aDepartment of Sports Science and Clinical Biomechanics, University of Southern Denmark, Odense M, Denmark
bNordic Institute of Chiropractic and Clinical Biomechanics, Odense, Denmark
cSchool of Physiotherapy and Exercise Science, Curtin University, Perth, Australia
dPrimary Care Centre Versus Arthritis, Research Institute of Primary Care and Health Sciences, Keele University, Staffordshire, United Kingdom
eDepartment of Physiotherapy and Occupational Therapy, Næstved-Slagelse-Ringsted Hospitals, Slagelse, Denmark
*Corresponding author. Address: Department of Sports Science and Clinical Biomechanics, University of Southern Denmark, DK-Campusvej 55, 5230 Odense M, Denmark Tel.: +4565504531. E-mail address: [email protected] (A. Kongsted).
Sponsorships or competing interests that may be relevant to content are disclosed at the end of this article.
Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal's Web site ( www.painrpts.com ).
This is an open access article distributed under the Creative Commons Attribution-ShareAlike License 4.0 (CC BY-SA) which allows others to remix, tweak, and build upon the work, even for commercial purposes, as long as the author is credited and the new creations are licensed under the identical terms.
The overall quality of care for musculoskeletal pain conditions is suboptimal, partly due to a considerable evidence-practice gap. In osteoarthritis and low back pain, structured models of care exist to help overcome that challenge. In osteoarthritis, focus is on stepped care models, where treatment decisions are guided by response to treatment, and increasingly comprehensive interventions are only offered to people with inadequate response to more simple care. In low back pain, the most widely known approach is based on risk stratification, where patients with higher predicted risk of poor outcome are offered more comprehensive care. For both conditions, the recommended interventions and models of care share many commonalities and there is no evidence that one model of care is more effective than the other. Limitations of existing models of care include a lack of integrated information on social factors, comorbid conditions, and previous treatment experience, and they do not support an interplay between health care, self-management, and community-based activities. Moving forwards, a common model across musculoskeletal conditions seems realistic, which points to an opportunity for reducing the complexity of implementation. We foresee this development will use big data sources and machine-learning methods to combine stepped and risk-stratified care and to integrate self-management support and patient-centred care to a greater extent in future models of care.
Substantial overlap between interventions and models of care for osteoarthritis and low back pain suggests potential for one common model, which may facilitate implementation.
Musculoskeletal pain conditions are the largest causes of disability worldwide. 30 Among the most disabling are knee and hip osteoarthritis (OA) and low back pain (LBP), which affect approximately 303 and 577 million people, respectively, resulting in very large annual societal costs. 30,35,86 An important contributor to these costs is poor-quality health care, including the application of non–evidence-based treatments. Examples of this are the overuse of imaging, surgery, and opioids, in circumstances when this is not aligned with current evidence-based recommendations for first-line care. 8,12,27,98,99
This suboptimal quality of care is in part due to the poor implementation of evidence-based guidelines for both OA and LBP. 24,27,32,37,62,88,94 Bridging this evidence gap to provide effective and affordable health care while ceasing the use of harmful, non–guideline-adherent practices is a major challenge. 12,24,27,31 However, it has great potential to significantly improve health and reduce costs. 12,26
Models of care that support best practice clinical decision-making are potentially very useful for implementing guideline-recommended care options ( Table 1 for definitions of key terms). 10,25 In OA and LBP, stepped care and stratified care models have been suggested to support decisions about care. With stepped care, all patients are initially offered the same basic care, and subsequent treatment efforts are only increased if patients have not benefited sufficiently. An example of a stepped care model is the “Beating Osteoarthritis” (BART) model. 95 Similarly, the “Good Life with osteoArthritis in Denmark (GLA:D)” program is considered the initial level in a stepped approach 92 (Appendix A: Summary of example models; available as supplemental digital content at http://links.lww.com/PR9/A76 ). In a risk-stratified care model, treatment is targeted to patient subgroups when patients initially seek care, with more comprehensive care offered to patients at risk of poor outcomes. The most widely known example of this approach is the STarT Back Screening Tool with treatments matched to risk profiles of people with LBP. 44,96
Table 1 - Definitions of key terms as used in this article.
