Social media narratives in non-communicable disease: their dynamics and value for patients, communities and health researchers
Background: Usage of social media is now widespread and growing, as is the number of people living with Non-Communicable Diseases (NCDs) such as diabetes and cancer. This thesis examines how social media are being used to share or discuss NCDs and the benefits, challenges and implications of these trends as a manifestation of digital public health. Aim and research questions: The aim of this research is to address the gap in empirical, evidence-based research into the secondary use of data from social media to understand patient health issues and inform public health research into NCDs. To this end, seven research questions, each linked to a sub-project, were defined and tested during the course of the six-year programme: 1.What is the status of the existing multi-disciplinary research literature based on analysis of data posted on social media for public health research, and where are the gaps in this research? 2.Can existing systematic review methods be re-purposed and applied to analyse data posted on social media? 3.How are research sponsors and researchers addressing the ethical challenges of analysing data posted on social media? 4.To what extent are diabetes-related posts on Twitter relevant to the clinical condition and what topics and intentions are represented in these posts? 5.In what ways do people affected by Type 1 diabetes use different social media (e.g. for social interaction, support-seeking, information-sharing) and what are the implications for researchers wishing to use these data sources in their studies? 6.Are these differences in platform usage and associated data types also seen in people affected by lung cancer? 7.Can characteristic illness trajectories be seen in a cancer patient’s digital narrative and what insights can be gained to inform palliative care services? Methods: A range of different qualitative and quantitative methods and frameworks were used to address each of the research questions listed. Arksey and O’Malley’s five-stage scoping review framework and the PRISMA guidelines are applied to the systematic scoping review of existing literature. The PRISMA guidelines and checklist are re-purposed and applied to the manual extraction and analysis of social media posts. Bjerglund-Andersen and Söderqvist’s typology of social media uses in research and Conway’s taxonomy of ethical considerations are used to classify the ethics guidelines available to researchers. The findings of these were used to inform the research design of the four empirical studies. The methods applied in the conduct of the empirical studies include a content and narrative analysis of cross-sectional and longitudinal data sourced from Twitter, Facebook, the Type 1 diabetes discussion forum on Diabetes.co.uk and the lung cancer discussion forum on Macmillan.org.uk, as well as the application of Bales’ Interaction Process Analysis and Emanuel and Emanuel’s framework for a good death. Results : Of the 49 systematic, quasi-systematic and scoping reviews identified, 24 relate to the secondary use of data from social media, with eight of these focused on infectious disease surveillance and only two on NCDs. Existing reviews tend to be fragmented, narrow in scope and siloed in different academic communities, with limited consideration of the different types of data, analytical methods and ethical issues involved, therefore creating a need for further reviews to synthesise the emerging evidence-base. The rapid increase in the volume of published research is evident, from the results of RQ1, with 87% of the eligible studies published between 2013-2017. Of the 105 eligible empirical studies that focused on NCDs, cancer (54%) and diabetes (20%) dominate the literature. Data is sourced from Twitter (26%), Facebook (14%) and blogs (10%), conducted, published and funded by the medical community. Since 2012, automated methods have increasingly been applied to extract and analyse large volumes of data. Those that use manual methods for extraction did not apply a consistent approach to doing so; the PRISMA guidelines and checklist were therefore re-purposed and applied to analyse data extracted from social media in response to RQ2. The deficit of ethical guidance available to inform research that involves social media data was also identified as a result of RQ3 and the guidelines provided by the ESRC, BPS, AoIR and NIHR were prioritised for the purposes of this research project. Results from the four empirical studies (RQ4-7) reveal that different forms of social interaction and support are represented in the variety of social media platforms available and that this is influenced by the type and nature of the condition with which people are affected, as well as the affordances offered by such platforms. In the pilot study associated with RQ4, Twitter was identified as a ‘noisy’ source of data about diabetes, with only 66% of the sample being relevant to the clinical condition. Twelve per cent of the eligible sample was associated with Type 2 diabetes, compared to 6% for Type 1, and most were information-giving in nature (49%) and correlated with the diagnosis, treatment and management of the condition (44%). A comparison of Twitter to the Type 1 Diabetes community on Facebook and the discussion forum on Diabetes.co.uk for RQ5 indicated that all three social media platforms were used to disseminate information about the condition. However, the Type 1 Diabetes Group on Facebook and the Type 1 discussion forum on Diabetes.co.uk were also used for social interaction and peer support, hence defying the generalisations made in public health studies, where social media platforms were often considered equal or synonymous. The results from the third empirical study into lung cancer (RQ6) support this, indicating that, by virtue of their digital architecture, user base and self-moderating communities, the Lung Cancer Support Group on Facebook and the lung cancer discussion forum on Macmillan.org.uk are more successful in their utility for social interaction and emotional and informational support. Meanwhile, the sample derived from Twitter hashtags showed greater companionship support. The final empirical study in this PhD research project is associated with RQ7 and used longitudinal data posted by a terminally ill patient on Twitter. This revealed that patient activity on social media mirrors the different phases of the end-of-life illness trajectory described in the literature and that it is comparable to or compliments insights garnered using more traditional qualitative research techniques. It also shows the value of such innovative methods for understanding how terminal disease is experienced by and affects individuals, how they cope, how support is sought and obtained and how patients feel about the ability of palliative care services to meet their needs at different stages. Conclusions: The analysis of health data posted on social media continues to be an expanding and evolving field of multi-disciplinary research. The results of the studies included in this thesis reveal the emergence of new methods and ethical considerations to inform research design as well as ethics policy. The re-purposed PRISMA guidelines and checklist were presented at the 2014 Medicine 2.0 Summit and World Congress whilst the review of ethical guidelines was published in the Research Ethics journal. The four empirical studies that extracted and analysed data from social media provide novel insight into the social narratives of those impacted by diabetes and cancer and can be used to inform future research and practice. The results of these studies have, to date, been presented at four international conferences and published in npj Digital Medicine and BMC Palliative Care. Although this thesis and associated publications contribute to an emerging body of knowledge, further research is warranted into the manual versus automated techniques that can be applied and the differences in social interaction and support needed by people affected by different NCDs.