Issues in the statistical analysis of clinical time-series data

Abstract
Interrupted time-series experiments are used with growing frequency to monitor and evaluate the delivery of social services. Practitioners who wish to analyze statistically the results of such experiments face problems related to the lack of familiarity with the nature of human behavior measured over time, short or nonexistent baseline data, and difficulties in summarizing results from several studies. Findings from the reanalysis of 70 interrupted time-series studies from practice settings provide the basis for addressing these issues.