A single before-and-after number cannot tell you whether your quality improvement change actually worked, or whether the process was always going to bounce around that much on its own. This guide explains run charts and control charts in plain terms: what each one shows, how to build them for a DNP QI project, and how to write up what you find without overclaiming.
Quick answer. A before-and-after comparison cannot separate a real improvement from ordinary process variation. Run charts and control charts plot your measure over time so a genuine shift, called special-cause variation, can be distinguished from the common-cause variation every process has.
Most DNP quality improvement projects compare a measure before an intervention with the same measure after it. That comparison is intuitive, and it is also incomplete. Every process, from fall rates to hand-hygiene compliance to catheter-associated infection counts, moves up and down from week to week even when nothing about the process has changed. If you only look at one number before and one number after, you cannot tell whether the difference you observed is a real, lasting shift or whether it is simply where the process happened to land that particular week.
This is exactly the distinction that statistical process control, often shortened to SPC, exists to make. SPC methodology was developed originally for manufacturing but has been adapted widely for healthcare improvement work, including by the Institute for Healthcare Improvement, because the underlying problem is the same: how do you know a change produced an effect, rather than coincidence? Charting your measure over time, both before and after the intervention, is what lets you answer that question with more confidence than a simple two-point comparison ever could.
A DNP project is evaluated in part on whether your evaluation approach is methodologically sound, not just on whether the after-number looked better than the before-number. Committees and faculty who are familiar with improvement science will often ask directly whether you used a run chart or control chart, and a project that relies solely on a pre-post average invites the question of whether the "improvement" is anything more than noise. Building a process-behavior chart into your evaluation plan from the start, rather than retrofitting one at the end, also shapes how you collect data, since these charts need a series of time-ordered data points rather than two aggregate numbers.
Understanding this distinction is the conceptual foundation for everything else in this guide, and it is worth stating clearly before moving into chart mechanics.
| Type of variation | What it is | What it means for your project |
|---|---|---|
| Common-cause variation | The normal, expected variation that is baked into any process simply because of routine differences in staffing, patients, timing and countless small factors | Ups and downs within the usual range should not be read as evidence that your intervention worked or failed |
| Special-cause variation | A signal that something genuinely different happened to the process, distinguishable from routine noise by specific pattern rules | This is the kind of shift a successful QI intervention is meant to produce, and what your chart is designed to help you detect |
A process that shows only common-cause variation is described as being "in statistical control," which is a technical way of saying it is behaving predictably within its usual range, even if that range is not where you want it to be. The goal of a QI intervention is often to move the process to a new, better level of performance, which shows up on a chart as a special-cause signal followed by a new, sustained pattern.
A run chart is the simpler of the two chart types covered in this guide, and for many DNP projects with a modest number of data points, it is the appropriate starting point. A run chart plots your data points in time order against a center line, which is usually the median of the data rather than the mean, because the median is less influenced by unusual outlier points.
The value of a run chart is not any single point but the pattern the points make relative to the center line over time. Analysts use a small set of non-random-pattern rules to decide whether a pattern is unlikely to have occurred by chance alone. Commonly cited rules include a long run of consecutive points all on one side of the median, a clear and sustained trend of points consistently moving in one direction, and an unusually long run of points overall. The Institute for Healthcare Improvement publishes widely used run-chart rules of this kind, and its materials are a reasonable, well-known reference point for a DNP methods section. Because the exact numeric thresholds used for these rules can be described slightly differently across sources and have been updated over time, confirm the specific thresholds against the current IHI guidance or whatever methodology your program specifies, rather than relying on a number reproduced secondhand.
A run chart is a useful, low-barrier first step, but its signal rules are simpler heuristics rather than calculated statistical limits. It does not, on its own, quantify how much variation is "normal" for your specific process the way a control chart's control limits do. Many QI teams start with a run chart during early PDSA cycles, when the number of data points is still small, and move to a control chart once there is enough baseline data to calculate stable limits.
A control chart takes the same idea, plotting data over time against a center line, and adds calculated upper and lower control limits based on the natural variation already present in your own process data. Points that fall outside those limits, or that follow other recognized non-random patterns within them, are read as a more statistically rigorous signal of special-cause variation than a run chart's simpler rules provide.
An important conceptual point, and one that is easy to get wrong in a methods write-up, is that control limits are calculated from the variation already present in your process's own baseline data. They are not an arbitrary target or a benchmark borrowed from another organization. This is what makes a control chart a statement about whether your process changed relative to its own historical behavior, rather than a comparison to an external standard.
