Kaplan-Meier analysis allows a description of the behaviour of a binary variable over time. The Kaplan-Meier method determines the number of patients included in one modality at the beginning of the study passing to the other modality over time. The resulting chart displays step-shaped curves.

In practice

Kaplan-Meier curves are often used to describe the number of death’s evolution over time in medical research, but it can be used for all statistics that could be divided in two categories of interest and have measures over time. For example, the variable “pain level below 3” taking the modalities “yes” or “no”, measured each day of the week, is a dichotomous variable that can be described with a Kaplan-Meier curve over time.

To test the difference between groups based on the Kaplan-Meier curves a Log-rank test is usually performed.

How to use it on EasyMedStat?

Which data is required?

In order to perform a survival analysis (Kaplan-Meier), you need 3 dates for each of your patients:

  1. Date of inclusion: the date at which the analysis will start
  2. Date of event: the analyzed event
  3. Date of last follow-up: the date at which the patient is not followed anymore

If you have delays instead of dates, you will not be able to perform survival analysis. However, you could generate dates with Excel or your usual spreadsheet from these delays.

Perform survival analysis

  1. Go to Statistics > Descriptive
  2. Look for the Event variable you want to analyze and click on " Kaplan-Meier" to see the Kaplan-Meier curve within the whole sample
  3. Kaplan-Meier curves are generated as well as tables with confidence intervals for each delay

Compare survivals between groups

  1. Go to the panel "Compare survival by" and choose the variable you want to compare your groups by to see the Kaplan-Meier curves by group.
    For example, if you want to compare the survival of Men and Women, choose the variable "Gender" in the List

You can only compare survival by a discrete variable, ie, a List or Yes-no variable. It is not possible to use a Numeric variable or an Event variable to compare two survivals. You can learn more about survival comparisons on the Log-rank test page.

See also