Belitsoft > Run Charts in Healthcare Data Analysis

Run Charts in Healthcare Data Analysis

Run charts, trend charts, and time series charts are essential tools for clear data visualization in healthcare analytics. They display data over time, revealing patterns and trends in healthcare settings. This article delves into the significance of these charts and explains how they can optimize data analytics processes in healthcare. At Belitsoft, we extend these insights by developing custom healthcare software equipped with advanced analytics and visualization tools to track and manage patient outcomes and workflows effectively. Keep reading to get the full scope of data’s role in enhancing healthcare decisions.


What is a Run Chart?

A run chart is a line graph that plots data points along a timeline. It highlights trends, shifts, or unusual patterns within a process. Run charts reveal insights about your process over time that tables or summaries might miss. These charts are valuable for analyzing smaller data sets before using more complex Shewhart control charts. 

In our experience, run charts serve not just as tools for analysis, but as narratives of the process being studied. You can see if things are stable, improving over time, getting worse, or showing dramatic variability. 

You also can observe shifts in data upwards or downwards after making changes. Improving a process is one thing, sustaining the improvement is another. Run charts help monitor over time to see if the gains you worked hard for are lasting.

Consider, for instance, how we can support primary care clinics in monitoring depression care improvement efforts. By employing run charts to track the percentage of patients attending follow-up visits, we enable clinics to visualize and quantify the impact of targeted changes aimed at improving patient attendance.

what is vision and scope document
Run chart example

This run chart tracks the percentage of outpatient follow-up attendance for depression care

The median is the middle value represented by a horizontal line in the chart. 

With a goal of achieving 85% attendance, the chart not only showcases an increasing trend over time but also highlightes the effectiveness of three key interventions: 

  • the introduction of a support group for outpatients with depression in February 2019
  • implementation of a new follow-up appointment system to simplify scheduling 
  • deployment of follow-up reminders to encourage patients to keep their appointment 

The resulting data vividly demonstrates a marked improvement in attendance rates post-intervention.

Belitsoft offers expert custom healthcare software development services, helping you turn complex healthcare data into clear insights with advanced analytics and visualization tools. Reach out for guidance.

Constructing a Run Chart

There are seven steps to construct a run chart. While software can automate most of the process, it's crucial to verify that the software's output follows established guidelines. 

Here's how we guide our clients through each step:

Formulating a focused question to evaluate the effectiveness of interventions

Every run chart begins with a clear, investigative question. The question may be, “Are more patients arriving on time for their appointments compared to the past?” This question sets the direction for our analysis. 

Data collection to monitor the impact of these interventions across a selected timeframe

Percentage of on-time appointments in clinic by month
Percentage of on-time appointments in clinic by month

Accordingly, the table may present the monthly percentage of on-time appointments, aggregating data from multiple patients.

Developing horizontal scale

To determine the horizontal scale, we select time intervals (such as days, weeks, or months) that accurately represent the collected data, and the observed changes.  

To prepare the chart for long-term use, we include space for future data.

Developing vertical scale

The vertical scale is designed to highlight the data effectively. We aim to have most of your current data points fall within the middle half of the chart's vertical axis.

Also we implement clear and consistent increments for your axis labels, ideally using round numbers that are evenly spaced.

For instance, if aiming for 85% on-time appointments, we might use 10% increments for clarity.

Our expertise ensures that the chart remains balanced, informative, and easy to interpret. For that we place tick marks at consistent intervals and make them visually distinct to help estimate values between labeled points.

We leave space on the scale to accommodate potential future data points that may fall outside the initial range.

In case you have a benchmark or standard for comparison, we include it on the scale to provide context and assess performance relative to the benchmark.

For the best visual presentation, we keep a ratio of 2 parts vertical to 5 parts horizontal. This ratio creates a rectangular chart area with enough vertical space for data and labels, and enough horizontal space to display the sequential order.

Visual Data Plotting

Our specialists accurately plot data points with clear symbols, connecting them to depict the story over time. 

This visual representation allows for easy identification of trends, shifts, or patterns.

