Charting patients' course: a comparison of statistics used to summarize patient course in longitudinal and repeated...

Investigators conducting longitudinal studies of psychiatric illnesses often analyze data based on psychiatric symptom scales that were administered at multiple time points. This study examines the statistical properties of seven indices that summarize patient long-term course. These indices can be used to compare differences between two or more groups or to test for changes in symptoms over time. They may also be treated as outcome measures and correlated with other clinical variables.The performance of each of the seven indices was assessed using data from two large ongoing studies of psychiatric patients: a longitudinal study of affective disorders and a longitudinal study of first-episode psychosis. These two datasets were subjected to bootstrapping techniques in order to calculate both type I error rates and statistical power for each summary statistic. Of the seven indices, Kendall's tau performed the best as a measure of patients' symptom course. Kendall's tau appears to offer more statistical power to detect change in course, yet its average type I error rate was comparable to the other indices.