Almost everyone will experience low back pain (LBP). Most of us also know someone who has persistent LBP – pain that comes and goes, or never goes; that limits work, or life or enjoyment. When our own back hurts we worry (a little or a lot) that we might end up like them.
The majority of adults who suffer a new episode of LBP recover within a few weeks or months. For the smaller proportion however, ongoing pain and disability can generate great personal and financial costs – such that the disability burden of LBP is higher than that for any other health condition world-wide.
The disability burden of LBP is higher than that for any other health condition world-wide.
There is widespread interest in the benefits of identifying early those who are at risk of a poor outcome after an episode of LBP. The driving principle is that, rather than ‘waste’ healthcare resources on treating those who will recover quickly (and avoid potential harms of over-treatment), screening questionnaires can be used to allocate care based on the prognostic information provided.
Sensibly, the concept of timely delivery of more complex management to those who need it most offers promise; and practically, there is evidence of benefits.
Appropriate care allocation relies on the ability of screening instruments to accurately differentiate between those who will recover well and those who will not. These self-report questionnaires generally assess characteristics of an individual’s pain experience (including pain intensity and functional impairment) and psychosocial factors known to have prognostic relevance (e.g. beliefs, catastrophisation, anxiety and depression). Recent guidelines (e.g. NICE) recommend their use to guide LPB management in primary care.
The uncertainty surrounding the prognostic performance of available screening instruments instigated this systematic review, published today in BMC Medicine.
Reviewing screening instruments
We asked: when adults consult a healthcare professional within the first 3 months of an episode of LBP, how accurately can screening instruments classify those who will have a poor outcome, a minimum of 12 weeks later? We defined ‘poor outcome’ uniformly (for pain and disability outcomes) to allow comparisons between different tools and pooling of data from multiple studies.
Thorough database searching identified 18 studies investigating 7 screening instruments. The instruments had either been developed for the purpose of providing prognostic information for LBP patients, or are being used for this purpose. Five studies investigated the STarT Back Tool (SBT), 7 studies investigated the Orebro Musculoskeletal Pain Screening Questionnaire (OMPSQ), 2 studies investigated the Vermont Disability Prediction Questionnaire and 4 further instruments were investigated in single studies only.
We were able to pool data for the SBT studies (for both pain and disability outcomes), and for the OMPSQ (for pain, disability and absenteeism outcomes). The SBT seems to differentiate patients with poor disability outcomes from those who recover, with an accuracy of around 74% – which for prognostic models is regarded as ‘acceptable’. The SBT was considerably less informative at predicting pain outcomes, differentiating between those who did and did not develop persistent pain with an accuracy little better than chance (59%).
The performance of the OMPSQ was comparable to the SBT for disability outcomes (75% accuracy), but superior to the SBT for differentiating pain outcomes (classifying patients accurately approximately 69% of the time). Notably, the pooled performance of the OMPSQ for predicting prolonged absenteeism at 6 months was regarded as “excellent” (83% accuracy).
That finding suggests that screening instruments are (consistently) poor at differentiating who will and will not develop persistent pain (at least in the way we’ve defined it).
When the studies investigating individual instruments were plotted alongside the pooled performances of the SBT and OMPSQ (with separate plots for pain and disability outcomes) – the confidence intervals of the individual study results overlapped with the pooled results. That finding suggests that screening instruments are (consistently) poor at differentiating who will and will not develop persistent pain (at least in the way we’ve defined it).
The tools are a bit better at estimating the risk of persisting disability (again, according to our definition of poor outcome). And perhaps prognostic screening can inform us most accurately about work-related outcomes (using the OMPSQ).
Indeed, there are limitations to the conclusions that can be drawn from our analyses, which have been discussed in the manuscript. Insightful comments by expert reviewers have raised some important issues which we have been careful to clarify.
As an investigator and author of this review I am sure to have a bias towards overstating the significance of this study’s results. Nonetheless, in light of this study’s findings: Do I think it’s time to debunk the use of currently available screening instruments for adults with recent onset LBP? No, probably not. Do I think that clinicians who adopt a screening instrument for use in clinical practice should be cautious about accepting its prognostic accuracy? Most definitely.