A comparison in cardiac output data: a random effects model for repeated measures

Introduction

A random effects model can be used to estimate the within-subject variation after accounting for other observed and unobserved variations, in which each subject has a different intercept and slope over the observation period. On the basis of the within-subject variance estimated by the random effects model, Bland-Altman plots can be created.

Methods

In 28 cardiac surgery patients, cardiac output data LiDCO™plus, PICCO, FloTrac/Vigileo pulse contour and CCO (PAC-Vigilance) was collected at 1 hour (T1), 2 hours (T2), 4 hours (T3), 8 hours (T4), 12 hours (T5), 24 hours (T6), 36 hours (T7), and 48 hours (T8) after ICU admission and compared against intermitted thermodilution COtd (ICO). Within patient variation was calculated using Linear Mixed Models (SPSS). Percentage error is calculated as: PE = [(2.SD of CO difference)/(CO_mean)] × 100%.

Results

The results of the random effects model on continuous cardiac output data are presented in Figure 1.

Figure 1

Bland-Altman statistics from CO data: random effects model (LMM).

Conclusions

The variation of the differences of the original measurement will be underestimated by this practice because the measurement error is, to some extent, removed. The bias between these two methods will not be affected by averaging the repeated measurements.