Understanding Variation in Healthcare Data: A Practical Guide for Improvement

Why this topic matters

Misinterpreting variation can lead to significant problems in healthcare. Reacting to common cause variation as if it were special cause (tampering) can destabilise a process, waste resources, or even worsen performance. Conversely, failing to identify special cause variation can mean missing crucial signals that a process is out of control, potentially impacting patient safety or efficiency.

For QI leads, clinical audit teams, and department managers, a clear understanding of variation is essential for:

• Effective Decision-Making: Knowing when to intervene and when to let a process stabilise.
• Targeted Improvement Efforts: Focusing resources on fundamental system changes rather than chasing random fluctuations.
• Accurate Performance Monitoring: Distinguishing genuine improvement (or deterioration) from natural process noise.
• Reduced Burnout: Preventing individuals from being unfairly targeted for 'poor' performance that is actually a product of the system.

Practical explanation: Common vs. Special Cause Variation

At its core, variation helps us understand the predictability and stability of a process.

Common Cause Variation (Expected Variation)

This refers to the inherent, random, and predictable variation that is always present in a stable process. It's the 'system noise' that arises from the way a process is designed and typically operates. If you observe any process long enough, you'll see this kind of variation. It's usually within a predictable range, forming a stable pattern.

Characteristics:

• Random and statistical in nature.
• Affects all process outputs or measurements.
• Part of the system itself, reflecting its current capability.
• To reduce common cause variation, the system (process design, resources, training) itself must be fundamentally changed.

Example: The slight differences in the time it takes for a patient to be discharged each day, assuming no extraordinary events occurred. Some days it might be a bit quicker, some days a bit slower, but generally it falls within a predictable range.

Special Cause Variation (Unexpected Variation)

This refers to variation that arises from specific, identifiable, and often transient factors that are not inherent to the process. It's an 'event' or an 'anomaly' that causes the process to behave differently from its usual pattern.

Characteristics:

• Assignably caused by specific events or factors.
• Appears sporadically and outside the normal expected range.
• Indicates a process is out of statistical control or has been affected by an unusual event.
• To address special cause variation, the specific event or factor needs to be investigated and managed.

Example: A sudden, sharp increase in patient waiting times during a particular week due to an unexpected staff sickness outbreak, or a new malfunctioning piece of equipment.

Statistical Process Control (SPC) Charts

SPC charts (often called control charts) are graphical tools used to monitor a process over time, helping to distinguish common from special cause variation. They plot data points sequentially and establish control limits based on the process's own historical performance.

How SPC Charts Work

An SPC chart typically includes:

1. A Centre Line (CL): Represents the average or mean of the process data.
2. Upper Control Limit (UCL) and Lower Control Limit (LCL): These are statistically derived boundaries (usually ±3 standard deviations from the mean) that define the expected range of common cause variation. They are not targets or specification limits but are calculated from the process data itself.

Data points falling within the control limits suggest that the process is stable and only experiencing common cause variation. Points or patterns outside these limits signal the presence of special cause variation, warranting investigation.

Common Rules for Identifying Special Cause Variation

While there are various rules, commonly accepted ones (e.g., Western Electric Rules, Nelson Rules) include:

• Points Outside Control Limits: Any single point above the UCL or below the LCL.
• Runs of Points: Seven or more consecutive points all above or all below the centre line.
• Trends: Seven or more consecutive points all steadily increasing or decreasing.
• Proximity to Control Limits: Two or three out of three consecutive points near a control limit.

It's important to use these rules consistently and understand their statistical basis. Local QI teams or statisticians can provide guidance on rule selection and application.

Common pitfalls

• Reacting to Common Cause: Making changes to a stable process in response to random fluctuations. This often leads to 'tampering' and can degrade performance.
• Ignoring Special Cause: Missing crucial signals that a process is out of control, leading to prolonged suboptimal performance or patient safety risks.
• Using Control Limits as Targets: Control limits describe the process's current capability; they are not desired performance levels. Setting targets based on control limits is a misuse of the tool.
• Calculating Limits on Unstable Data: Control limits should ideally be calculated from a period when the process was stable. If there's special cause variation in the baseline data, the limits will be skewed.
• Over-reliance on Automated Systems: While software can generate SPC charts, understanding the underlying principles and interpreting patterns still requires human oversight and clinical context.

Step-by-step approach to using SPC Charts for Improvement

1. Define Your Process and Measure: Clearly identify what you want to improve and select a measure (e.g., 'time from decision to admit to ward bed', 'percentage of patients receiving appropriate VTE prophylaxis'). Ensure data is collected consistently.
2. Collect Baseline Data: Gather at least 20-25 data points under normal operating conditions. This allows for stable calculation of control limits.
3. Choose the Right SPC Chart: Different data types require different charts (e.g., Xbar-R for continuous data, P chart for proportions, C chart for counts of defects). Consult with a QI lead or statistician if unsure.
4. Plot the Data and Calculate Limits: Use software (e.g., Excel, specialist QI tools) to plot your data and calculate the Centre Line, UCL, and LCL.
5. Interpret the Chart: Look for patterns that indicate special cause variation using the rules mentioned above. If special causes are present, investigate their root causes.
6. Act on Special Causes: If special cause variation is detected, investigate the specific factors that led to it. Implement targeted actions to eliminate undesirable special causes or embed desirable ones.
7. Sustain and Monitor: Once special causes are addressed and improvements are made, continue to monitor the chart to ensure sustained performance. Re-calculate control limits if significant, stable changes to the process have occurred.

Example in clinical practice: Reducing Delays in Discharge Summaries

A hospital department aims to reduce the time for discharge summaries to be completed and sent to the GP. They decide to monitor the daily percentage of discharge summaries completed within 24 hours of discharge.

1. Measure: Percentage of discharge summaries completed within 24 hours.
2. Baseline Data: They collect 30 days of data, recording the daily percentage.
3. Chart Type: Since they are tracking a proportion (successes out of total discharges), a P-chart is appropriate.
4. Plotting: The data is plotted, and the average completion rate is 82%, with a UCL of 95% and an LCL of 69%.
5. Interpretation: For the first couple of weeks, most points fall within the control limits, indicating common cause variation around 82%. However, one week, three consecutive points fall below the LCL (e.g., 65%, 62%, 60%).
6. Action: This 'special cause' pattern triggers an investigation. It's discovered that two junior doctors on the ward were new and unfamiliar with the digital discharge summary system, causing significant delays. Additional training and buddying were implemented.
7. Sustain: Following the intervention, the percentage returns to working within the control limits, and the team aims to further improve the centre line by streamlining the system itself, addressing common cause variation next.

This example shows how an SPC chart can provide actionable insights, prompting investigation only when a true signal emerges from the noise.

How Lazomis can help

Lazomis offers tools that can streamline the process of understanding and acting on variation in your healthcare data. Our platform allows for easy data input and provides intuitive visualisations, including pre-built SPC chart templates, to help you quickly identify trends and special cause variation without needing specialist statistical software. By integrating data collection with charting, Lazomis can help your QI teams focus more on interpretation and action, rather than manual data processing. For those new to SPC, our resources and templates can simplify initial setup and ongoing monitoring, encouraging data-driven decision-making across your department or organisation.

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