Table 4 Expected return and mortality under different polices
From: Towards more efficient and robust evaluation of sepsis treatment with deep reinforcement learning
Policies | \({\textbf{V}}_{{\textbf{DR}}}\) | Mortality |
|---|---|---|
\(Reward_{3.0}\) | \(-0.0284\) | 14.5% ± 0.6% |
\(Reward_{3.0^+}\) | \(-0.1800\) | 17.2% ± 0.5% |
\(Reward_{4.0}\) | \(-0.0253\) | 14.7% ± 0.6% |
\(Reward_{3.0+3.0^+}\) | 0.0291 | 14.1% ± 0.6% |
\(Reward_{3.0+4.0}\) | 0.0365 | 13.9% ± 0.5% |
\(Reward_{3.0^++4.0}\) | \({\textbf {0.2307}}\) | 11.3% ± 0.4% |
\(Reward_{all}\) | 0.1546 | 12.2% ± 0.4% |
Clinician | \(-0.0294\) | 14.5% ± 0.5% |