01 January 1970 0 6K Report

z-score is widely used in pediatric echocardiography, and sometimes in adult echocardiography. The benefit of z-score is the growth and change of body size can be corrected so that comparison of the dimensions or other growth related parameters of the heart can be achieved. However, how much change of z-score during the follow up is significant is not clearly defined.

Suppose there are visit1 and visit2. Z-score1=(measurement1-mean1)/standard deviation while z-score2=(measurment2-mean2)/standard deviation. If there is no other physiological or pathological reason for a change of cardiac parameter, the z-scores for one individual in a population should be constant during the follow up. So we have z-score=(measurment1-mean1)/standard deviation=(measurement2-mean2)/standard deviation. The only variation of z-score is then from observation, i.e. echocardiographic acquisition and measurement. If coefficient of variance for an individual echocardiographic parameter is known for an echo lab, let's say x%, both measurement1 and measurement 2 could have a range 1±2x%, and their difference could be to an extreme when measurement1 is +2x% and measurment2 -2x%. When we transfer them into z-score, the difference between z-cores would be (2x% of measurement1)/standard deviation versus (-2x% of measurement2)/standard deviation. If the standard deviation is known and constant, then the acceptable range of z-score can also be calculated.

Example: for LVED (LV enddiastolic dimension), coefficient of variance is 5%. The standard deviation, or mean standard error for the regression for z-score is 0.7cm. Measurment1 is 3.9 cm, measurment2 is 4.4cm. The tolerable difference of z-scores (95% confidence interval) is 4.4X2x5%/0.7-3.9X(-2)X5%/0.7=1.19. In other words, depending on the BSA at measurement 1 and 2, if the actual difference of z-scores between the visits are more than 1.19, significant change of z-score can be assumed.

This method at least can give reference for a significant z-score change. But one has to be well aware of the coefficient of variance of an echo parameter in the lab, also has to be knowledgeable of the regression formula used for z-score calculation. In case more complicated regression instead of simple linear regression and constant standard error are used, this method can be complicated to use as well.

More Wei Hui's questions See All
Similar questions and discussions