Run Charts for Operational Teams: Practical Guidance for NHS Improvement

Why this topic matters

NHS operational teams are constantly striving to improve efficiency, patient flow, and quality of care. Whether you're tracking emergency department waiting times, theatre utilisation, clinic DNA rates, or medication errors, you're dealing with data that varies over time. The challenge is discerning whether observed ups and downs represent true changes in the underlying process or just common cause variation – the natural, random 'noise' inherent in any system.

Misinterpreting common cause variation as a signal for action can lead to 'tampering' – making unnecessary changes that often worsen the system. Conversely, failing to recognise a genuine signal (special cause variation) means missing opportunities to embed improvements or address emerging problems. Run charts provide a visual and statistically sound method to navigate this challenge, offering a clear way to distinguish between these types of variation and guide appropriate action.

Practical explanation: What is a Run Chart?

A run chart is a simple line graph of data collected over time, with a median line added. Its power comes from a set of rules used to interpret the sequence of data points relative to this median. These rules help to identify patterns that are unlikely to occur by chance, thereby signalling that a 'special cause' has influenced the process.

Key Components of a Run Chart:

• Data Points: Individual measurements plotted in the order they occurred (e.g., daily number of discharges, weekly average length of stay).
• Time Axis (X-axis): Represents the chronological order of data collection.
• Measurement Axis (Y-axis): Represents the value of the measurement being tracked.
• Median Line: The central line on the chart, representing the median of all data points. The median is preferred over the mean for run charts as it is less affected by outliers.

Rules for Interpreting Run Charts (Western Electric Rules):

These rules help identify non-random patterns, indicating special cause variation. It's important to remember that these rules are indicators, not definitive proof, and should prompt investigation.

1. Shift: Six or more consecutive points above or below the median. (Points on the median do not count towards a run and are ignored for this rule).
2. Trend: Five or more consecutive points all going up or all going down. (A flat line is not considered part of a trend).
3. Too Many/Too Few Runs: A 'run' is a series of consecutive points on one side of the median. The number of runs observed is compared against expected values (often using a table or software). Too few runs suggest a shift or trend, while too many runs suggest oscillation.
4. Astronomical Point: A point that is clearly an outlier, visually separate from the rest of the data. While subjective, these often warrant immediate investigation.

When a rule is met, it signals that something has changed in the process beyond common cause variation. This is an opportunity to investigate – what happened at that time? Was it a new intervention, a change in staffing, or an external factor? This investigation is key to learning and improvement.

Common pitfalls

• Mistaking Common Cause for Special Cause: The most frequent error is reacting to every data fluctuation as if it's a signal. This leads to 'tampering' and instability.
• Using the Mean Instead of the Median: The mean is susceptible to outliers, which can distort the central line and make it harder to detect true shifts.
• Insufficient Data Points: Run charts become more reliable with more data. Aim for at least 10-12 points before applying the rules rigorously. For rule 3 (too many/too few runs), more points (e.g., 20+) are generally needed.
• Ignoring the Context: Data never tells the full story alone. Always combine chart interpretation with local knowledge and team insights.
• Over-interpreting Single Points: One isolated point above or below the median is usually common cause variation unless it's an astronomical point.
• Lack of Action: Detecting a special cause is only useful if it prompts investigation and action. The chart is a tool for understanding, not an end in itself.

Step-by-step approach to creating and interpreting a run chart

1. Define Your Measure: Clearly identify what you are tracking (e.g., 'proportion of patients discharged before noon', 'average theatre turnaround time', 'number of falls per ward'). Ensure it's measurable and relevant to your improvement goal.
2. Collect Data: Collect data points over time. Consistency in data collection is vital. Daily or weekly data is often suitable for operational processes.
3. Plot the Data: Create a simple line graph with time on the X-axis and your measure on the Y-axis. Plot each data point in chronological order.
4. Calculate and Draw the Median: Order all your data points from smallest to largest and find the middle value (the median). Draw a horizontal line across your chart at this median value.
5. Apply the Rules: Systematically review your plotted data against the four run chart rules:
   • Look for six or more consecutive points above or below the median (shift).
   • Look for five or more consecutive points all going up or all going down (trend).
   • Count the number of runs (series of points on one side of the median). Refer to a statistical process control (SPC) resource or software for guidance on expected number of runs for your data points.
   • Look for any obvious 'astronomical points'.
6. Interpret and Act:
   • If no rules are met, the process is stable and exhibiting common cause variation. Focus on improving the system itself, rather than reacting to individual fluctuations.
   • If a rule is met, a special cause variation has been detected. Investigate what might have caused this change. Did an intervention occur? Was there an unusual event? Use this insight to embed improvements or address new problems.
7. Sustain and Review: Continue tracking the data. If you implement a change, you might expect to see a new shift or trend on your run chart, indicating the change has had an effect.

Example in clinical practice: Reducing ward discharge delays

A medical ward team aimed to reduce the proportion of patients discharged after 4 pm. They decided to track the 'percentage of discharges before 4 pm' daily for three months.

1. Measure: Percentage of patients discharged before 4 pm.
2. Data Collection: Daily percentages were recorded for 60 days.
3. Plot Data: A line graph was created showing the daily percentage over time.
4. Calculate Median: The median percentage of discharges before 4 pm was calculated as 65%.
5. Apply Rules:
   • Initial Period (Days 1-30): The chart showed points fluctuating around the 65% median. No clear shifts or trends were observed. The team concluded this was common cause variation – the process was stable, but not achieving their desired level of improvement.
   • Intervention (Day 31): The team implemented a new 'Discharge Huddle' at 9 am daily, involving nursing, medical, and therapy staff, to proactively identify and address barriers to early discharge.
   • Post-Intervention (Days 32-60): The run chart showed a noticeable pattern. From day 35 onwards, eight consecutive points were above the 65% median. This met the 'shift' rule.
6. Interpret and Act: The shift indicated that the Discharge Huddle intervention had likely caused a statistically significant improvement in early discharges, moving the process to a new, higher level of performance (e.g., now consistently achieving 75-80% early discharges). The team decided to embed the huddle as standard practice and monitor for sustainment, potentially recalculating a new median after a period of stable performance at the improved level.

This resource supports, but does not replace, clinical judgement. Local policy, formulary and specialist advice should be followed.

How Lazomis can help

Lazomis provides intuitive tools to support your data collection and run chart analysis. Our Data Capture & Visualisation module allows you to easily input your daily or weekly operational data. The system can then automatically generate clear run charts, complete with median lines, and even highlight potential special cause signals based on the run chart rules. This reduces the manual effort of plotting and calculating, freeing up your team to focus on understanding the 'why' behind the data and designing effective interventions. Our dashboards can then help you present these insights clearly to your team and stakeholders, demonstrating the impact of your improvement work.

Key takeaways