| Sumario: | Background: Especially in an intensive care setting, the cumulative fluid balance (CFB) provides easy-to-assess and valuable information on the patient's current health status [1] and the amount of excess fluid currently accumulated within the body. A fluid overload of 10% of a patient's baseline body weight is associated with an increased mortality [2-4] and should therefore be avoided. Estimating a patient's CFB course as a response to different fluid application regimes may be difficult. Modeling an individual patient's fluid transfer characteristics by considering as many relevant patient parameters as possible can be challenging and easily results in high dimensional and complex models, whose introduction into clinical practice can be difficult. Control system analysis provides efficient tools for the description of complex systems and is commonly used in other areas aiming to model physiological behavior [5-9]. Objectives: The identification of individual transfer functions commonly used in system analysis may help in detecting patients being non-responsive to late conservative fluid therapy at an early stage of postoperative fluid management. Clustering the individual patients' transfer functions within a large patient population with respect to diagnosis or other patient features might furthermore allow the definition of cohort-specific model parameters. The use of cohort transfer functions in decision support systems might assist in assessing each patient's actual fluid needs, facilitating fluid management by preventing severe fluid overloads and minimizing the risk of therapies such as renal replacement therapy in advance. Methods: The CFB course of critically ill patients recovering from trauma has already been described qualitatively in literature [10-13]. Malbrain et al. [10] suggested the ROSE model, which divides the recovery process into four subsequent stages: Resuscitation, Optimization, Stabilization and Evacuation. Figure 1 shows the CFB course to be targeted in intensive care. In general, the lengths of the four subsequent recovery phases depend on the clinical course of the respective patient. A second order discrete-time transfer function was identified using a selected cardiac patient's individual cumulative fluid intake (CFI) and CFB as input series and output series respectively. The patient's transfer function was estimated using the MATLAB System Identification Toolbox. Model verification was performed using MATLAB Simulink, whereby an approximated intake function fitted to the patient's CFI was used as input series. Results: The identified transfer function comprises a holistic description of the patient's characteristics influencing the individual reaction to administered fluids without necessity for measuring multiple and/or complex vital parameters. The model output of the estimated transfer function for the selected patient after application of the approximated CFI compared to the patient's actual CFB versus the averaged CFBs including four patients with similar lengths of stay are shown in Figure 2. Conclusions: Second order transfer function models provide a valuable option for describing fluid transfer characteristics of ICU patients. The estimated transfer function shows a good congruence with the documented preliminary patient data. A transfer function of higher order does not result in a justifying increase of goodness of fit. Patient- -specific transfer functions might act as a key tool reflecting the actual patient within control loops being an essential base for providing decision support in fluid administration. Acknowledgments: "CBmed" is cooperating together with B.Braun Melsungen AG for this research project. Work done in "CBmed" was funded by the Austrian Federal Government within the COMET K1 Centre Program, Land Steiermark and Land Wien.
|
|---|