Data analysis
Manual encoding of Twitter content is more accurate than automatic encoding because humans can perceive linguistically refined text more efficiently than computer-based systems.15 Therefore, data were recorded and evaluated manually in this study. The data were recorded using the Microsoft Excel program and qualitatively evaluated by thematic analysis16 and the themes were determined based on this analysis.
The tweets were recorded under the following codes: (1) date, (2) number of retweets, (3) number of likes, (4) tweet source, (5) tweet, (6) theme, (7) positive/negative/neutral, (8) tweet uploader status. The tweets were posted by 3 different groups of people: (1) community, (2) layperson, (3) news. 5 themes were identified: (1) criticism of bullying, (2) news about bullying in CLP, (3) parental experience of a child being bullied, (4) personal experience of being bullied, (5) social support against bullying. Tweet uploader status was classified into 3 groups: (1) CLP subjects, (2) CLP subjects’ parents, (3) irrelevant individuals (individuals who were unaffected). All tweets were evaluated by 2 experienced orthodontists who worked independently. When a case of conflict existed, two investigators exchanged ideas and determine the final theme together after the first evaluation.