How Can Social Media Platforms like Twitter Serve Public Health Departments?

BACKGROUND: Since 2010 there have been over 300 publications on how social media can assist public health surveillance and situational awareness. The increasing number of techniques and platforms utilizing social media can overwhelm health department decision makers. This literature review is aimed to isolate techniques and strategies that public health departments can feasibly use to collect and analyze social media. The 3 social media categories of interest are emergency response (Hurricane Sandy), crowd surveillance (sentiment towards policy) and disease surveillance (ILI).

METHODS: We conducted a PubMed search of peer-reviewed English articles from 2009-2014. Each article was manually sorted into one or more of the following categorizes; emergency response, program evaluation/crowd surveillance, and disease surveillance. Each article was documented for; study type, tools utilized, use of geographic information, feasibility, reproducibility and type of analysis.

RESULTS: Overall, 237 articles were identified, 157 were excluded based on exclusion criteria and 80 were reviewed. Twenty-six unique data analysis tools were mentioned in the literature but analysis methods were unique to study; 94% of the articles were retrospective analyses, and 61% mentioned the use of geotagging. We evaluated the workflow, consisting of data collection, data storage, analysis and data visualization. Methods in the first three steps were summarized. Data collection methods separated into 3 categories: purchasing data from a vendor, sourcing data from a public application programming interface (API) or utilizing an analysis tool which accessed the API in real-time. Data storage methods included local disk storage, local server storage, cloud storage or no storage due to lack ownership. Data analysis methods were the most diverse. Text analysis tools included Boolean searches, manual coding, automated classification, usage of natural language processing, machine learning techniques, and online analysis tools. Non-text modes of analysis included frequency/location of tweets and analysis of users.

CONCLUSIONS: Social media provides a novel and powerful source of data and can add value to a range of public health projects. Selected publications demonstrated the successful use of social media by public health departments resulting in a request to incorporate social media as a novel and complementary data source. However, social media analysis methods are unique, and for in-depth and real-time analysis, capacity must be built internally or sourced through public-private partnerships. This review outlined the mechanisms, feasibility, strengths and limitations of how social media data can successfully be implemented for public health purposes.