DNP Project Data Analysis: Run Charts, Statistical Process Control, and Pre/Post Comparisons

The most common analytical mistake in a DNP scholarly project is treating it like a small research study. A candidate collects a baseline measure, implements an intervention, collects the measure again, runs a paired t-test, and reports whether the difference was significant. That analysis is not wrong, exactly, but it is rarely sufficient. It often misses what the committee actually wants to know. Did the process change, did the change hold, and was the change real rather than the ordinary fluctuation every clinical measure shows week to week? A DNP project is quality improvement work, not knowledge-generating research, and quality improvement has its own analytical toolkit built precisely for that question. Run charts and statistical process control charts display data over time and distinguish real signals from noise, which two summary numbers cannot do. This guide sets out how to analyze DNP project data properly, when a pre-/post-test is still appropriate, and how to report the whole thing under SQUIRE 2.0. When you want the analysis built and documented to that standard, our dissertation data analysis service covers DNP and quality improvement work directly.

Quick Answer:

A DNP scholarly project is practice-focused quality improvement, so its analysis usually centers on data displayed over time rather than a single before-and-after comparison. A run chart plots your measure in time order against the median and uses four rules, shift, trend, number of runs, and astronomical point, to identify non-random signals. A control chart adds statistically derived control limits and separates common-cause variation, the ordinary noise of a stable process, from special-cause variation, which signals real change. Where a pre/post comparison is still needed, use a paired-samples t-test for a continuous outcome, the Wilcoxon signed-rank test when its assumptions fail, and McNemar's test for a paired categorical outcome. Report the project under SQUIRE 2.0, the reporting standard for healthcare improvement work.

Why a DNP Project Is Analyzed Differently

The distinction that drives everything else is the one between a practice doctorate and a research doctorate. A PhD dissertation generates new, generalizable knowledge. A DNP scholarly project translates existing evidence into practice and evaluates whether the change improved care in a specific setting. The American Association of Colleges of Nursing has framed the DNP project in exactly these terms since the 2006 Essentials. The practice-focused student carries out a practice application-oriented final project rather than a knowledge-generating research effort. The 2021 competency-based Essentials continue to emphasize quality, safety, and data literacy as core capabilities.

This changes the analytical question. Research asks whether an effect exists in a population. Improvement work asks whether this process, in this unit, actually got better and stayed better. Answering that requires seeing the data in time order, because a process is a thing that unfolds, and two aggregate numbers taken months apart conceal everything that happened in between. A unit whose infection rate fell, rose again, and settled back where it started can produce the same pre/post difference as a unit that improved steadily and held the gain. Only a time-ordered display tells them apart. Getting this framing right from the start also depends on a well-defined practice problem, which our guide on the DNP practice gap addresses, and on a clinical question framed with PICOT, covered in our guide on writing a PICOT question.

The Run Chart: Your Primary Analytical Tool

A run chart is a line graph of a quality measure plotted in time order, with the median of the baseline data drawn as a center line. It is the workhorse of improvement analysis because it is simple to build, simple to read, and answers the question that matters: Is something non-random happening here?

The logic rests on probability. If a process is stable and unchanged, data points should fall above and below the median in a random pattern. Departures from randomness are unlikely by chance and therefore suggest a real change in the process. Perla, Provost, and Murray set out the standard approach to constructing and interpreting run charts for healthcare in their widely used BMJ Quality and Safety paper, drawing the rules from the statistical process control literature (Perla, Provost, & Murray, 2011). Their central argument is worth carrying into your defense: plotting data over time yields richer information and more accurate conclusions than summary statistics taken before and after a change.

To build one, plot your measure on the vertical axis and time, usually days, weeks, or months, on the horizontal. Calculate the median of your baseline points and extend it as the center line. Annotate the chart to show when the intervention was introduced, and add annotations for any other events that could plausibly affect the measure, such as a staffing change or a policy update. Then apply the rules.

The Four Run Chart Rules

Four rules identify non-random signals on a run chart. Three are probability-based; the fourth is explicitly a matter of judgment, and saying so signals that you understand the tool rather than applying it mechanically.

Rule 1: a shift. Six or more consecutive points all above or all below the median indicate a shift in the process. Points that fall exactly on the median are skipped: they neither add to nor break a run.

Rule 2: a trend. Five or more consecutive points, all increasing or all decreasing, indicate a trend. Where two consecutive points are equal, one is ignored when counting.

Rule 3: the number of runs. A run is a series of consecutive points on one side of the median. Too few or too many runs, judged against a published table of critical values for your number of data points, indicates a non-random pattern.

Rule 4: an astronomical point. A data point that is obviously, even blatantly, different from all the others, such that anyone looking at the chart would agree it is unusual. Perla and colleagues are explicit that while the first three rules are probability-based, this fourth rule is subjective, and they caution that an astronomical point should not be confused with the simple highest or lowest point, since every chart has one of those.

