Table 4 Utility and challenges of social media listening
From: Utility of social media and crowd-intelligence data for pharmacovigilance: a scoping review
Utility and challenges of social media listening | Count (%) |
|---|---|
Utility of social media listening for pharmacovigilance | |
 Supplemental data to traditional post-marketing safety surveillance | 31 (44.3%) |
 Captures perceptions and consequences of treatment and adverse events | 14 (20.0%) |
 Large publicly available data source | 14 (20.0%) |
 Able to discover undocumented or rare adverse events | 11 (15.7%) |
 Promising early warning system | 10 (14.3%) |
 Computationally efficient | 7 (10.0%) |
 Captures prescription drug misuse/abuse | 4 (5.7%) |
 Not biased towards severe adverse events | 7 (10.0%) |
 Captures large geographical area | 3 (4.3%) |
 Useful for risk communication | 3 (4.3%) |
 Able to extract complex medical concepts | 2 (2.9%) |
 Can be more accurate than spontaneous reporting systems | 2 (2.9%) |
 Hypothesis-generating | 2 (2.9%) |
 Able to identify undocumented drug interactions | 2 (2.9%) |
 Findings are similar to traditional systems | 1 (1.4%) |
 Captures information on adherence related to adverse events | 1 (1.4%) |
Challenges of social media listening for pharmacovigilance | |
 Non-standard reporting format (informal language, format used to report information, amount of information provided by each user) | 30 (42.9%) |
 Difficult to draw complex semantic relationships from unstructured texts | 14 (20.0%) |
 May not be a representative population | 13 (18.6%) |
 Noise may exist in signal detection | 12 (17.1%) |
 Inadequate information to draw causality | 9 (12.9%) |
 Lacks comprehensive medical and demographic information | 8 (11.4%) |
 Subjective, incomplete or misinformation | 6 (8.6%) |
 Not a balanced coverage of all drugs and medical conditions | 5 (7.1%) |
 Data acquisition challenges due to host site restrictions | 4 (5.7%) |
 Duplication of data (double-counting) | 4 (5.7%) |
 Processing multi-lingual texts | 3 (4.3%) |
 Resource-intensive to process big data | 2 (2.9%) |