Interobserver variability in the interpretation of pulmonary artery catheter pressure tracings.

STUDY
OBJECTIVE: We evaluated the ability of three independent reviewers (R1, R2, R3) using waveform analysis to accurately identify confirmed valid PCWP tracings, and their ability to consistently report the PCWP numerical value.
DESIGN: Sixty PA and PCWP tracings were prospectively obtained and blindly reviewed by three independent critical care physicians.
SETTING: The medical ICU of Wilford Hall USAF Medical Center. PATIENTS OR PARTICIPANTS: Twenty mechanically ventilated patients with PA catheters inserted for hemodynamic assessment.
INTERVENTIONS: Sixty PA and PCWP tracings were reviewed blindly and independently for acceptability using waveform criteria by three critical care physicians. While recording all 60 tracings, blood was aspirated from the distal port of the PA catheter with the balloon "wedged" and blood gas analysis was done. Each reviewer analyzed the PCWP tracings for validity using waveform criteria, and reported a numerical PCWP reading for those tracings judged valid by waveform criteria. Reviewer sensitivity, specificity and accuracy in performing waveform analysis were assessed by comparing their predictions with those tracings that were confirmed their predictions with those tracings that were confirmed valid by the aspiration of pulmonary capillary blood. Inter-reviewer agreement upon which validity of PCWP tracings was based and reviewer agreement on the numerical PCWP reading were also assessed. All tracings were blindly reviewed by each physician, first without and then with an AP tracing to define end-expiration. MEASUREMENT AND
RESULTS: Thirty-eight of 60 PCWP tracings were confirmed valid by the aspiration of pulmonary capillary blood. In the remaining 22 tracings, mixed venous blood was aspirated with the balloon wedged, and tracing validity was unconfirmed. Reviewer accuracy in identifying was 50 percent for R1, 65 percent for R2 and 57 percent for R3. No reviewer's accuracy was significantly different from a random guess which would yield an accuracy of 50 percent. Agreement by all three reviewers in identifying valid PCWP tracings using waveform analysis varied from 37 percent in the absence of an AP tracing to 66 percent when an AP tracing was available to identify end-expiration (p less than 0.003). Agreement by all three reviewers on the PCWP numerical reading (within 4 mm Hg) was 79 percent without an AP tracing and 96 percent with an AP tracing (p = NS). The numerical reading reported by the ICU nurses and house staff correlated closely with the reviewers' readings. Agreement with the reported PCWP reading was improved only for R2 by the addition of an AP tracing.
CONCLUSION: We conclude that the validation of PCWP tracings by waveform analysis is subject to interobserver variability, and reviewer accuracy in identifying confirmed valid tracings was no better than a random guess. Agreement on the numerical PCWP reading was high among the reviewers as was agreement by each individual reviewer with the reported PCWP. Finally, the presence of an AP tracing, to define end-expiration, adds little to the interpretation of the PCWP numerical reading by experienced physicians.