Subgrouping patients with sciatica in primary care for matched care pathways: development of a subgrouping algorithm

Sciatica (radicular pain) has significant impact on patients, healthcare and societal costs [1]. The annual impact on the United Kingdom (UK) economy is £268 million in direct medical costs, and £1.9 billion in indirect costs, based on a Dutch study which indicated that the societal cost of sciatica represents 13% of all back pain related costs (inflated from 1998 figures) [2]. Many patients with sciatica do well over time and are treated successfully with conservative interventions, with injections and/or surgery reserved for patients with persistent, disabling pain [3].

In the absence of suspected serious pathology or profound neurological deficits, current usual management of sciatica in the UK National Health Service (NHS) follows a ‘stepped’ care approach, with an initial period of ‘low level’ treatment for most patients, including advice, reassurance and analgesic medication, and for those not improving, onward referrals are initiated to services such as physiotherapy or specialist spinal services in intermediate and secondary care settings [4]. Up to 30% of patients with severe sciatic pain continue having significant symptoms at 1 year follow-up [5]. However, a direct referral to spinal specialist services for all sciatica patients at the point of first consultation in primary care is not a cost-effective care model [6], and it is unlikely that this is needed for most patients, except when sinister pathology is suspected. In the UK, most primary care patients with sciatica, who may eventually require a specialist opinion, are referred to NHS spinal specialist services at the interface (intermediate) level, between primary and secondary care services. Interface spinal services are predominately staffed by non-medically qualified clinicians (for example physiotherapists) with advanced training and expertise in spinal care and direct referral access to surgical and injection services [4]. The challenge for clinicians (such as general practitioners (GPs)) managing sciatica in primary care is how to identify, early in the presentation, patients who are likely to only need and do well with conservative management in primary care, and patients who may need early, fast-track referral to spinal specialist opinion.

For non-specific low back pain (LBP), research has shown that a model of stratified care that does not over-treat those with a good chance of improvement but identifies those likely to need more intensive treatment, has led to superior clinical and cost effectiveness compared to non-stratified care [7, 8]. This approach in LBP utilises a brief self-completed screening tool (the Keele STarT Back tool) [9] which captures eight modifiable physical and psychological prognostic indicators. The STarT Back tool has 9 items, 4 are physical constructs, with 2 questions capturing back pain/leg pain related disability, and 5 are psychological constructs. A score of 3 or less indicates the patient is at low risk of future persistent back pain-related disability, a score of 4 or more of the 5 psychological items indicates the patient is at high risk, any other score identifies patients as at medium risk of persistent disability [9]. By estimating future risk of persistent back pain-related disability, the STarT Back tool supports early clinical decision-making about conservative treatments (such as GP care and physiotherapy management) [7, 9]. Although neither the STarT Back trial [7] nor the subsequent implementation study (IMPaCT Back) [8] excluded patients with sciatica, the STarT Back tool and matched treatments were not specifically developed for sciatica. Sciatica patients have more severe pain and take longer to recover [1]. Currently there is no screening tool or algorithm available to support clinical decision-making in directing patients with sciatica to matched care pathways, when they first consult with sciatica symptoms which are not suspicious of sinister pathology. This is explained, at least in part, by the currently limited number of prognostic factors that have been shown to be independently associated with outcome in patients with sciatica [10‐12]. This paucity of evidence for prognostic factors, with only pain and disability severity being consistently associated with eventually having spinal surgery (which is taken as proxy of poor outcome for natural course and conservative management) [11, 13], impedes development of prognostic models that can guide treatment decision-making [11], yet it is not feasible or necessary to refer all patients with high levels of pain and disability to spinal specialist or surgical services. However, early referral to investigations and more intensive treatments may improve outcomes and quicker recovery for some sciatica patients consulting in primary care, instead of waiting the outcome of initial conservative management before referring on when there is no improvement.

The aim of this research was to develop an algorithm with which to support clinical decision-making about subgrouping patients with sciatica consulting in primary care for matched care pathways.