Recommendations for clinical practice informed by systematic reviews of evidence and expert opinion, including an assessment of the benefits and harms of alternative care options.
Model of care
Framework that describes the principles of disease‐specific, evidence‐informed health care that should be delivered to consumers in a given setting, that is, the right care, at the right time, delivered by the right team, in the right place, using the right resources 10
A factor (or combination of factors) that affects the direction and/or strength of the relationship between the exposure (eg, treatment) and the outcome. For example, a factor that affects the response to a treatment.
A model/tool consisting of factors that in combination help predict the likely outcome of an individual with a given intervention
Treatment decision tool
A tool consisting of factors that in combination helps predict the likely effect of one treatment compared with another treatment
Risk stratification tool
A tool that combines a prediction model (of risk of poor outcome) with treatment options for each risk stratum
To engage with the international challenges around improving the management of musculoskeletal health conditions, there is a need for models of care that can be widely implemented. With recent reviews highlighting several similarities between recommended treatment options and clinical practice guidelines for OA, LBP, and other musculoskeletal conditions, it seems that similar models of care might also be appropriate, which would reduce the complexity of implementation. 5,58 Exploring the generalisability of models of care across conditions would inform clinicians, researchers, and decision-makers about ways of developing models of care and may facilitate implementation in clinical practice.
Therefore, the aim of this overview was to examine the conceptual similarities and differences between stepped care and risk-stratified care, discuss lessons learned from work on these models, and describe possible future directions for the field.
2. Key message #1: models of care for osteoarthritis and low back pain share many similarities
Multiple national and international guidelines exist for the care and management of OA and LBP. 58 These clinical guidelines typically offer a series of recommendations for clinical procedures and treatments with some offering explicit or implicit recommendations about models of care including stratified care models, stepped care models, and hybrids of the 2. For example, the NICE guidelines from the United Kingdom explicitly recommend risk-stratified care for LBP, 68 whereas the OARSI guidelines for knee and hip OA recommend a stepped care approach including a core initial treatment for all patients consisting of education and exercise with or without weight management. 8 Furthermore, the OARSI guidelines advocate elements of stratified care by differentiating care for people with and without comorbidities.
Examples of guidelines that implicitly recommend stepped care are the recommendations for OA and LBP management stating that surgery is only considered for specific groups of patients and only if nonsurgical care of a sufficient dose has already been provided without adequate symptom relief. 8,58,73 Another example is the recommendation that pharmacological treatment is only initiated for chronic LBP in patients who had inadequate response to nonpharmacological therapy. 79
In OA, stepped care is the most commonly advocated model of care, whereas in LBP hybrid models, elements of both stepped and risk-stratified care are included in clinical guidelines. 8,73 Both models have basic treatments (recommended for everybody with the condition), adjunct treatments (for some), and surgical options (for a minority). 8,44,73 Despite some differences in models of care, many aspects of treatment essentially remain the same. For example, self-management advice and education are included for all patients, and should be continuously delivered throughout all steps of the treatment pathway. 8,68,73
Exercise therapy is also a core element in the treatment of both OA and LBP due to its therapeutic effects in these conditions, the concurrent benefits of physical activity for common comorbidities and general health, and because it is a safe treatment option. 9,75,80,91 Examples in OA are the Enabling Self-management and Coping with Arthritis Knee Pain through Exercise (ESCAPE-knee pain) and Beating Osteoarthritis (BART) programs in the United Kingdom, the PARTNER model in Australia, and GLA:D for knee and hip offered in several countries in the world. 3,47,48,84,92 These are all stepped care models that include supervised exercises as part of the treatment, either for all patients (ESCAPE, PARTNER, and GLA:D) or where self-management advice has proven insufficient (BART). In LBP, most clinical guidelines do not recommend supervised exercises for acute LBP, but consistently do so for persistent LBP. 4,73 Following the stratified care principle, the NICE guidelines recommend that the decision about exercise therapy is guided by the STarT Back model for LBP with supervised exercises provided for people at medium risk and high risk of having persistent activity limitation after 6 months, irrespective of episode duration. 68
