Here contains dataset(csv), analysis program (Stata do-file;in zip), preprint(pdf). Preprint - Abstract Objective: The number of suicides in Japan is about 30 000 people a year. In Japan, Suicide Prevention Weeks (September 10 to September 16) and Suicide Support Measure Reinforcement Months (March) have been established, and efforts to appropriately educate the public are implemented. The impact of posting on social media is growing in today's world. However, efforts taking into consideration the spread of suicide-related information via social media during these periods have been inadequate. In this study, we assess whether the number of postings (“tweets”) on Twitter changes during these periods. Subjects and Methods: I targeted tweets posted between January 1, 2011 and December 31, 2014. I defined the number of suicide-prevention-related tweets as “number of preventive tweets” and the number of other suicide-related tweets as “number of related tweets”. The number of tweets posted each day was calculated and analyzed using an ARMA model as time series data. Because there was no stationarity in the number of related tweets, we analyzed differences in this number compared with the previous day. Results: The number of preventive tweets increased by 15.62 tweets (95% CI 4.16, 27.09 p=0.008) during suicide support periods. However, the interday difference in the number of related tweets did not change significantly during suicide prevention periods (65.98 tweets 95% CI -330.00, 461.96 p=0.744). Discussion: The ARMA model showed that preventive tweets significantly increased during suicide support periods. It will be necessary in future studies to more closely examine the influence of social media on suicide incidents by linking with external data. Data set The 1st column (label: "suicide_all") = All sampled Tweets including "Suicide" or "Self Death" The 2nd column (label: "suicide_prevent") = All sampled Tweets including 'Suicide Precautionary Term' The 3rd column (label: "date") = Date Stata do-file Save all do-files in the same folder (directory) as the data set and execute "master.do" to get all the analysis results. These do-files were executed in Stata 14.

This study was supported by JSPS KAKENHI Grant Number. 26870387.