Table 2.
Most promising tweet features to help distinguish the accuracy level of tweets.
| Metric | Description | Trustworthy tweets (N=75), n (%) | Untrustworthy tweets (N=34), n (%) | P valuea |
| URLs | The tweet contains a URL | 6 (8) | 8 (24) | .03 |
| Listed | The author is listed | 53 (71) | 17 (50) | .051 |
| Favorited | The tweet is favorited | 46 (61) | 14 (41) | .06 |
| Hashtags | The tweet contains a hashtag | 20 (27) | 4 (12) | .13 |
| Tashkeel | The tweet contains a tashkeel | 22 (29) | 5 (15) | .15 |
| Exclamation mark | The tweet contains “!” | 3 (4) | 4 (12) | .20 |
| Semicolon | The tweet contains “;” | 42 (56) | 14 (41) | .21 |
| Retweeted | The tweet was retweeted | 43 (57) | 15 (44) | .22 |
| Kashida | The tweet contains a kashida | 6 (31) | 8 (24) | .49 |
aP values were calculated using the Fisher exact test with P≤.05 indicating statistical significance.