Back to presentation overview

Title:Validation of a new low back pain sub-grouping tool for primary care (the STarT Back tool).

Authors and affiliation: Jonathan C Hill, Kate M Dunn, Chris J Main, Elaine M Hay

Primary Care Sciences Research Centre, Keele University, Staffordshire, UK

Email address: j.hill@cphc.keele.ac.uk

Introduction: Despite the popularity of the “flag” approach for managing low back pain, systematic reviews have shown that effect sizes of interventions aimed at tackling unhelpful cognitions are modest, and benefits are relatively short lived. One reason for this might be an over-general approach, which fails to target appropriate sub-groups. We have therefore developed a new sub-grouping tool to classify patients with low back pain into appropriate groups for targeted treatment.

Objective: To test the STarT Back sub-grouping tool for validity and test-retest reliability.

Methods: Construct validity was tested in a cross-sectional survey of 244 consecutive primary care LBP consulters. Questionnaires containing the STarT Back tool and corresponding complete scales for each prognostic construct measured by the tool (including the Roland-Morris Disability Questionnaire (RMDQ), the Pain Catastrophising Scale (PCS), the Tampa Scale of Kinesiophobia (TSK) and the PHQ-2 (a depression screen) were mailed to participants. Construct validity was tested by comparing responses to individual items in the tool with their respective complete scales. Discriminant validity was assessed by comparing pain intensity, bothersomeness, disability, work loss and psychological distress in participants classified by the tool as “high”, “medium” and “low” risk of chronicity. Test-retest reliability of sub-group allocation was examined with on a sub-sample of 53 early responders sent repeat questionnaires (two-weeks after baseline). Predictive validity will be assessed within a prospective cohort of 500 primary care LBP consulters currently being followed-up. Outcome is defined as disability at six months, dichotomised above and below median values measured by the RMDQ. The ability of the tool to predict change in pain intensity, work loss and psychological distress will also be reported.

Results: 131/244 (54%) responded to the cross-sectional survey with 40% allocated to low, 35% to medium and 25% to high risk groups. Construct validity was demonstrated for all individual items: e.g. the catastrophising item identified people above the median of the PCS with 75% sensitivity (84% specificity). Similarly, a disability item identified people above the median on the RMDQ with 89% sensitivity (61% specificity). Sub-group discrimination was high: over 91% of ‘high risk’ patients had scores above the median on the RMDQ, TSK, PCS and PHQ-2. Test-retest reliability for sub-group allocation was moderate (kappa = 0.54). Follow-up of the prospective cohort will be complete in early 2006, and data on the predictive validity of the sub-grouping tool will be presented at the meeting.

Conclusion: The sub-grouping tool has been shown to allocate primary care LBP patients to groups with high levels of validity and moderate reliability. The usefulness of the sub-grouping tool to allocate patients to specifically targeted treatment groups will be tested in a future RCT.