Table 1: The Four Run Chart Rules for Detecting Non-Random Signals

Applying these rules lets you make a defensible statement, such as that a shift occurred beginning three weeks after implementation and was sustained for the remainder of the measurement period. That is a far stronger finding than a p-value from a single comparison, because it locates the change in time and shows whether it held.

Control Charts: Adding Statistical Limits

A statistical process control chart, also called a Shewhart chart, extends the run chart by adding a center line at the mean and upper and lower control limits derived from the variation in the data itself, conventionally at three standard deviations. Its purpose is to distinguish two fundamentally different kinds of variation.

Common-cause variation is the ordinary, inherent noise of a stable process: the week-to-week fluctuation that occurs even when nothing has changed. Special-cause variation is variation arising from something outside the usual process, a signal that something real has happened. This distinction is the conceptual heart of improvement analysis, because reacting to common-cause variation as if it were a signal, a habit sometimes called tampering, wastes effort and can make a process worse. A control chart tells you when a change is worth acting on.

Two points about control limits are worth stating precisely, because candidates confuse them. First, control limits are calculated from the process data; they are not targets, specifications, or goals, and a process can be perfectly stable while consistently missing its target. Second, choosing the right chart type matters, and it depends on the kind of data: charts for continuous measurements differ from charts for proportions, rates, and counts. Selecting the wrong chart type produces misleading limits. Statistical process control has an established evidence base in healthcare improvement, and its application has been reviewed systematically in the quality and safety literature, so citing that basis strengthens your methods section. This is the level of analytical detail our DNP capstone support is built around.

The Model for Improvement and Where Analysis Fits

Most DNP projects sit inside a formal improvement framework, and the most widely used is the Model for Improvement, associated with the Institute for Healthcare Improvement. It asks three questions: what are we trying to accomplish, how will we know that a change is an improvement, and what changes can we make that will result in improvement, and it tests changes through Plan-Do-Study-Act cycles.

The second question is where your analysis lives. "How will we know that a change is an improvement?" is answered by your measures and the charts that display them, which is why measure selection deserves as much attention as the intervention. Improvement work conventionally uses three kinds of measures: outcome measures, which capture the result that matters to patients; process measures, which capture whether the intervention is actually being delivered; and balancing measures, which check that improving one thing has not degraded another. Reporting all three is a mark of a mature project, because a process measure explains why an outcome did or did not move, and a balancing measure shows you looked for unintended harm. Each measure gets its own chart, plotted over time.

When a Pre/Post Comparison Is Still Appropriate

None of this means hypothesis testing has no place in a DNP project. Committees frequently expect a pre/post comparison alongside the time-series display, and it is appropriate when you have discrete before-and-after measurements rather than a long series of time points. The test follows the same logic as any analysis: it depends on your measurement level and the fact that the observations are paired.

For a continuous outcome measured before and after on the same patients or units, use the paired-samples t-test. When its assumptions do not hold, particularly with small samples or clearly non-normal data, use the Wilcoxon signed-rank test. For a paired categorical outcome, such as whether each patient was compliant before and after, use McNemar's test, which is designed for paired binary data. This is the test DNP candidates most often overlook, defaulting incorrectly to a chi-square test of independence that assumes independent groups. Where you are comparing independent groups rather than paired observations, the usual independent-samples tests apply instead.

Table 2: Choosing a Pre/Post Test for a DNP Project

Whatever test you use, report it fully: the statistic, the exact p-value, an effect size, and a confidence interval. And be careful about the inference. A statistically significant pre/post difference does not by itself establish that your intervention caused the change, because a before-and-after design without a control has no protection against secular trends, seasonal effects, or other concurrent changes. This is precisely why the time-ordered chart matters: it shows whether the change coincided with your intervention and whether it was sustained. Present the two together, with the chart carrying the causal argument and the test quantifying the magnitude.

Statistical Significance Versus Clinical Significance

A distinction that carries particular weight in nursing and improvement work is the one between statistical and clinical significance. A change can be statistically significant and clinically trivial, especially with a large number of observations, and it can be clinically important but fall short of significance in a small project. Your committee cares about whether the change matters to patients and to the unit. Report the magnitude of the change in its natural units, such as percentage points of compliance or minutes of delay, alongside any test result, and discuss what that magnitude means in practice. This is also where sustainability belongs: an improvement that held for the duration of measurement is a different finding from one that decayed, and only your chart can show which you have. Framing the clinical meaning well depends on having framed the clinical question well, which our EBP and PICOT support, addressing directly.

