The current retrospective study included publicly accessible tweets,
i.e.
Twitter messages,
in English with no geographic limitations.
Using Twitter’s Advanced Search feature,
a total of 288 tweets were collected for a two-year period,
spanning from January 1,
2015,
to December 31,
2016.
The top tweets for each month within the studied period were manually collected and processed.
Two sets of keywords were used.
The first search was performed for tweets containing both the terms “mammography” and “radiation” in any order of appearance,
while a similar second run searched for tweets containing both the terms “mammo” and “radiation”.
Common misspellings for "mammography" were also investigated.
The data derived from all searches were finally merged.
The fore mentioned tweet total was concluded after filtering out duplicates and tweets that although contained the desired keywords their meaning was out of scope.
The distribution of tweets per month searched is presented in Table 1.
Interestingly enough,
October which is the Breast Cancer Awareness Month,
i.e.,
an annual campaign to increase awareness of the disease,
did not present the maximum number of tweets in either year.
Additionally,
19% of the tweets studied included the term "digital".
Two reviewers used the criteria delineated below to independently categorize collected tweets,
whilst a third reviewer resolved discrepancies.
For data analysis,
four categories were predefined,
each one with its own set of subcategories.
The four main categories were “User Type”,
“Linked Source Type”,
“Favourability” and “Informative Value”.
Every single tweet was then reviewed and categorised according its content.
According to the Twitter user’s profile,
each reviewer categorized every tweet in the following “User Type” sub categories: “person” – a user who is not a healthcare professional,
“physician” – medical doctors with specialties not directly related to radiology,
“radiologist”,
“physicist” – physicist specializing in medical physics,
“technologist”,
“organization” – healthcare related scientific or government organizations,
“private company” – e.g.,
private hospitals and practices,
“unspecified” – this subcategory contains tweets by users with incomplete or vague background information in their profiles (it also contains users that generically describe their account as a healthcare related news feed).
The majority of the reviewed tweets included a link to other websites,
blogs,
etc.
that either supported or/and supplemented their content.
Therefore,
tweets were categorized based on their “Linked Source Type” in the following subcategories: “peer reviewed medical journal” – links forwarding directly to a scientific journal or to articles that strictly refer to a medical journal without the author's personal analysis or opinion,
“not peer reviewed” – links to an article by an author that reviews radiation risk from mammography expressing his own opinion of the matter based on scientific references,
“lay press” – links forwarding to unofficial sources of healthcare related news usually found in websites of generic interest,
“blog” – links forwarding to the author’s or third party's personal blog sites and Facebook,
“advertisement” – tweets advertising healthcare services originating mostly from private hospitals,
medical imaging centres or private practices,
“informational website or pamphlet” – links forwarding to professional healthcare service websites which provide information to patients about the mammography procedure and reference its potential radiation risks,
“remark”- tweets which do not include a link but briefly express the user’s personal opinion/remark on the current matter.
The collected tweets were also classified in terms of their “favourability” regarding mammography and its benefits as opposed to its theoretical radiation-induced risk.
In this respect,
each tweet was categorized as “favourable”,
“unfavourable” or “neutral”.
The tweets were also grouped according to how informative their content was considered to be,
specifically about radiation exposure and its projected risks during the mammography procedure.
These tweets were assorted as “informative”,
“uninformative” or “question” – when a user simply asks for information about radiation in mammography.
The processing of the reviewed data produced the results presented in Figures 1-4.