Department of Population Medicine, Harvard Medical School and Harvard Pilgrim Health Care Institute, 133 Brookline Avenue, Boston, MA 02215, USA; Global Health Unit, Norwegian Knowledge Centre for the Health Services, PO Box 7004, St. Olavs pl, 0130 Oslo, Norway; Department of Community Medicine, Institute of Health and Society, University of Oslo, PO Box 1130 Blindern, 0318 Oslo, Norway.
Department of Population Medicine, Harvard Medical School and Harvard Pilgrim Health Care Institute, 133 Brookline Avenue, Boston, MA 02215, USA.
J Clin Epidemiol. 2015 Mar;68(3):324-33. doi: 10.1016/j.jclinepi.2014.10.003. Epub 2014 Dec 11.
There is often substantial uncertainty about the impacts of health system and policy interventions. Despite that, randomized controlled trials (RCTs) are uncommon in this field, partly because experiments can be difficult to carry out. An alternative method for impact evaluation is the interrupted time-series (ITS) design. Little is known, however, about how results from the two methods compare. Our aim was to explore whether ITS studies yield results that differ from those of randomized trials.
We conducted single-arm ITS analyses (segmented regression) based on data from the intervention arm of cluster randomized trials (C-RCTs), that is, discarding control arm data. Secondarily, we included the control group data in the analyses, by subtracting control group data points from intervention group data points, thereby constructing a time series representing the difference between the intervention and control groups. We compared the results from the single-arm and controlled ITS analyses with results based on conventional aggregated analyses of trial data.
The findings were largely concordant, yielding effect estimates with overlapping 95% confidence intervals (CI) across different analytical methods. However, our analyses revealed the importance of a concurrent control group and of taking baseline and follow-up trends into account in the analysis of C-RCTs.
The ITS design is valuable for evaluation of health systems interventions, both when RCTs are not feasible and in the analysis and interpretation of data from C-RCTs.
卫生系统和政策干预措施的影响往往存在很大的不确定性。尽管如此，该领域很少进行随机对照试验（RCT），部分原因是实验可能难以实施。影响评估的另一种方法是中断时间序列（ITS）设计。然而，对于这两种方法的结果如何比较，人们知之甚少。我们的目的是探讨 ITS 研究是否会产生与随机试验不同的结果。
我们根据集群随机试验（C-RCT）干预组的数据进行了单臂 ITS 分析（分段回归），即丢弃对照组数据。其次，我们通过从干预组数据中减去对照组数据点，将对照组数据纳入分析中，从而构建了代表干预组和对照组之间差异的时间序列。我们将单臂和对照 ITS 分析的结果与基于试验数据常规汇总分析的结果进行了比较。
研究结果基本一致，不同分析方法得出的效应估计值具有重叠的 95%置信区间（CI）。然而，我们的分析揭示了在 C-RCT 分析中，同期对照组和考虑基线和随访趋势的重要性。
ITS 设计对于评估卫生系统干预措施非常有价值，无论是在 RCT 不可行时，还是在分析和解释 C-RCT 数据时。
