Converting SRS-24, SRS-23, and SRS-22 to SRS-22r Establishing Conversion Equations Using Regression Modeling

Study Design. Cross-sectional mail questionnaire. Objective. Assess the feasibility of translating total and domain scores from Scoliosis Research Society (SRS)-24, SRS-23, and SRS-22 to SRS-22r. Summary of Background Data. Three successive editions of the original SRS-24 health-related quality-of-life questionnaire have resulted from efforts to improve its psychometric properties and validate its use in patients down to 10 years of age. This resulted in the need to establish, if possible, conversion equations to the last and most thoroughly validated version, SRS-22r. Methods. A consolidated questionnaire of 49 questions that incorporated the various questions in the four questionnaires was mailed to a consecutive series of 235 patients who had received primary posterior or anterior instrumentation and arthrodesis to treat adolescent idiopathic scoliosis. Regression modeling was used to establish conversion equations from the SRS-24, SRS-23, and SRS-22 to the SRS-22r. Results. One hundred twenty-one of the 235 patients (51%), aged 23.3 +/- 4.52 years (range 14.2-34.6 years), returned the questionnaire at 8.6 +/- 4.00 years (range 2.3-15.9 years) following surgery. Estimation of SRS-22r questionnaire and nonmanagement domains total scores and mean scores from SRS-22 and SRS-23 scores is excellent (R(2) scores of 0.97-0.99) and good for SRS-24 scores (R(2) scores of 0.80-0.82, improving to 0.86 and 0.87 after minimal domain reconfiguration). Estimation of SRS-22r individual domain total scores and mean scores from SRS-22 and SRS-23 is good to excellent (R(2) scores of 0.81-0.99). Minimal domain reconfiguration improves conversion from SRS-24 pain from R(2) = 0.71 to 0.76, which are both fair; SRS-24 function from R(2) = 0.69 and 0.74 to 0.83, from poor and fair to good; and SRS-24 satisfaction/dissatisfaction with management from R(2) = 0.64 to 0.80, from poor to good. Conversion of SRS-24 self-image is poor (R(2) = 0.60) despite the correlation being statistically significant. Conclusion. With one exception, SRS-24, SRS-23, and SRS-22 questionnaire, nonmanagement domains, and individual domain total scores and mean scores can be translated to SRS-22r scores with fair to excellent accuracy, which is further improved in some instances by minimal domain reconfigurations. The sole exception is SRS-24 self-image, which translates poorly.