Titling and labeling the graphs

To confirm clarity, we title and label the run chart accurately, with the horizontal axis (x-axis) representing time and the vertical axis (y-axis) providing details on the specific measure being tracked. 

If the x-axis units are obvious, we don't include an extra label, ensuring that the chart is clear and easy to understand.

Median calculation 

The median is the middle value when data is sorted from highest to lowest. If you have an even number of data points, the median is the average of the two middle values.

Our team calculates the median to serve as a centerline, providing a benchmark against which changes can be measured.

However, a median line may not be necessary for run charts with very few data points or those displaying multiple data series, as it can add unnecessary complexity.

How we Determine the Median for Your Run Chart

  • we plot your data in chronological order
  • then reorder the data values from highest to lowest
  • and find the middle value directly. This is your median.
Method 1: Run chart data reordered and median determined
Method 1: Run chart data reordered and median determined
Run chart with labels and median
Run chart with labels and median

Enriching Run Charts for Deeper Insights

Besides these steps, we enhance run charts by adding a goal line and annotations for significant changes or interventions.

This added layer of context adds visual reference for progress and turns raw data into actionable insights for healthcare providers.

Run chart data with goal line and tests of change annotated
Run chart data with goal line and tests of change annotated

The potential of run charts in healthcare data visualization is just the beginning. Belitsoft's expertise in cloud analytics modernization is showcased with our healthtech client on AWS, highlighting our ability to tackle challenges in scalability, security, and data customization. Discover how our end-to-end analytics services can bring similar innovations to your healthcare data strategies.

Small Multiples

Small multiples are a visualization technique that arranges similar charts in a grid layout. Each chart uses the same scales and axes, making it easy to compare different data subsets.

This is especially useful for run chart applications, as it allows our clients to monitor the same metrics across different segments or locations, or providers.

Our use of small multiples makes it easy to compare trends and patterns across different groups in healthcare.

By keeping scales and axes consistent, we can visually highlight whether changes are widespread or limited to specific areas.

Despite the focus on individual locations or providers, the overall results can still be understood in one view.

Run chart used as small multiples
Run chart used as small multiples

Chart with Multiple Measures

Run charts can track multiple related measures over time. In this case, all measures share the same vertical axis (percentage) and horizontal axis (time).

Plotting multiple measures on the same chart allows for direct, side-by-side comparison of their trends, revealing the evolution of each measure and their connections to one another.

Run chart displaying multiple measures
Run chart displaying multiple measures

This chart focuses on three measures of diabetes care: foot exams, eye exams, and self-management goal setting.

The chart shows that both foot and eye exam percentages have sustained improvement. However, self-management goal setting initially improved but then plateaued at around 35%.

Dual-axis Run Chart

When dealing with measures that have different scales, such as clinic wait times and visit volumes, we use dual-axis run charts,since using a single vertical axis can make visualization difficult. A dual-axis run chart solves this problem by having separate, appropriately scaled vertical axes on each side of the chart.

Run chart displaying multiple measures for each axis
Run chart displaying multiple measures for each axis

This type of chart enable us to track median clinic wait time in minutes and clinic workload in the number of visits. The chart reveals a significant decrease in wait time. 

However, this decrease is not due to a drop-in clinic visits, as the workload remained stable. 

This suggests that other factors, such as improved processes or staffing, are contributing to the decrease in wait time. Our analysis aims to uncover the underlying factors that contribute to changes in performance, providing a deeper understanding of the data.

Displaying multiple related statistics on a single run chart

Averages can sometimes mislead us because they are affected by extreme values (outliers) and may mask the actual performance of most of your data points.

To address this issue, our approach involves using a run chart to track multiple statistics related to the same healthcare measure, revealing a more nuanced and accurate understanding of changes over time.

Run chart displaying multiple statistics for the same measure
Run chart displaying multiple statistics for the same measure

For instance, let's consider our efforts in monitoring glycated hemoglobin (HbA1c) levels, a crucial indicator for managing diabetes. 

Instead of just monitoring the average HbA1c values across all patients, we also track the percentage of patients who achieve HbA1c targets below 7.

