The Statistically Meaningful Display of Analog Data

Background: Analog recordings of data are used in many fields and it is often important to determine if portions of a recording deviate from normal. In many cases, the accurate interpretation of analog recordings is time-consuming and requires considerable skill and experience. The present paper describes a statistics-based method for quickly detecting possible abnormalities in analog tracings and for assessing the likelihood that any apparent abnormality is genuine, rather than simply a variant of normal. Methods: In the proposed method, each difference in the recording being analyzed vs. an appropriate normal reference is expressed as a Z score. For all the Z score that are statistically significant, the corresponding portions of the analog tracing are color-coded to represent the directions and relative magnitudes of the differences. The usefulness of the Z score method was tested using the electrocardiogram (ECG). The diagnostic performance of the Z score method for diagnosing prior inferior and anterior myocardial infarction was compared that of each of two widely used commercial ECG diagnostic algorithms. Differences in diagnostic sensitivity at 95% specificity were tested using chi square analysis. The presence or absence of myocardial infarction (MI) in each subject was determined by cardiac catheterization and the interpreter of the ECGs was blinded to the results of the catheterizations. Results: By cardiac catheterization, there were 497 normal subjects, 366 with inferior MI and 275 with anterior MI. For inferior MI, the sensitivities of the Z score method and the commercial diagnostic algorithms were 79%, 56% and 64%, respectively (P<0.0000001 and P<0.000001). For anterior MI, the sensitivities at 95% sensitivity of the Z score method and the commercial diagnostic algorithms were 75%, 55% and 63%, respectively (P <0.000001 and P <0.002). Conclusions: Color-coding based on Z scores to designate differences from normal reference data is an objective and accurate method to detect abnormalities in analog recordings.