SpO_{2} coefficient of variation (%) | SpO_{2} <91% (episodes/hour) | SpO_{2} <91% (episode duration, s) | SpO_{2} >95% (episodes/hour) | SpO_{2} >95% (episode duration, s) | |
---|---|---|---|---|---|

Manual control (original recordings) | 4.2 (3.3–4.8) | n.a. | n.a. | n.a. | n.a. |

Core PID | 1.8 (1.7–2.3)^{g} | 16 (14–19)^{a,h} | 7.9 (7.2–8.8)^{a,e} | 17 (14–28)^{a,i} | 11 (9.4–21)^{a,g} |

Core PID+SC | 2.9 (2.3–5.4)^{a} | 24 (19–40)^{c} | 9.1 (8.2–9.3) | 27 (18–30)^{c} | 9.2 (8.6–20)^{c} |

Core PID+adaptive K_{p} | 1.7 (1.5–2.1)^{b} | 10 (7.7–12)^{b,d} | 9.2 (8.7–11)^{b} | 13 (7.1–17)^{b,d} | 16 (13–25)^{b,d} |

Core PID+TRA | 1.7 (1.6–2.2)^{b} | 13 (9.6–14)^{d} | 8.9 (8.2–9.6)^{b} | 15 (9.5–24)^{d} | 13 (11–25)^{d} |

Enhanced PID−SC | 1.7 (1.5–2.1) | 11 (8.0–13)^{e} | 9.7 (9.0–11)^{d} | 13 (7.3–18)^{g} | 15 (13–26)^{e} |

Enhanced PID−adaptive K_{p} | 2.7 (1.8–3.8)^{c} | 18 (15–25)^{f} | 9.4 (8.9–10) | 17 (13–25)^{e} | 9.7 (8.7–23)^{f} |

Enhanced PID−TRA | 1.8 (1.6–2.4)^{d} | 13 (11–16)^{f} | 9.2 (8.7–10)^{c} | 11 (5.3–17)^{f} | 9.3 (8.3–24)^{f} |

Enhanced PID | 1.7 (1.6–2.3)^{g} | 13 (11–15)^{f,g} | 9.7 (9.0–11)^{d,g} | 12 (5.2–15)^{f,h,j} | 9.7 (8.6–26)^{f,g} |

Enhanced PID 30 s lockout | 2.2 (2.0–2.8)^{h} | 15 (12–17) | 13 (12–15)^{f,h} | 13 (5.8–16)^{j} | 14 (12–29)^{h} |

PD control | 4.5 (3.0–7.4)^{h} | 27 (14–56)^{h} | 21 (17–27)^{f,h} | 15 (7.1–36) | 28 (11–42)^{h} |

Indices of SpO

_{2}instability with different algorithms. Median (IQR). Within-column statistical comparisons (Friedman analysis of variance with Dunn's post hoc test):^{a}differs from^{b}, p<0.05;^{c}differs from^{d};^{e}differs from^{f};^{g}differs from^{h};^{I}differs from^{j}.K

_{p}, proportional coefficient; PD, proportional–derivative; PID, proportional–integral–derivative; SC, Severinghaus compensation; TRA, target range attenuation.