Table 3 Associations between risk identification model and recurrent stoke events
From: A deep learning system for detecting silent brain infarction and predicting stroke risk
Stroke events Incidence per 1,000 person-years, number of cases/number of people | aRR (95% CI) | |||
|---|---|---|---|---|
Integrated community management programme | IM group (n = 56) | DeepRETStroke low risk (AI-Low) | Metadata-low risk (Meta-Low) | −38.86 (−91.5 to 297.93) |
73.95 (2/28) | 82.05 (2/26) | |||
DeepRETStroke high risk (AI-High) | Metadata-high risk (Meta-High) | 48.09 (−62.42 to 598.57) | ||
324.12 (8/28) | 292.47 (8/30) | |||
Non-IM group (n = 162) | DeepRETStroke low risk (AI-Low) | Metadata-low risk (Meta-Low) | −97.14 (−99.49 to −90.81) | |
27.14 (3/112) | 136.94 (6/48) | |||
DeepRETStroke high risk (AI-High) | Metadata-high risk (Meta-High) | 543.61 (53.68 to 2,572.37) | ||
202.17 (9/50) | 53.93 (6/114) | |||
Comprehensive interventions: [(AI-High + AI-Low) − (Meta-High + Meta-Low)] in IM group − [(AI-High + AI-Low) − (Meta-High + Meta-Low)] in Non-IM group | 82.44 (1.58 to 324.47) | |||
