Interrupted time-series analysis yielded an effect estimate concordant with the cluster-randomized controlled trial result.

View Abstract

OBJECTIVE

We reanalyzed the data from a cluster-randomized controlled trial (C-RCT) of a quality improvement intervention for prescribing antihypertensive medication. Our objective was to estimate the effectiveness of the intervention using both interrupted time-series (ITS) and RCT methods, and to compare the findings.

STUDY DESIGN AND SETTING

We first conducted an ITS analysis using data only from the intervention arm of the trial because our main objective was to compare the findings from an ITS analysis with the findings from the C-RCT. We used segmented regression methods to estimate changes in level or slope coincident with the intervention, controlling for baseline trend. We analyzed the C-RCT data using generalized estimating equations. Last, we estimated the intervention effect by including data from both study groups and by conducting a controlled ITS analysis of the difference between the slope and level changes in the intervention and control groups.

RESULTS

The estimates of absolute change resulting from the intervention were ITS analysis, 11.5% (95% confidence interval [CI]: 9.5, 13.5); C-RCT, 9.0% (95% CI: 4.9, 13.1); and the controlled ITS analysis, 14.0% (95% CI: 8.6, 19.4).

CONCLUSION

ITS analysis can provide an effect estimate that is concordant with the results of a cluster-randomized trial. A broader range of comparisons from other RCTs would help to determine whether these are generalizable results.

Abbreviation
J Clin Epidemiol
Publication Date
2013-08-01
Volume
66
Issue
8
Page Numbers
883-7
Pubmed ID
23810027
Medium
Print
Full Title
Interrupted time-series analysis yielded an effect estimate concordant with the cluster-randomized controlled trial result.
Authors
Fretheim A, Soumerai SB, Zhang F, Oxman AD, Ross-Degnan D