Determinants of reach counts on medical social media: Evidence from a facebook community page metrics

Abstract Sharing medical information using social media became popular. We aimed to identify what sort of posts can increase reach counts. We used metrics data of 40 posts on the Japan Primary Care Association Commission on Social Determinants of Health (JPCA‐CSDH) Facebook community page. The most popular post was the one about the position paper on health inequality (Mie‐statement). The video posts significantly earned 175% more reach counts (vs posts only with letters). Each one percent word counts increase was associated with a 31% increase in reach counts. Videos and more words may deliver medical information effectively.


| INTRODUC TI ON
Sharing medical information via social media has been attracting attention recently in medical education and public health. 1,2 Medical societies and medical care providers have shared information on social networking services (SNSs) (Facebook, Twitter, etc…), weblog communities, and blogs. 3 Several studies have identified that visual articles including photographs or video can gain people's reach counts effectively in the marketing field. 4,5 However, the evidence of determinants of effective information delivery on social media is poorly documented and still inconclusive in the medical field, 6 especially in Japan. Moreover, there are many restrictions on the collection of metrics data on social media, making it difficult to use the data for research. 7 To close the gap, we analyzed the data of the Japan Primary Care Association Commission on Social Determinants of Health (JPCA-CSDH) Facebook community page, 8 whose metrics data were available with administrator authority of the community page.
From the data, we aimed to identify what sort of posts can deliver the information more effectively.

| Data
The JPCA-CSDH runs its Facebook page to share information on the activities and has 913 followers and two administrators who can post on the community, as of Dec. 15, 2020. For the present analysis, we analyzed data of Jun 2018 to December 2019. Facebook metrics data were obtained through the insight interface of Facebook community page functions. Each metrics data contained the date of posts; counts of reach, engagement, and reaction; contents of posts; and publishing statuses. We retrieved the data in January 2020.
Because of a fixed available sample, the number of posts during the study period determined our sample size. 9 | 301 NISHIOKA ANd TSUBOYA

| Outcome variables
We identified reach counts of each post from metrics data. Reach count was defined as "the number of unique people (not number of times) who saw a content" in Facebook, which we could use this variable as a number of people that information was delivered.
Before we conduct the analyses, we log-transformed reach counts to achieve normality.

| Covariatess
Based on data availability, we used conference dummy variable that explains the posts about JPCA conferences and seminars (Yes/No), dummy variable of person who posted the content on the JPCA-CSDH community page, to adjust for unmeasured characteristics of posts (A/B), and date of submission that was categorized into three terms; June to December 2018, January to June 2019, and July to December 2019, to adjust the effect of the number of followers on engagement.

| Statistical analysis
First, we described the metrics and characteristics of overall posts.
Second, we performed univariable linear regression analysis and calculated the crude coefficient of reach counts and its 95% confidence interval (CI) of each explanatory variable. Third, we performed multiple linear regression analysis to calculate the multivariable-adjusted coefficient of each explanatory variable. There were no missing data for the variables. Analyses were performed by using STATA SE Ver.16.2 for these statistical analyses (Stata Corp.).

| Ethical consideration
Our study did not include any personal information related to the posts and was thus exempted from ethical review. Also, Ross reported that posts with photographs and videos were associated with more interactions globally. 5 Our results implicated that social media enable more people or health professionals to access medical information. For example, members of a medical society who were unable to attend the annual congress or nonmembers of a medical society would become to be able to access medical information, which would result in improving the information gap, and making society better. Thus, medical societies could be accountable and increase their presence in the society. In the COVID-19 era, remote congress and webinars become much more popular; thus, it is important and required to provide more information through video social media, to analyze the relevant factors in more detail, and to work on improving the quality of information dissemination.

| RE SULTS
There were several limitations to our study.

ACK N OWLED G EM ENTS
This research was presented at the 11th JPCA annual meeting. We would like to thank Members of JPCA-CSDH for their helpful comments and support for the analysis and data interpretation. Abbreviations: CI, confidence interval; coef., Coefficient.

CO N FLI C T O F I NTE R E S T
Multivariable linear regression analysis was adjusted by types of posts, log word counts, conference dummy variable, term of submissions, and posted person.