Reliable Change Index

Test scores will very from one administration to another due to imperfect reliability. In order to test if the change in scores is greater than might would be expected from random variation alone, many researcher's advocate the use of the Reliable Change Index Statistic. If the change score exceeds the Reliable Change Index (which is calculated based on the reliability of the measure), then the patient can be said to be "significantly improved", meaning that the observed change would be expected by chance alone at a probability of less than 5%.

The formula for the Reliable Change Index is:

square root (2 X (standard error of measurement)²).

I have attached an Excel utility to calculate the Reliable Change Index and clinical cutoff score for any outcome measure.

• ACORN_RCI_and_clinical_cutoff_calculator.xls

The largest problem with a simplistic use of the RCI statistic is that it does not take into account regression artifacts: patients reporting more severe symptoms are much more likely to show significant change on the outcome questionnaires that those reporting mild symptoms. See RegressionToMean

For more information, see Dr. Lee Becker's discussion of reliable change.

Primary topics:

• Measure What? Why?
• Global Distress Factor
• Absenteeism & Presenteeism
• Items or Questionnaires?
• Outcomes Measurement 2.0
• Reliability and Validity
• Measurement Error
• Measuring Change
• Regression to the Mean
• Clinical Cutoff Score
• Reliable Change Index
• Effect Size
• Meta-analysis
• Case Mix Adjustment
• Severity Adjusted Effect Size
• Outcomes Benchmarking
• Trajectory of Change
• Random Walk Hypothesis
• Rational Resource Allocation
• Therapist Effects
• Value Index

You may also view all topics in alphabetical order in WebTopicList.

-- JebBrown - 15 April 2007