Both these statistics move together. Just looking at the average HbA1c alone could be misleading because it might be influenced by a few outlier patients who made significant improvements.

This dual-statistic method allows us to have a comprehensive view of the situation. While the average HbA1c value gives us a broad overview, analyzing the percentage of patients who meet specific targets gives us direct insight into the effectiveness of diabetes management strategies.

Median of the 'Before' Data and Median of the 'After' Data

At Belitsoft, we place great importance on the nuances within healthcare data, especially in projects where data points are limited. Fluctuations in such datasets can distort the assessment of process performance, much like short-term stock price movements may not accurately reflect a company's overall well-being.

In these cases, our experts rely on median analysis to provide a more precise evaluation. Medians are less influenced by extremes or outliers that can distort the analysis of a small sample. 

By comparing the median before and after a change, we get a broader sense of whether there has been a shift in the overall performance of the process, rather than just focusing on a few individual data points.

Run chart with little data
Run chart with little data

How to Interpret Run Charts

Visual analysis of run charts is powerful but subjective. What may appear as improvement to one person may not seem significant to another.

Run chart rules offer a standard, statistical way to identify meaningful patterns in data that may not be immediately obvious. These rules are especially helpful when there is not enough data to create a more sophisticated Shewhart control chart.

To detect actual changes in processes, we use four important rules for identifying "nonrandom signals of change." These rules are based on statistical principles and look for patterns such as shifts, trends that are unlikely to occur due to pure chance, suggesting that an actual change in the process has likely happened.

  1. Rule One - A shift 
  2. Rule Two - A trend 
  3. Rule Three - Too many or too few runs
  4. Rule Four - An astronomical data point
Four rules for identifying nonrandom signals of change
Four rules for identifying nonrandom signals of change

Rule One—Shift

A "shift" in your run chart data is a sustained period where your data points consistently deviate from the median.

  • We draw the median line on the run chart.
  • Our experts look for sequences of six or more consecutive data points either ALL above or ALL below the median, disregarding data points that precisely align with the median.
  • Then they exclude any data points that fall exactly on the median line. 

For example, if the run chart tracks patient wait times, seeing 6 or more points in a row below the median suggests consistently shorter wait times. This suggests the changes had a positive impact.

Note that Rule One requires a minimum of ten data points to be practical.

Rule Two—Trend

The Trend Rule identifies a sequence of five or more consecutively increasing or decreasing data points. This signals a gradual and a sustained change in your process.

The threshold of five points helps teams avoid unnecessary reactions and wasted time investigating false alarms. It also prevents overreacting to short-term random fluctuations that may appear as a trend with only 2 or 3 points.

If there are consecutive points with the same value, they should be ignored as they do not contribute to the upward or downward trend.

Simulations have shown that the Trend Rule is effective at detecting changes with five points, and increasing it further to six or seven does not significantly improve detection.

Rule Three—Runs

We utilize Rule Three to assess the stability of your processes. An unusually high or low number of runs may indicate hidden complexities or instabilities within your process.

Sometimes, data points can bounce around a lot, causing frequent ups and downs.

These fluctuations may have underlying causes you haven't thought about. Rule Three serves as a warning sign for potential hidden complexities in your data that trends and shifts alone might miss.

It helps identify data instability by looking for "too few" or "too many" runs.

If the data points stay clustered on one side of the median for a long time, it creates very few runs. This suggests that the process is not fluctuating as much as it should be. It's possible that our measurement isn't sensitive enough, making everything appear the same.

Too many runs is also a problem. Imagine your data line rapidly zig-zags above and below the median, creating lots of runs. It indicates an unusual level of fluctuation. Something in your process might cause wild or inconsistent results. In such cases, we make a further investigation to find the root cause.

Combining data from different sources without separating them, like in the day/night shift example, can create a false sense of hyperactivity in the run chart.

To make Rule Three (number of runs) practical, we need at least ten data points.