Reporting Your Project Under SQUIRE 2.0

Quality improvement work has its own reporting standard, and using it signals to your committee and to any journal that you understand the genre. SQUIRE 2.0, the Standards for Quality Improvement Reporting Excellence, was published by Ogrinc and colleagues after a multi-year consensus revision and appears in both BMJ Quality and Safety and, for a nursing readership, the Journal of Nursing Care Quality (Ogrinc et al., 2016). It follows an IMRaD structure and emphasizes three components that distinguish improvement work from research: the use of formal and informal theory in planning and evaluating the work, the context in which the work was done, and the study of the intervention itself.

For your analysis section specifically, SQUIRE 2.0 expects you to describe the approach used to assess whether the observed outcomes were due to the intervention, and the methods used to understand variation within the data, including the effects of time as a variable. That phrasing is essentially an invitation to present run charts or control charts, and it is why a project analyzed only with a single pre/post test can look thin against the standard. Report your measures, your charts, the signals you identified, and the rules that flagged them, any pre/post tests with effect sizes, and your interpretation of both statistical and clinical significance. The full guideline and its explanation documents are available at the SQUIRE statement website, and citing them in your methods is clear evidence of rigor. Structuring the whole project this way is what our capstone project support is built to deliver.

Frequently Asked Questions

How is DNP project data analysis different from PhD dissertation analysis?

A DNP scholarly project is practice-focused quality improvement that translates evidence into practice and evaluates the result in a specific setting, whereas a PhD dissertation generates new generalizable knowledge. That difference changes the analysis: improvement work asks whether a process actually changed and whether the change held, which requires displaying data over time using run charts or control charts, rather than answering a population-level question with a single hypothesis test.

What is a run chart, and how do I interpret it?

A run chart is a line graph of your measure plotted in time order with the baseline median as a center line. If the process is unchanged, points fall above and below the median randomly. Four rules identify non-random signals: a shift (six or more consecutive points on one side of the median), a trend (five or more consecutive points all rising or all falling), too few or too many runs, and an astronomical point. Annotate the chart to show when the intervention began.

What is the difference between common-cause and special-cause variation?

Common-cause variation is the ordinary, inherent fluctuation of a stable process, the noise that occurs even when nothing has changed. Special-cause variation comes from something outside the usual process and signals that something real has happened. A control chart separates the two using control limits derived from the data. The distinction matters because reacting to common-cause variation as if it were a signal wastes effort and can make a process worse.

When should I use a control chart instead of a run chart?

Use a run chart when you want a simple, accessible display and have a modest number of time points; it requires no assumptions about the distribution and is easy for a clinical audience to read. Use a control chart when you want statistically derived limits that formally separate common-cause from special-cause variation, and when you have enough data points to estimate those limits reliably. Choose the chart type to match your data, since charts for continuous measures, proportions, rates, and counts differ.

Which statistical test should I use for a pre-/post-DNP comparison?

It depends on the measurement level and on the fact that the observations are paired. For a continuous outcome measured before and after on the same units, use the paired-samples t-test, or the Wilcoxon signed-rank test when its assumptions fail. For a paired categorical outcome, such as compliant or not compliant before and after, use McNemar's test. Candidates frequently err by using a chi-square test of independence, which assumes independent groups rather than paired observations.

What is SQUIRE 2.0, and do I have to use it?

SQUIRE 2.0 is the Standards for Quality Improvement Reporting Excellence, the reporting guideline for healthcare improvement work, published by Ogrinc and colleagues in 2016 in BMJ Quality and Safety and the Journal of Nursing Care Quality. It uses an IMRaD structure and emphasizes theory, context, and the study of the intervention. Most DNP programs expect it, and their analysis items specifically ask you to describe how you assessed whether outcomes were due to the intervention and how you examined variation over time.

Do I need statistical significance for my DNP project to succeed?

No. DNP projects are evaluated on whether they translated evidence into practice and produced a meaningful, sustained improvement in a specific setting, not on achieving a significant p-value. Clinical significance, the practical magnitude of the change, and whether it held, often matter more to a committee than statistical significance. Report the change in its natural units alongside any test result, and use your time-series chart to show whether the improvement was real and sustained.

Showing That the Change Was Real

A DNP project succeeds analytically when it can show, not merely assert, that a process changed and that the change held. Plot every measure over time, annotate when the intervention began, and apply the run chart rules or control limits to identify genuine signals. Add a pre-/post-test where your committee expects one, reported with its effect size and interpreted for clinical as well as statistical meaning. Then report the whole thing under SQUIRE 2.0. Do that, and your project stops resting on a single p-value and starts demonstrating improvement the way improvement science does.

If you would rather a methodologist build your charts, run the right pre/post tests, and write the analysis to SQUIRE 2.0, send us your project measures and data. You will have an itemized quote within 2 to 4 business hours, with no obligation.