Kappa values evaluating the inter-rater reliability (agreement) between human coders and supervised training of three-topic training models
| Topics | Train percentage | |||
| 0.2 | 0.4 | 0.6 | 0.8 | |
| 124 | 0.67 (±0.01) | 0.70 (±0.02) | 0.68 (±0.03) | 0.70 (±0.04) |
| 135 | 0.66 (±0.02) | 0.66 (±0.02) | 0.67 (±0.05) | 0.67 (±0.06) |
| 145 | 0.67 (±0.02) | 0.71 (±0.02) | 0.70 (±0.02) | 0.72 (±0.02) |
| 245 | 0.67 (±0.04) | 0.69 (±0.03) | 0.71 (±0.03) | 0.72 (±0.06) |
| 257 | 0.68 (±0.02) | 0.70 (±0.01) | 0.71 (±0.02) | 0.70 (±0.02) |
Topic labels are: (1) distrust, (2) media messaging, (3) trusted sources of information, (4) personal medical concerns, (5) family concerns, (6) societal concerns, (7) barriers to recommendations, (8) no worries, (9) other Broad.