health-related quality of life as reported in social media and forum messages written by patients.

Methods:

Using a web crawler, 19 health-related forums in France were harvested and messages relating a patient’s experience with a disease or a treatment were specifically collected. The algorithm was based on the two clinically validated questionnaires SF-36 and EQ-5D. Models were trained using cross-validation (a machine learning technique which obtains the best combination between different data samples) and hyperparameter optimization. Over-sampling was used to increase the infrequent dimension: after annotation, SMOTE was used to balance the proportion of the dimension among messages.

Results:

The training set was composed of 1400 messages, randomly taken from a 20 000 batch of health-related messages coming from forums. The algorithm was able to detect a general impact on health-related quality of life (sensitivity of 0.83 and specificity of 0.74), a physical impact (0.67 and 0.76), a psychic impact (0.82 and 0.60), an activity-related impact (0.73 and 0.78), a relational impact (0.73 and 0.70) and a financial impact (0.79 and 0.74).

Conclusions:

Real-time assessment of patients’ health-related quality of life through the use of Social Media Listening is useful to a patient-centered medical care. Social media as a source of Real World Data are a complementary point of vue to understand patients’ concerns, unmet needs and how diseases and treatments can be a burden in their daily lives. Trial Registration: Not applicable (not a trial)