2004 OPEN FORUM Abstracts
COMPARABILITY OF THE ACTIVE SERVO LUNG 5000 SIMULATOR WITH A PNEUMOTACHOMETER/ PRESSURE TRANSDUCER SYSTEM
Alotaibi, MS, RRT (University of Medicine and Dentistry of New
Jersey, Newark, NJ); Craig Scanlan, EdD, RRT, FAARC (UMDNJ,
Newark, NJ); Robert Kacmarek, PhD, RRT, FAARC (Massachusetts
General Hospital and Harvard Medical School, Boston, MA); Al Heuer,
PhD, RRT (UMDNJ, Newark, NJ).
BACKGROUND: Most ventilator bench testing is conducted on bellows/spring lung models, with key variables measured by pneumotachometer and pressure transducer (PnPT), interfaced with data acquisition hardware and software. The Active Servo Lung 5000 (ASL; Ingmar Medical, Pittsburgh, PA) integrates all these components into a single computer-controlled device that could serve as a new standard for ventilator testing. Unfortunately, there is no performance data yet published on this device. To address this limitation, we assessed whether the ASL's primary measures (pressure and volume) were comparable to those obtained via a calibrated PnPT system.
METHOD: A method agreement study was conducted using the approach recommended by Chatburn (1996), with data analysis based on that of Bland and Altman (1999). Acceptable levels of ASL repeatability (repeatability coefficient or RC), average bias between methods, and method limits of agreement (LOA) were set a priori for pressure (RC = 0.5 cm H2O; bias = ±2 cm H2O; LOA = no differences > ± 2SD of bias) and volume (RC = 5 mL; bias = ±20 mL; LOA = no differences > ± 2SD of bias). The PnPT system was calibrated using a standard 2-point procedure for pressure, flow and volume, with all data signals digitized and recorded using WINDAQ software. The calibrated PnPT system was then placed in line with the ASL and 15 observations of two breaths each were obtained over a range of pressure (10-55 cm H2O) and volume (200-900 mL), with simultaneous recording of PnPT and ASL 5000 data. RESULTS: The repeatability coefficients for the ASL's measurements were 0.10 for pressure and 1.50 for volume, both superior to the PnPT system. Pressure and volume bias and LOAs between these two systems, including their 95% confidence intervals (95% CI) are reported in the following table.
|Measurement||Bias (95% CI)||Limits of Agreement|
|Upper (95% CI)||Lower (95% CI)|
|Pressure cm H2O||–0.83 (± 0.49)||0.74 (± 0.75)||-2.41 (± 0.75)|
|Volume mL||13.14 (±1.62)||-3.83 (± 8.16)||30.11 (± 8.16)|
Bland-Altman plots revealed that all volume observations were within their expected LOA, but that one pressure observation (mean P = 54.62 cm H2O; P= -2.53 cm H2O) fell outside its expected LOA. However, these plots also revealed systematic trends in measurement bias between the two systems. Pressure bias trended positive (ASL > PnPT) at low pressures and negative (ASL < PnPT) at high pressures. Volume bias was always positive (ASL > PnPT), with a trend toward greater volume differences at progressively higher volumes.
CONCLUSIONS: As compared to the traditional PnPT ventilator bench test standard, the ASL exhibits good repeatability and acceptable bias and levels of agreement within the range of pressure and volume commonly encountered with adult patients. Systematic trends in measurement differences indicate the potential need for correction formulae when interpreting data obtained between these systems.