Modifying the STarT Back tool for use with patients with other musculoskeletal conditions: does it work?

Talk Code: 

The problem

The Keele STarT Back Screening Tool (SBST) has predictive validity for stratifying primary care patients with non-specific low back pain (Hill et al, 2008). We tested whether a modified SBST could be used with a broader group of musculoskeletal pain patients.

The approach

A generic, modified SBST was embedded within prospective data collection of patients in 1) physiotherapy services (PhysioDirect trial n=2247) and 2) musculoskeletal primary/secondary care interface services, (SAMBA n=1039). The tool's ability to predict 6-month physical function (SF-36 upper tertile Physical-Component-Score) for patients with back, neck, upper limb, lower limb, or widespread pain was examined using Area-Under-the-Curve (AUCs) and sensitivity/specificity.


In PhysioDirect, AUCs for back, neck, upper limb, lower limb and multi-site pain were 0.72, 0.82, 0.79, 0.77 and 0.81 respectively. However, different subgroup cut-offs were required depending on site of pain to optimize sensitivity and specificity sufficiently. These findings were replicated in the SAMBA dataset.


A modified SBST moderately predicts outcome in other musculoskeletal sites within physiotherapy and interface service settings. However, the use of consistent subgroup cut-offs was not possible and resulted in poor sensitivity (too few patients ‘at risk' were identified) or specificity (too many were inaccurately classified ‘at risk') for some pain sites. The draft tool is now being amended and validated within a new programme of research for this broader patient population in a primary care general practice sample.


  • Jonathan Hill
  • Kate Dunn
  • Ed Roddy
  • Danielle van der Windt
  • Nadine Foster