Emerging technology review
Assessment of Trending Ability of Cardiac Output Monitors by Polar Plot Methodology

https://doi.org/10.1053/j.jvca.2011.01.003Get rights and content

Objectives

To develop a valid statistical method of showing acceptable cardiac output (CO) trending ability when new CO monitors are compared to a reference standard, such as thermodilution, using polar coordinates.

Design

Developing a new statistical analytic method using historic data.

Setting

University Hospital Anesthesia and Intensive Care Department.

Participants

Data taken from previously published CO validation studies.

Interventions

Cartesian data were reanalyzed, being uplifted using Data Thief 3.0 software (http://datathief.org/). Polar plots were constructed from this data. Central zone data (<0.5 L/min or <10% change) were excluded because they introduced statistical noise. Trial polar criteria were set using data from a study that compared 5 CO monitors against thermodilution. Then, these criteria were further validated using data extracted from 15 other studies. Mean (95% confidence intervals) polar angles were used.

Measurements and Main Results

Trial data suggest ±5° (angle) ±30° (95% confidence interval) as acceptance limits. Concordance rates (ie, >95%-90%) from 5 articles supported trending, and polar data from these studies concurred with the authors' pilot criteria. Favorable comments on trending also were found in 8 of 15 articles in which radial limits were less than ±32°. Good calibration was associated with a mean polar angle of less than ±5°.

Conclusions

Polar plots can be used to show the trending ability of CO monitors in comparative validation studies. They overcome the deficiencies of concordance analysis, which uses the direction of change as a statistic and ignores the magnitude of change in CO.

Section snippets

Methods

A number of steps were required to transform the X-Y ΔCO data into polar data, and these steps are presented later (Fig 1). Statistical variables that assessed trending ability and described the polar plot needed to be identified. In Bland and Altman analysis, the bias and limits of agreement are used to assess agreement.3 The following variables based on the polar plot and polar angle (θ) were chosen: (1) the mean polar angle (or angular bias) and (2) 95% CIs (limits of radial agreement) of

Results

A total of 18 articles with 4-quadrant plots were identified, 2 of which were excluded because the range of values of the data was too small for polar analysis.21, 22

Discussion

Polar plots present the data from a 4-quadrant plot in a similar format to a Bland and Altman plot but with a radial distribution of data points about a polar origin3 (Fig 2). Full- and half-circle formats can be drawn. The bias becomes the mean of the polar angles formed by these data points and reflects the difference in calibration between the reference and test methods. The radial limits of agreement replace the horizontal limit lines in the Bland and Altman plot. A modified concordance

References (24)

  • A.C. Perrino et al.

    Transtracheal Doppler cardiac output monitoring: Comparison to thermodilution during noncardiac surgery

    Anesth Analg

    (1994)
  • R.B. de Wilde et al.

    An evaluation of cardiac output by five arterial pulse contour techniques during cardiac surgery

    Anaesthesia

    (2007)
  • Cited by (194)

    View all citing articles on Scopus

    Gerard R. Manecke, Jr, MD

    Marco Ranucci, MD

    Section Editors

    View full text