Adjunct and more intensive treatments, such as psychologically informed physiotherapy using a combined physical and psychological approach, are perhaps where the greatest differences between OA and LBP models are presently seen. Some LBP guidelines stress the importance of early identification of those at high risk of poor clinical outcome due to psychosocial obstacles to recovery in order that they are fast-tracked to therapists who can address these issues, 68 whereas OA guidelines have until recently only included psychological treatments to a lesser extent. 67 However, recent OA guidelines have recommended addressing psychological factors in certain patient subgroups and are, in this aspect, suggesting a stratified approach. 8 Also, return-to-work interventions are recommended without delay in both OA and LBP for subgroups who are struggling with their capacity to work or have been absent from work. Differences in treatment recommendations include that weight loss, orthoses, and corticosteroid injections can play a role in the treatment of OA, whereas that is not the case in LBP. Best practice for pain medication also differs to some extent for OA and LBP, with stepped models in OA suggesting different tiers of medication be tried in a stepped fashion, whereas in LBP, the role of pain medication is being questioned because of lack of effect and risk of harm. 87
Finally, decision-making about referral for surgical opinion is similar for OA and LBP in the sense that surgery is only considered if best-practice nonsurgical treatments have not provided sufficiently good outcomes. However, currently, there is a lack of evidence for surgical treatment as an effective option in nonspecific LBP. 27,73,97
In summary, clinical guidelines for both OA and LBP recommend elements of stepped care and stratified care models, including some explicit recommendations of stepped care in OA and stratified care in LBP. Although the same core treatments are recommended for patients with OA and LBP, some differences emerge in the clinical decision-making process around some adjunct therapies.
3. Key message #2: it is not a choice between stepped care or stratified care—rather, it is about using the best from both of them
A fundamental difference between these 2 models of care is that stepped care assumes a substantial number of patients will improve with core treatment, and patients who do not are not harmed by waiting for more comprehensive treatment to be initiated. 59 By contrast, risk-stratified care assumes that patients with poor outcomes can be identified at an early point of care-seeking and their risk of poor outcomes reduced by early targeted interventions. 45,59
The risk of overtreatment and undertreatment with stepped care clearly depends on the content of care at each step. If the initial step only includes, eg, self-management advice, more people are potentially being undertreated with a stepped approach than by risk stratification. By contrast, where more comprehensive core treatment packages are the initial level of care (eg, including pain medication), there is a risk of overtreatment and adverse events in patients who would improve sufficiently with self-management advice alone. The likelihood that patients are overtreated or undertreated with risk-stratified care depends on the accuracy with which patients benefitting from more intensive care can be identified. 33,51 In addition to an accurate screening tool, better outcomes rely on there being suitably effective treatment options available for each risk stratum. Both models of care guide which general approach to follow, with room for clinical judgement and individualisation of care that directs treatment at patients' individual prognostic factors, impairments, and treatment preferences.
Despite differences, the fundamental approach that informs both of these models is that treatment choice differs between patients based on individual patient information. In stepped care, this information is the response to prior treatment and in risk-stratified care, it is information on prognostic factors (the predicted outcome). 36,72 Consequently, patients seeking care for the first time with an estimated good prognosis will have the same treatment with stepped or stratified care. This may also apply to patients seeking care early after symptom onset because there is greater uncertainty with prediction in the acute stage. 63
To the best of our knowledge, no trials have directly compared the effectiveness of different models of care in musculoskeletal conditions. Stratified care for LBP based on the STarT Back Screening Tool was cost-effective when compared with usual care (no specified model of care) in a U.K. effectiveness trial and in an implementation study, with usual care being determined by the individualised treatment decisions made by general practitioners. 28,44 In an Irish nonrandomised controlled study, improved outcomes were observed for high-risk patients with risk stratification, seemingly without affecting outcomes for low-risk and medium-risk patients. 66 Also, an implementation strategy, including the use of the STarT Back Screening Tool in Danish general practice, led to lower rates of referral to hospital settings. 83 However, the implementation of the STarT model was not as successful in a U.S. study, 15 illustrating that effective implementation strategies for models of care may vary across different international health service contexts. The effectiveness of a stepped care model for other musculoskeletal conditions has not been investigated using a controlled design. However, a randomised controlled trial in knee OA has demonstrated that a substantial proportion of patients considered eligible for surgery achieved satisfactory outcomes with alternative nonsurgical treatment and decided not to proceed with surgery. 93 This implied that the stepped approach of offering core nonsurgical treatments before considering surgery is an effective strategy for some patients.
