Validation and treatment of missing values for the reduced WOMAC function scale

Abstract

Introduction: The reduced WOMAC function scale has been developed and initial validity performed. However, further validation and recommendations for the treatment of missing values is required.

Aim: To further assess the validity of the reduced function scale of the WOMAC and recommend a protocol for the treatment of missing values.

Method: Further validation of the reduced scale was performed via a cross-over study of 100 pre-operative total joint replacement patients, each being randomised to receive either the full or reduced scale along with the pain scale, and then the alternate version upon admission. Data utilised in the development of the reduced scale was used to develop a missing value protocol, where the number of valid responses for several protocols was examined, as well as comparison of the means and standard deviations.

Results: Of the consenting 100 patients, 66 continued onto admission. The median time between administrations of the questionnaires was 14 days (range 0 to 72 days). There was no significant difference between pain scores for each questionnaire using the paired t-test (p=0.56). Similarly, there was no significant difference between the full and reduced function scales (p=0.65). The standard protocol for the full scale is that if there are 4 or more missing items, the patients’ response is invalid. But when there are 1-3 items missing, the average value for the sub-scale is substituted in lieu of these missing values. Examining the frequencies of valid responses, means and standard deviations when using different missing value protocols (none missing, 0 or 1, up to 2 and up to 3 missing), indicated that there was no substantial benefit between the ‘up to 2 missing’ and ‘up to 3 missing’ response protocols. However, for this small gain, the supposition that the completed items are representative of the missing ones rises from 29% (2 of 7 items) to 43% (3 of 7 items) should be considered unacceptable.

Conclusions: The reduced WOMAC function scale has been further validated. It is proposed that where 3 or more responses are missing, the patients’ response is regarded as invalid. Where there are 1 or 2 items missing, the average value for the sub-scale is substituted in lieu of these missing values