*Qing Zeng, PhD
University of Utah
Oral Presentation – Research Track
Sunday, Sept 30, 2012: 12:48 PM – 1:08 PM – LK120

← Back to Medicine X 2012 Proceedings

*Presenting Speaker

Background
Clinical trials and patient records have been the main information sources for clinical research. While well-designed clinical trials can produce high quality data, they are generally very expensive and time consuming. Further, patients enrolled in clinical trials are not necessarily representative of the intended patient population. Chart reviews avoid some of the drawbacks of the clinical trial approach. However, studies that use chart reviews are limited by the accuracy and completeness of the data in the patient records. In the past decade, online social networks have grown exponentially. We hypothesized that information from online social networks has the potential to serve a new and complementary information source for clinical research. To test this hypothesis, we conducted two separate studies. In the first study, we compared the prevalence of fatigue and depression for patients of amyotrophic lateral sclerosis (ALS), multiple sclerosis (MS) and Parkinson’s disease (PD), as reported on the online social network PatientsLikeMe and a large medical record data repository. In the second study, we compared clinicians’ and patients’ perspectives on the symptomatic treatment of ALS by comparing data from a traditional survey study of clinicians with data from a patient social network.

Methods
In the first study, multivariable logistic regression was performed on the probability of reporting fatigue or depression as predicted by age, gender, data source, type of neurological disease and the interaction of data source and type of neurological disease. We report on the effects of the interaction of data source and type of neurological disease on the probability of reporting fatigue or depression. Our analysis addresses whether the association of reporting fatigue and depression with disease type differs between the two data sources, and, equivalently, whether the association of reporting fatigue and depression with data source varies between disease types. These results are controlled for the effects of age and gender. In the second study, we first extracted the 14 symptoms and associated top four treatments and then selected twenty symptom-treatment pairs to compare the clinicians’ and patients’ perceptions of treatment prevalence and efficacy.

Results
In the first study, overall, both fatigue and depression were more likely to be reported if the data source was PatientsLikeMe regardless of disease. The odds for reporting fatigue and depression were greater from the PLM source across all diseases (i.e. PLM users are more likely to report fatigue and depression). The odds ratio for reporting fatigue was 33.9 for ALS, 36.3 for MS, and 18.7 for PD. The odds of reporting depression were 6.1 for ALS, 9.7 for MS, and 4.91 for PD. In the second study, similarities and discrepancies were found between clinicians’ and patients’ perceptions of treatment prevalence and efficacy. In 10 out of the 20 pairs, the symptom-treatment differences between the two groups were above 10%. In three pairs the differences were above 20%.

Conclusions
Online social network data, reflecting patients’ perspectives, do provide somewhat different information regarding symptoms and symptomatic treatment from the traditional research data sources like survey results and medical records.