VALIDATING GENERIC HEALTH MEASURES IN SPINAL DISEASE

Abstract

Objective: Measuring outcomes from chronic disease in terms of generic, health-related quality of life (HRQoL) instruments is of increasing importance to allow valid comparison of interventions and to accurately assess efficacy of treatment from the patient’s perspective. In this context we sought to establish the role of the generic SF-36 health survey in measuring outcomes from spinal surgery.

Method: A prospective observational study of patients undergoing elective cervical discectomy, lumbar discectomy, and lumbar laminectomy using both disease specific (Myelopathy Disability Index [MDI], Roland Morris Disability Scale [RMDS], Visual Analogue Scales [VAS], Hospital Anxiety and Depression Scales [HADS]) and SF-36 assessment pre-operatively and at 3 months and 12–24 months following surgery. The generic instrument was tested for the components of construct validity in comparison to the established specific measures. Analysis was performed with non-parametric statistics within SPSS.

Results: Six-hundred and twenty patients were followed between 1998 and 2005 (median age 53 years; 203 lumbar discectomy, 177 lumbar laminectomy, 240 cervical discectomy). The principal SF-36 physical domains (Physical Functioning, Bodily Pain) strongly correlated with disease specific scores in all patients (Spearman’s ρ=0.5–0.74, p< 0.001) and similarly SF-36 mental domains correlated with the HADS subscales (ρ=0.30–0.45, p< 0.001) indicating concurrent/convergent validity. Discriminant validity was confirmed by the absence of significant correlation between SF-36 physical domains and the HADS (ρ=0.014–0.14, p> 0.05). In the lumbar laminectomy and cervical discectomy patients disease-specific physical scores prior to surgery strongly predicted early and late outcome (area under the receiver-operating characteristics curve [AUC] = 0.79–0.86, p< 0.001) and the same pattern was mirrored in the SF-36 physical domains (AUC = 0.76–0.78, p< 0.001) demonstrating the predictive validity of the generic measure. Physical Function and Bodily Pain SF-36 domains both had excellent response to change by Cohen’s criteria with effect sizes (standardised mean difference) of 0.86–1.57.

Conclusion: The SF-36 has been shown to possess the necessary features of construct validity in relation lumbar and cervical surgery to be considered as a suitable adjunct or alternative to measuring outcome with disease specific scores. As a widely employed HRQoL instrument the SF-36 should be a convenient means of assessing patients with spinal morbidity in all healthcare settings and the generic measure will permit easier comparison of the clinical and economic efficacy of different interventions.