Run chart evaluating number of runs
Run chart evaluating a number of runs
Measure with too few runs
Measure with too few runs
Run chart with too many runs
Run chart with too many runs

Rule Four—Astronomical Point 

An astronomical point is not evidence of a change, but it highlights the need to investigate its cause.

It stands out from other data points like a bright star in the night sky. It is significantly different and easily noticeable on a chart.

Every dataset has its highest and lowest points, but not all of them are considered astronomical. An astronomical point is exceptionally abnormal for your specific data.

Rule Four, unlike other rules, does not rely on statistical probabilities. It requires judgment and agreement from those familiar with the tracked process.

Astronomical points can signify something unusual happened – a sudden spike or drop.  

Regardless of whether these points suggest a sudden shift caused by new processes, equipment failures, or data entry errors, we investigate to uncover and comprehend their underlying causes.

By paying meticulous attention to detail, we ensure our clients receive a thorough analysis that considers all aspects of their data, ultimately offering a clear direction for.

Special Considerations Using Run Charts


Run chart rules like Shift (Rule One) and Too Few/Too Many Runs (Rule Three) rely on the median acting as a balance point, with data roughly evenly distributed above and below. 

Our team monitors for data distribution issues, if too much data falls directly on the median or clusters at the extreme edges (like always at 0% or 100%), this balance is broken. The statistical basis of the rules is no longer valid. We can't reliably detect shifts or unusual patterns of data fluctuation.

Two cases when median ineffective on run chart
Two cases when median ineffective on run chart

Trend Lines

If the run chart shows only normal variation (no signals according to the four rules), a trend line can make it look like a meaningful change is happening when it's not.

Recognizing the potential for misinterpretation, we only incorporate trend lines when a clear, statistically significant change (shift, trend, too few/too many runs) is identified through the application of the four rules.

This cautious strategy prevents the overinterpretation of normal variations, focusing instead on genuine signals of change.

Run chart with inappropriate use of trend line
Run chart with inappropriate use of trend line


We tackle the challenge of autocorrelation head-on, particularly prevalent in healthcare data at regular intervals (e.g., monthly). This means that data points from consecutive months are related or similar because patient information can be carried over from one month to the next if new measurements aren't taken.

In chronic conditions like diabetes, patients may not need new health assessments every month. This leads to the reuse of previous data, creating artificial similarity between consecutive months.

Autocorrelation makes it difficult to identify genuine trends or changes in patients' health status because the data isn't truly independent. It can skew statistical analysis.

Imagine a clinic measuring blood glucose levels in patients with diabetes. A patient with well-controlled levels might have appointments every three months. If the clinic's data summary uses their last measurement for the months in between, the monthly data will appear artificially stable, potentially masking subtle changes in the patient's condition.

To avoid the skewing effect of autocorrelation, we analyze data only for patients who actually had appointments and measurements within that month.

We help healthcare organizations analyze data and make strategic decisions with our Business Intelligence services. Our BI solutions offer advanced analytics, customizable dashboards, and detailed reporting features. With these tools, you can convert complex healthcare data into a valuable strategic asset. Contact us for a personalized consultation.

Never miss a post! Share it!

Written by
Head of BI Department
With over 15 years in data analysis, I can make your data work for your business success with customized BI, AI, and ML solutions.
1 review

Rate this article

Recommended posts

Belitsoft Blog for Entrepreneurs


Cloud Analytics Modernization on AWS for a Healthtech Company
Cloud Analytics Modernization on AWS for a Healthtech Company
Belitsoft designed a cloud-native web application for our client, a US healthcare solutions provider, using AWS. Previously, the company relied solely on desktop-based and on-premise software for its internal operations. To address the challenge of real-time automated scaling, we embraced a serverless architecture, using AWS Lambda.

Our Clients' Feedback

Let's Talk Business
Do you have a software development project to implement? We have people to work on it. We will be glad to answer all your questions as well as estimate any project of yours. Use the form below to describe the project and we will get in touch with you within 1 business day.
Contact form
We will process your personal data as described in the privacy notice
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply
Call us

USA +1 (917) 410-57-57

UK +44 (20) 3318-18-53

Email us

[email protected]

to top