Statistical process control and interrupted time series: a golden opportunity for impact evaluation in quality improvement

Time series plots are widely used, across sectors and media, probably because many find them easy to understand. Figure 1 is a time series plot of how the readmission rate in a hospital changed over time (constructed data set). Figure 1 Example of fictive time series data displaying the percentage of discharged patients readmitted to the hospital within 30 days for each month. The time series stretches over 31 months. After month 17, a quality improvement programme was introduced to reduce the proportion of patients readmitted to the hospital. Statistical process control (SPC) and interrupted time series (ITS) designs are two closely related methodologies in the field of quality improvement. In both approaches, data are organised in time series and presented using time series plots. Both SPC and ITS use data to assess whether observed changes reflect random variation or ‘real’ change. SPC is a popular method in quality improvement in the health sector worldwide, with scores of time series data collected. These data represent a golden but largely lost opportunity for learning and improving quality of care. First, because findings from SPC projects are rarely published: Thor and colleagues searched comprehensively for reports on the use of SPC in healthcare quality improvement and identified only 57 articles published between 1990 and 2004.1 Second—we will argue—because the potential for rigorous impact evaluation based on SPC data is not fully exploited. Simplified, SPC is a tool for monitoring processes by means of time series plots. Two key concepts in SPC are ‘common cause variation’ and ‘special cause variation’. Common cause variation implies that the observed variation reflects random fluctuations. When this is the case, the process is ‘in control’. On the other hand, a process is not in control when there is more variation than can be expected by chance alone, that is, the variation …