Table 6.
Subgroup results of the BP control RL algorithm
| Subgroup | No. of encounters | RL benefit relative to clinician policy | ||
|---|---|---|---|---|
| SBP under RL | SBP under clinician | Benefit | ||
| Male | 8108 | 131.32 (0.09) | 132.45 (0.17) | − 1.13 (0.17) |
| Female | 12,143 | 132.07 (0.08) | 132.29 (0.14) | − 0.22 (0.14) |
| Age >60 years | 16,151 | 131.43 (0.07) | 132.34 (0.12) | − 0.90 (0.12) |
| Age ≤60 years | 4100 | 133.12 (0.13) | 132.43 (0.25) | 0.68 (0.24) |
| White ethnicity | 11,925 | 130.35 (0.07) | 131.22 (0.14) | − 0.87 (0.14) |
| Black ethnicity | 6536 | 134.19 (0.11) | 135.12 (0.20) | − 0.93 (0.19) |
| Other ethnicity | 1790 | 132.41 (0.23) | 129.79 (0.39) | 2.62 (0.38) |
| Smoker | 951 | 132.34 (0.31) | 132.55 (0.54) | − 0.21 (0.53) |
| Non-smoker | 19,300 | 131.74 (0.06) | 132.35 (0.11) | − 0.60 (0.11) |
BP blood pressure, RL reinforcement learning, SBP systolic blood pressure