Construct validity and reliability of the Rivermead Post-Concussion Symptoms Questionnaire

Abstract

Objectives: To provide further evidence of reliability and internal and external construct validity of the Rivermead Post-Concussion Symptoms Questionnaire (RPQ), which measures severity of postconcussion symptoms following head injury. Design and setting: A cross-sectional study of consecutive patients presenting with a head injury in two urban teaching hospitals and a community trust. Patients: Three hundred and sixty-nine patients returned a questionnaire from 1689 consecutive adult patients (18 years and above) referred to radiology for a skull X-ray following a head injury, and those who were currently under the care of a community-based multidisciplinary head injury team. Method: Internal construct validity tested by fit to the Rasch Measurement model; external construct validity tested by correlations with Rivermead Head Injury Follow-up Questionnaire (RHFUQ); test-retest reliability tested by correlations at two-week intervals. Outcome measures: Rivermead Post-Concussion Symptoms Questionnaire and Rivermead Head Injury Follow-up Questionnaire. Main results: RPQ scores ranged from 0 to 64 (17.3% floor, 0.3% ceiling). Overall fit to the Rasch model was poor (item fit mean-0.416, SD=1.989, chi-squared=172.486, p < 0.01) suggesting a lack of unidimensionality. The items headaches, dizziness and forgetful displayed misfitting residuals and the first two items also displayed significant item trait fit statistics (p < 0.0006). After removing the items headaches, dizziness and subsequently nausea the RPQ demonstrated good fit at overall and individual item levels, both for the remaining 13 items (RPQ-13) and the three items (RPQ-3) which now formed a subsidiary scale. All items functioned consistently across age and gender. The RPQ-13 and RPQ-3 scales showed test-retest reliability coefficients of 0.89 and 0.72 (both p-values < 0.01) and positive correlations with RHFUQ scores (0.83 for RPQ-13, 0.62 for RPQ-3, both p-values< 0.01). Conclusions: As currently used, the RPQ does not meet modern psychometric standards. Its 16 items do not tap into the same underlying construct and should not be summated in a single score. When the RPQ is split into two separate scales, the RPQ-13 and the RPQ-3, each set of items forms a unidimensional construct for people with head injury at three months post injury. These scales show good test-retest reliability and adequate external construct validity.