Using physiological models and decision theory for selecting appropriate ventilator settings.

OBJECTIVE: To present a decision support system for optimising mechanical ventilation in patients residing in the intensive care unit. METHODS: Mathematical models of oxygen transport, carbon dioxide transport and lung mechanics are combined with penalty functions describing clinical preference toward the goals and side-effects of mechanical ventilation in a decision theoretic approach. Penalties are quantified for risk of lung barotrauma, acidosis or alkalosis, oxygen toxicity or absorption atelectasis, and hypoxaemia. RESULTS: The system is presented with an example of its use in a post-surgical patient. The mathematical models describe the patient's data, and the system suggests an optimal ventilator strategy in line with clinical practice. CONCLUSIONS: The system illustrates how mathematical models combined with decision theory can aid in the difficult compromises necessary when deciding on ventilator settings.

Hypoxaemia after cardiac surgery: clinical application of a model of pulmonary gas exchange.

BACKGROUND AND OBJECTIVE: To investigate the clinical application of a mathematical model of pulmonary gas exchange, which ascribes hypoxaemia to shunt and ventilation/perfusion mismatch. Ventilation/perfusion mismatch is quantified by deltaPO2, which is the drop in oxygen pressure from alveoli to lung capillaries. Shunt and deltaPO2 were used to describe changes in oxygenation after coronary artery bypass grafting. METHODS: Fourteen patients were studied 2-4 h after surgery and on postoperative days 2, 3 and 7. On each occasion inspired oxygen fraction was changed in four to six steps to obtain arterial oxygen saturation (SaO2) in the range of 90-100%, enabling construction of FEO2/SaO2 curves. Measurements of ventilation, circulation and oxygenation were entered in a previously described mathematical model of pulmonary gas exchange. RESULTS: We found that oxygenation was most impaired 3 days after surgery. By fitting the mathematical model to the FEO2/SaO2 curve, we found that shunt remained constant throughout the study period. However, deltaPO2 increased from 0.5 kPa (median, range 0-3.8) 2-4 h after surgery, to 3.2 kPa (range 1.2-6.4, P < 0.05) on day 2, and to 4.0 kPa (range 1.2-8.3) on day 3. On day 7, deltaPO2 decreased to 2.2 kPa (range 0-3.5, P < 0.05). CONCLUSIONS: Ventilation/perfusion mismatch (deltaPO2), rather than shunt, explains the changes in postoperative oxygenation. The model of pulmonary gas exchange may serve as a useful and potentially non-invasive clinical tool for monitoring patients at risk of postoperative hypoxaemia.

Modelling of hypoxaemia after gynaecological laparotomy.

BACKGROUND: Late postoperative arterial hypoxaemia is common after major surgery, and may contribute to cardiovascular, cerebral or wound complications. This study investigates the time course of hypoxaemia following gynaecological laparotomy, and estimates parameters of mathematical models of pulmonary gas exchange to describe hypoxaemia. METHODS: Twelve patients were studied on four occasions; preoperatively, 2, 8 and 48 h after surgery. On each occasion inspired oxygen fraction (FIO2) was varied, changing end-expired oxygen fraction (FEO2) to achieve arterial oxygen saturations (SaO2) ranging from 90% to 100%. Measurements of ventilation and blood gases were taken. Oxygenation was characterized plotting FEO2 against SaO2. The shape and position of the FEO2/SaO2 curve was described using two mathematical models including parameters describing gas exchange: either shunt and resistance to oxygen diffusion (Rdiff); or shunt and asymmetry of ventilation-perfusion (fA2). RESULTS: Two hours after surgery SaO2 was reduced from 97.5%+/-1.2% (mean+/-SD) to 93.8%+/-2.7% (mean+/-SD) (P<0.001). Values of shunt, Rdiff and fA2 were significantly changed at 2 and 8 h postoperatively. Forty-eight hours postoperatively Rdiff and fA2 were still significantly changed. CONCLUSION: Oxygenation in 12 patients preoperatively, 2, 8 and 48 h after gynaecological laparotomy is described. Two patients were hypoxaemic (SaO2 <92%) 48 h postoperatively. When two different models of oxygen transport are fitted to patient data, high values of Rdiff or low values of fA2 describe the right shift in the FEO2/SaO2 curve seen in patients with oxygenation problems. These models fit patient data identically, and may be useful in quantifying postoperative hypoxaemia.

Hypoxemia after coronary bypass surgery modeled by resistance to oxygen diffusion.

OBJECTIVE: To evaluate a model describing postoperative hypoxemia after cardiac surgery by using two variables, i.e., shunt and resistance to oxygen diffusion (Rdff). DESIGN: Estimation of these two variables in normal subjects and postoperative cardiac patients. SETTING: The pulmonary function laboratory for the normal subjects and the intensive care unit for the cardiac patients. PATIENTS/SUBJECTS: Nine postoperative cardiac patients and six healthy subjects. INTERVENTIONS: Inspired oxygen fraction was varied in normal subjects and in cardiac patients 3-6 hrs after surgery. This variation occurred in four to seven steps to achieve arterial oxygen saturations in the range 0.90-1.00. MEASUREMENTS AND MAIN RESULTS: Measurements were taken of arterial oxygen saturation, cardiac output, ventilation, and end-tidal gases at each inspired oxygen fraction. These measurements gave the following estimates for the normal subjects: shunt = 3.9+/-5.4% (mean +/- SD) and Rdiff = -5+/-16 torr/(L/min) [-0.7+/-2.2 kPa/(L/min)]; for the cardiac patients: shunt = 7.7+/-1.8% and Rdiff = 212+/-230 torr/(L/min) [28.2+/-30.6 kPa/(L/min)]. The increase in Rdiff (P = .01) was sufficient to explain the observed hypoxemia in these patients. The value for shunt was not significantly increased in the patients (p = .09). The two-variable model (shunt and Rdff) gave a better prediction of arterial oxygen saturation than a model with shunt as the only variable (p = .02). CONCLUSIONS: In cardiac patients requiring supplementary oxygen, the respiratory abnormality could, in our model, be best described by an increased Rdiff, not by an increased shunt value.