# Time Series Charts

A time series chart is a line graph which plots the indicator of interest on the Y (vertical) axis and the time interval over which the data are displayed on the X (horizontal) axis, using any interval of time (e.g., minute, hourly, daily, weekly, monthly, quarterly, yearly, etc.). Common types of indicators plotted on the Y axis are percentages (e.g., percent of patients receiving care according to standards), rates (e.g., case fatality rate), time (e.g., waiting time), quantities (e.g., stock levels), or numbers (e.g., weight).

Time series charts help us understand if the changes we are making are leading to a change in improving the quality of care from some initial level to a consistently sustained higher level. They are a simple yet effective tool to track the performance of a process over time and document the story of improvement work. They make trends or other non-random variation in the process easier to see and understand. With the understanding of patterns and trends of the past, groups can then use time series charts to help predict future performance.

Guidance developed by the USAID Health Care Improvement Project explains how to create and label and how to analyze and interpret a time series chart.

**When to Use a Time Series Chart**

Time series charts portray indicator data over time. While most graphs are like a photo that captures a point of time, the time series chart is like video rolling over time. The ongoing monitoring of an indicator through a time series chart is particularly valuable in quality improvement as it allows us to track when specific changes were introduced, see their impact on a process, and tell whether improvement is sustained over time. The more data points that can be plotted, the better the understanding that can be gained about the process over time.

An important element in construction of a time series chart is to calculate the *median* value of the data. The median represents the middle value in a set of data. Creating a horizontal line through the median of a data set allows you to detect shifts or changes in the tendency of the indicator on a time series chart. A minimum of 10 data points are needed to calculate and plot the median of a set of data.

**How to Use a Time Series Chart**

Time series charts are often referred to as “run charts” because the term *run *means a consecutive series of points running either above or below the median or center line of the data set. The points in a run chart mark the single events (how much occurred at a certain point in time). A run is broken once it crosses the center line. Values on the center line are ignored: they do not break the run, nor are they counted as points in the run.

Here is an example from Russia.

The basic steps in creating a time series chart follow.

**Step 1. **Collect at least 25 data points (number, time, cost), recording when each measurement was taken. Arrange the data in chronological order.

**Step 2. **Determine the scale for the vertical axis as 1.5 times the range. Label the axis with the scale and unit of measure.

**Step 3. **The horizontal axis marks the measure of time (minute, hour, day, shift, week, month, year, etc.). Make sure the axis is labeled as such.

**Step 4. **Plot the points and connect them with a straight line between each point. Be sure to include a center line for the median value of all the data points.

The following rules guide the interpretation of a time series chart:

- Five consecutive increasing (or decreasing) points suggest a trend
- Six consecutive points above (or below) the median line suggest a statistically significant shift in the process
- Fourteen successive points alternating up and down suggest a cyclical process

**Points to Remember**

Be careful not to use too many annotations on a time series chart. Keep it as simple as possible and include only the information necessary to interpret the chart.

Do not draw conclusions that are not justified by the data. Certain trends and interpretations may require more statistical testing to determine if they are significant.

When interpreting data points, the variation in the denominator can influence patterns and conclusions. Two common sources of variation in denominator size are: 1) changes in service utilization by clients, and 2) for pooled data, variations in the number of sites contributing (reporting) data.

To ensure that the time series chart does not mislead, take care to present scales in regular intervals.