Validation of Delayed Anesthesia Model Using Identification Methods and Correlation Analysis

Many successful control strategies for anesthetic processes do not adequately consider time delays despite potential 30-second delays from biophase and bispectral index (BIS) monitoring. This study introduces linear delayed Pharmacokinetic-Pharmacodynamic (PK-PD) patient models that correlate more to real patient outputs than common non-delayed models. The delayed models are obtained for both effect-site concentration (Ce) and BIS outputs. The model structures are identified using autoregressive exogenous input (ARX) and delayed-ARX (DARX) polynomial models and least squares estimations (LSE) using the real surgical data of the open-access database VitalDB. Validation of delayed PK-PD patient models involves using correlation analysis. Results indicate that PK-PD patient models with output delays exhibit higher correlations with real patient data, particularly when considering BIS output.