Model 1 | Model 2 | Model 3 | Model 4 | |
---|---|---|---|---|

Intercept variables | Urea | Urea | Urea | Urea |

PCO_{2} | PCO_{2} | PCO_{2} | PCO_{2} | |

Group | Group | Group | ||

Trajectory variables | Time^{0.5} | Time^{0.5} | Time^{0.5} | Time^{0.5} |

Time^{2} | Time^{2} | Time^{2} | Time^{2} | |

Trajectory interactions | Group × time^{0.5} | Net fluid balance × time^{0.5} | Group × time^{0.5} | |

Group × time^{2} | Net fluid balance × time^{2} | Group × time^{2} | ||

Group × net fluid balance × time^{0.5} | Group × net fluid balance × time^{0.5} | |||

Group × net fluid balance × time^{2} | Group × net fluid balance × time^{2} | |||

AIC | 2037.7 | 2057.2 | 2045.2 | 2048.2 |

A series of exploratory models is shown, examining the effect of 4 h net fluid balance on glucose-corrected sodium. The effects include: (1) intercept, how the independent variables influence glucose-corrected sodium at time zero (presentation to hospital); (2) trajectory, the relationship between glucose-corrected sodium and time; and (3) interaction effects, interacting variables that may alter the trajectories in (2). Model fit is quantified by the Akaike Information Criterion (AIC), with smaller values indicating a better fit. All models are adjusted for baseline urea and PCO

_{2}. Model 1 is the ‘reference model’ from table 3, quantifying the effect of the oedema-defined groups on corrected sodium. Model 2 ignores the oedema group classification and examines the role of 4 h net fluid balance on corrected sodium trajectory. Models 3 and 4 examine whether fluid balance affects the three oedema-based groups in different ways (a higher order interaction). The lowest AIC (best model fit) is for model 1, demonstrating that inclusion of fluid balance does not improve model fit.PCO

_{2}, partial pressure of dissolved carbon dioxide at hospital presentation. Group refers to: controls, early and late onset cerebral oedema.