Is my patient actually getting better? Application of the McNemar test for demonstrating the change at a single subject level

Abstract

Purpose: The aim of the current work is to provide a novel method for demonstrating the modification of a single patient’s performance on questionnaires and scales. The minimal detectable change (MDC), a statistics indicating the minimal change in measure not attributable to random variation, is commonly used in rehabilitation for this purpose. However, the MDC has some important drawbacks (e.g. it cannot be calculated on scores from ordinal tests and it can be only used for full questionnaire).Method: Review of the MDC and its limitations and application of the McNemar test on simulated data from single subjects.Results: We propose to use the McNemar test to check if the proportion of test items affirmed by a patient after rehabilitation is significantly different from the same proportion before rehabilitation. A significant McNemar test would indicate a non-random modification of the patient’s score and thus a true modification of his/her performance.Conclusions: The application of the McNemar test to questionnaires and scales offers a simple method for demonstrating the modification of a single patient’s performance. This use of the McNemar test overcomes the weaknesses of the MDC and gives support to the clinician in assisting him/her to convincingly communicate a non-negligible modification of the patient’s status.IMPLICATIONS FOR REHABILITATIONMeasuring the change in patients’ status is of paramount importance in medicine and rehabilitation. However, tracking the change in rehabilitation is difficult.For example, the minimal detectable change cannot be calculated on scores from ordinal questionnaires and tests, which are widely used as rehabilitative outcome measures.We propose here to use a McNemar test to check if the proportion of test items affirmed or passed by is significantly different between two conditions (e.g. before vs. after rehabilitation).Similar to the minimal detectable change, the significant McNemar test would indicate a non-random modification of the patient’s test score. In addition, the McNemar test can be calculated on ordinal data, thus overcoming some of the minimal detectable change weaknesses.