Variability of SF-36 scores within gose categories

Abstract

Measuring the outcomes of patients following head injury is important for research studies1, clinical audit2 and assessing rehabilitation requirements in individual patients. The tools that are used to quantify such outcome must be appropriate for the intended purpose. The Glasgow Outcome Score (GOS)3 and, more recently, the extended Glasgow Outcome Score (GOSE)4 have been used in clinical trials and audit studies to define outcomes in patient groups. However, these scores are sometimes used to characterise outcomes in individual patients. This is inappropriate, since functional outcomes in individual patients may be poorly described by these scoring tools. We have sought to address this issue by comparing GOS and GOSE against the Short Form-36 health survey (SF-36)5, a patient-reported score of health-related quality of life in a large cohort of patients surviving head injury. Outcome measurements were collected on 227 patients with head injury admitted to the regional neurosurgical intensive care unit. Participants were aged between 16 and 86 (median 31) with median (range) post resuscitation GCS of 7 (3–15). Twenty-two patients had mild, 53 moderate, and 152 severe head injury. Data obtained included the GOS, GOSE and the UK version of the SF-36. Forms were sent by post approximately 6 months post injury, and were completed by the patients (with the help of a family member or partner, if required). SF-36 scores in different measurement domains were plotted against the GOSE and relationships assessed using non-parametric statistical methods. We found a highly significant correlation between the GOSE and scores in individual SF-36 domains (p < 0.0001 for all comparisons), with Spearman Rank rho values between 0.47 and 0.68. However, we found large variations in the spread of SF-36 values within each GOSE category, with the largest interquartile ranges observed in the domains of Emotional Role and Social Function (Fig 1). The GOS and GOSE are extremely useful measures of outcome in groups of patients following head injury, and have proved valuable in the conduct of clinical trials and audit of clinical outcomes as a means of quality assurance. However, our data underline the pitfalls of using these scoring systems as descriptors of functional outcome and quality of life in individual subjects. A better understanding of the causes of intra-category variability in SF-36 scores might provide one means of devising new outcome scoring systems that improve on the GOSE. The failure of experimentally successful neuroprotective interventions in clinical trials has, in some part, been attributed to the lack of sensitivity of outcome assessment.6 Such improved scoring systems may go some way towards addressing this issue.