DNP 810 Module 6 Reading a Control Chart Example

Signal or Noise? Reading a P-Chart of Late Second Antibiotic Doses Before and After a Redosing Change

Student Name

Doctor of Nursing Practice Program, Aspen University

DNP 810: Evidence-based Practice for Quality Improvement

Instructor Name

Month Day, Year

Signal or Noise? Reading a P-Chart of Late Second Antibiotic Doses Before and After a Redosing Change

Every process varies. The proportion of late second antibiotic doses will differ from month to month even if nothing about care changes, simply because of which patients arrive and how busy the department is. The central question in evaluating an improvement is whether a change in the data is a signal, meaning a real change in the process, or noise, meaning ordinary variation. Statistical process control charts are designed to answer that question. This paper explains the chart chosen for the redosing project, shows how its center line and control limits are calculated, and reads a worked example using composite data for 12 baseline months and 6 intervention months.

Why Not Compare Two Averages?

A common alternative is to compare the average rate before the change with the average after it, perhaps with a chi-square test. That approach hides how the process behaved over time. It cannot distinguish a sudden shift that coincides with the change from a gradual improvement that began earlier, and it treats all months as interchangeable. Statistical process control keeps the time order of the data, distinguishes the everyday variation that is built into how the process works from the unusual variation that means the process itself has changed, and has been advocated as a tool for both improvement and research in health care for exactly these reasons (Benneyan et al., 2003). It also lets a team see results month by month while the project is under way, rather than waiting until the end to run a test.

Choosing the Chart

Control charts differ by the type of data. When the measure is a percentage, here the fraction of patients whose dose came late, and the count of eligible patients changes from month to month, the appropriate chart is the p-chart, which uses the binomial distribution and calculates control limits separately for each month's sample size. Tutorial guidance for health care practitioners describes this choice and the calculation of limits for attribute data such as proportions (Mohammed et al., 2008). A run chart, which plots data around the median without control limits, is simpler and useful early, since it can detect shifts and trends with as few as 10 to 20 points using probability-based rules (Perla et al., 2011). The project uses a run chart during the pilot and a p-chart once enough data exist.

The Baseline Data

The composite baseline covers 12 months. Across those months, 175 of 460 eligible patients had a major second-dose delay, giving a center line, the overall proportion, of 175 divided by 460, or 0.380. Each month's upper and lower control limits are the center line plus or minus three standard errors, calculated as the square root of 0.380 times 0.620 divided by that month's number of patients. For a month with 38 patients, the standard error is about 0.079, so the limits are 0.380 plus or minus 0.236, or 0.144 to 0.617.

Reading the Baseline

All 12 baseline months fall within their control limits, with points scattered above and below the center line and no runs or trends that would signal special causes. The process is therefore stable: late second doses occur at a predictable rate of about 38%, varying between about 34% and 43% by chance. Stability matters because it means the baseline is a fair comparison, and because a stable process that performs poorly cannot be improved by reacting to individual bad months. It needs a change to the system itself.

Reading the Intervention Months

Standard rules identify special cause variation, including a single point outside the control limits, and eight or more points in a row falling above, or below, the center line (Mohammed et al., 2008). In the composite data, the first intervention month, when the pilot covered only day shifts, fell to 0.289, below the center line but within limits. All four points from the third month onward fall below their lower control limits, a clear signal that the process has changed. All six intervention months fall below the baseline center line, and although six points do not yet meet the eight-point run rule, the points outside the limits are sufficient evidence of a shift.

The appropriate response to a sustained shift is to recalculate the center line and limits using the new data once enough points are available, typically after about 12 months, so that the new level becomes the standard against which future months are judged.

Cautions

Control charts show that a process changed, not why. Other changes during the period, such as a new hospitalist schedule or a fall in boarding times, could contribute, so the team must annotate the chart with known events and check process measures, such as the proportion of second doses ordered within 30 minutes, to confirm that the intervention was actually delivered. Small monthly samples also produce wide limits, so modest improvements may take several months to become visible as signals. Finally, the data here are composite and illustrative; the real project's chart will be read the same way when its data are available.

Conclusion

A p-chart turns monthly proportions of late second doses into a picture of a stable baseline and, in the composite example, a clear shift after the redosing change, with four months below their lower control limits. Choosing the right chart, calculating month-specific limits, applying standard rules and remembering what charts cannot show allow a doctoral nurse to judge improvement without overreacting to noise or missing a real signal.

References

Benneyan, J. C., Lloyd, R. C., & Plsek, P. E. (2003). Statistical process control as a tool for research and healthcare improvement. Quality and Safety in Health Care, 12(6), 458-464. https://doi.org/10.1136/qhc.12.6.458

Mohammed, M. A., Worthington, P., & Woodall, W. H. (2008). Plotting basic control charts: Tutorial notes for healthcare practitioners. Quality and Safety in Health Care, 17(2), 137-145. https://doi.org/10.1136/qshc.2004.012047

Perla, R. J., Provost, L. P., & Murray, S. K. (2011). The run chart: A simple analytical tool for learning from variation in healthcare processes. BMJ Quality & Safety, 20(1), 46-51. https://doi.org/10.1136/bmjqs.2009.037895