We have developed an algorithm for supporting clinical decisions regarding onward referral at the point of initial consultation, and allocation to matched care pathways, for patients consulting in primary care with sciatica. The algorithm combines information about patients’ risk of persistent back and leg pain related disability and four items from the clinical assessment. It allocates patients to one of three groups for matched care pathways: for group 1 this comprises of brief self-management support (up to 2 sessions), for group 2 a course of up to 6 sessions of physiotherapist-led care, and for group 3 a ‘fast-track’ referral to MRI and spinal specialist services.

Currently, prognostic research in sciatica patients shows that only one or two factors (pain, disability or impact of the condition/symptoms) are consistently associated with persistent disability or proceeding to surgery (when surgery is an appropriate and desirable management option) [11‐13, 25], but not all sciatica patients with severe leg pain will have a poor outcome and will not proceed to have surgery. This limits the development of useful prognostic models to inform stratified care and guide clinical decision-making about early referral to spinal specialists. In the absence of models purely based on probability of poor outcome for this condition, the next reasonable step is to consider investigating and combining information on prognosis with factors associated with referral to specialists. It is possible that referral to specialists for investigations and treatments early in the presentation of symptoms in primary care, may improve outcomes for some sciatica patients however, this will need to be investigated in clinical trials. To our knowledge, this is the first subgrouping algorithm developed to guide clinical management at first consultation in primary care for patients with sciatica.

It was important that the factors/characteristics used, would be acceptable to clinicians in terms of directing referrals to NHS spinal specialist services immediately after consultation in primary care, given the possible impact of early ‘fast-tracking’ of patients on the healthcare system. In choosing the ‘optimal’ combination of characteristics and cut-off points, we considered the trade-offs between sensitivity (correctly identifying patients who need a referral to spinal specialists) and feasibility, as it would be impractical to refer the majority of patients. And the input of our research and clinical advisory group was unanimous and fundamental in these decisions, indicating high clinical face validity. We acknowledge as a limitation that the clinical advisory group did not include patients. The model’s predictive performance is moderate (C-statistic: 0.70). The positive predictive value (PPV) of the algorithm for specialist referral is 22%, with sensitivity of 51%. It is clear that based on the proposed algorithm, a number of patients will be referred early for tests (MRI) and spinal specialist assessment, who may not have needed such referral. It is fully acknowledged that these values are low, however, in the absence of a reference standard or evidence-based guidance to underpin referral decisions at first consultation (in the absence of suspected sinister pathology), we consider this as a first step towards stratified care for this population. In the UK, these patients are referred to primary/secondary care interface (intermediate) clinics where they are further evaluated, appropriate management options are discussed with the patient, and referrals from that intermediate setting are streamlined to surgical, injections or pain management settings, in secondary care.

Further potential limitations relate to the factors, and their cut off points, we included in the algorithm development. We chose the list of factors based on published data and clinical opinion, and that were available in the ATLAS cohort study dataset. We included all factors that could, in principle, be associated with referral to spinal specialist services and could also be easily collected/recorded during a routine consultation with a GP or physiotherapist. We were practical about the need for brevity and simplicity for use of such an algorithm in busy clinical practice. Although the algorithm presented in Fig. 1 appears cumbersome, if its use for directing patients’ care at first consultation was to lead to better outcomes compared to the current model of stepped care, the algorithm can be fashioned in a ‘calculator’ form, embedded in the current electronic systems used by GPs, in the same way such systems are used for calculating cardiovascular risk for example.

Although we tested the final model using recommended internal validation statistics, we acknowledge the limitation of using data from only one cohort. Further testing in other similar populations will need to provide evidence on external validity, however, we are not aware that a similar primary care cohort is available. At this stage, we do not recommend the algorithm is used in clinical practice, as clinical trials are necessary first to assess the effects of the algorithm on patients’ outcomes. To that effect, we are currently testing the clinical and cost-effectiveness of using this subgrouping (stratification) algorithm with matched care pathways, compared to non-stratified care, in a randomised controlled trial (SCOPiC; ISRCTN75449581) [26].