Healthcare improvement literature describes several common control chart types, and choosing the right one depends primarily on what kind of data you are plotting. Getting this choice right matters more for the credibility of your methods section than most students expect, since a mismatched chart type is one of the more common reviewer comments on improvement-focused DNP projects.
| Data type | Example measure | Commonly used chart |
|---|---|---|
| Proportions or percentages | Percent of patients screened, infection rate per population | P-chart |
| Individual continuous measurements | Average length of stay, a single measured value per time period | XmR or I-MR chart |
| Counts of rare events | Number of falls per unit per month | C-chart or U-chart, depending on whether the opportunity for the event varies |
This guide deliberately does not walk through the underlying control-limit formulas for each chart type, since getting a formula slightly wrong from memory is worse than pointing you to a reliable source. Most DNP programs point students toward software, a statistics consultant, or a methods text for the calculation itself; your job as the project author is to choose the right chart type for your data and to interpret the output correctly, which is where most of the substantive academic reasoning actually happens.
The practical sequence below reflects how these charts are typically built for an improvement project, regardless of which specific chart type you use.
Presenting a "before" chart and an "after" chart as two separate images, side by side, is a frequent mistake. It reintroduces the exact problem this whole methodology exists to solve, because it invites the reader back into a simple before-and-after comparison instead of showing the continuous process behavior that a run or control chart is meant to reveal.
Share your data description, chart type and project brief, and a specialist can help you draft or edit the evaluation chapter, including how to describe and interpret your chart honestly. The price is shown before you pay, and every delivered paper includes 14 days of free revisions.
How you describe your chart's findings in the write-up matters as much as building the chart correctly. A shift that meets a recognized signal rule can reasonably be described as a likely special-cause improvement associated with your intervention. It should not be described as proof, and it should not be described using causal language stronger than your design supports.
| Overclaimed | Better fit for the evidence |
|---|---|
| The intervention caused a 100 percent improvement in compliance. | Following the intervention, the process showed a sustained shift consistent with special-cause improvement, based on [the specific rule met]. |
| The data prove the change worked. | The pattern observed is consistent with improvement, though the short observation window and single-unit setting limit how confidently this can be generalized. |
| Results were statistically significant. | A signal rule was met on the chart, which is a different claim than a hypothesis-test significance result, and should not be described in that language unless a formal test was also run. |
A strong write-up discusses, in the same section as the chart, the small sample size typical of a single-unit DNP project, the short time frame available for a capstone term, and other plausible explanations for an observed shift, such as a concurrent change the project did not control for. Naming these limitations does not weaken your findings; it is exactly the kind of honest, evidence-calibrated reasoning that improvement science and DNP faculty are looking for.
Illustrative example, not a real client. This short story is invented to show the pattern, and it contains no real people or numbers.
The problem. A DNP student had planned to compare a single "before" percentage against a single "after" percentage for a hand-hygiene compliance project, and her committee flagged that this design could not distinguish improvement from ordinary week-to-week variation.
The tension. She had already collected several weeks of baseline data as spot audits, but had not planned to continue the same audit through the intervention period on one continuous chart.
The turn. She restructured her data collection to continue the identical weekly audit through and after the intervention, plotted it as a single run chart with the intervention date marked, and worked with her chair to confirm which IHI run-chart rule would count as a signal for her project.
The proof. The chart showed a run of consecutive points above the baseline median beginning shortly after the intervention, meeting the rule her committee had agreed on in advance.
The payoff. Her evaluation section described the shift as a likely special-cause improvement rather than a proven cause-and-effect result, and her committee's feedback moved from questioning the design to refining the discussion of limitations.
Many programs accept charts built in spreadsheet software with add-ins, dedicated QI software, or statistical packages. Ask your program or chair which tools are acceptable and whether a specific one is expected.
Enough to establish a stable baseline is the general principle; the exact number recommended varies by source and chart type, so confirm the current guidance from your methodology reference rather than assuming a fixed number.
It depends on your data volume and your program's expectations. A run chart is a reasonable and commonly accepted starting point, especially with a limited number of PDSA cycles, while a control chart is often expected when enough baseline data exists to calculate stable limits.
Yes, run and control charts are one tool within a broader evaluation plan. Our DNP outcomes evaluation guide covers how this fits into the larger evaluation chapter, including outcome, process and balancing measures.
That is a legitimate and reportable finding. Describe honestly that the process remained within common-cause variation, discuss possible reasons, and consider what this means for practice recommendations, rather than searching for a way to claim a signal that the data do not support.
The same logic applies any time you are trying to show a process changed over time. See our evidence-based practice change project guide for how this fits alongside a broader implementation plan.
A chart that plots your measure over time, marks the intervention clearly, and applies a named signal rule tells a far more credible improvement story than two numbers and an arrow between them. Choose the chart type that matches your data, build baseline and post-intervention points on one continuous timeline, and write up what you find with the same honesty you would want from anyone else's improvement claim.
Want help drafting or tightening your QI evaluation chapter? Get my instant quote. The price is shown before you pay, every delivered paper includes 14 days of free revisions, and refund terms are on the money-back guarantee page. Please use any model paper in line with your institution's academic-integrity policy.