Currently, there are a number of trials investigating the effectiveness of risk-stratified care for LBP in different settings and countries, 13,38,64 and trials are being conducted to test stepped care approaches in OA 2 and LBP 81 (examples were identified from ClinicalTrials.gov ). Also, we identified one published pilot study and protocol for a trial testing stratified care in musculoskeletal conditions including OA. 39,40 No trials or protocols were identified on head-to-head comparisons of stepped and stratified models of care. Although there is limited research to inform whether models of care could be the same for OA and LBP or if one model should be chosen rather than the other, a recent focus group study indicated that practitioners consider a generic stratified model to be relevant across 5 musculoskeletal conditions. 78
Given the similarities and overlapping practical applications of stepped and risk-stratified care, these models may be considered parts of a common approach with elements of each model being present at different time points in decision-making ( Fig. 1 ). As described in key message #1, such a combined approach is implicitly advocated by current clinical guidelines. A stepped approach is used in the very early stage of symptoms, where all patients are offered basic care supporting self-management. 2,20,97 From primary care settings, which mostly managed nonacute LBP presentations, there is evidence that risk stratification can be cost-effective as a tool for making decisions about care pathways. The precise content of treatment is to some extent individualised within both stepped and stratified care. For example, in the GLA:D programs for knee, hip, and back pain and the ESCAPE-knee pain intervention, there is individualisation of exercise programs by adaptation of type and dose to the individual patient. 48,92 Similarly, in LBP, patients who are stratified to the same level according to the STarT Back Screening Tool have different treatments tailored to their individual risk factors. 36,92 In both OA and LBP, decisions about surgical care follow a stepped approach based on observed response to nonsurgical care (except for very special cases of trauma or progressive neurological deficits). 8,50 At the level of surgical assessments, risk stratification again has a role, which may mean that people with a high risk of poor outcome from surgery would not be recommended this treatment or are perhaps offered more intensive rehabilitation ( Fig. 1 ). However, although there are known risk factors for poor outcome with surgery, currently there are no well-established decision support tools for surgery or other secondary care settings. There is a model for predicting outcome of total knee arthroplasty albeit with only partial support for its external validity. 21,82 Also, a decision support tool for spinal fusion surgery is being developed, 85 but there is no evidence yet that implementation of risk stratification on the basis of these 2 approaches does actually improve patient outcomes. There is evidence that the STarT Back Screening Tool, which was developed for primary care, is not useful for predicting outcomes in secondary care settings, 52,65 but stratified care for LBP (matched treatment to risk profiles) has not been investigated in secondary care settings. Also, at this point, there is no evidence to inform exactly how models of care might most optimally be combined.
Principles of combining stepped and stratified models of care. Existing models of care coexist in musculoskeletal care to form a joint model with stepped and stratified approaches to making treatment decisions being used at different points along the clinical course. The figure illustrates principles and some parts of this combined model remain to be investigated.
Although musculoskeletal conditions are very often recurrent, 18,19,71,74 and many people with OA and LBP seek care repeatedly with periods of self-management in between, none of the existing models explicitly suggest how recurrence is handled. Episodic pain generally has less negative impact than persistent pain and recurrences do not necessarily indicate progression of the condition. 55 Therefore, recurrences should not automatically lead to more intensive care, and treatment decisions should rationally take previous treatment response and success with self-management into account.
4. Key message #3: risk stratification has different roles and implications depending on how, when, and where it is applied
In general, within risk stratification tools, there are 2 different types of prognostic factors: those that are treatment-modifiable (eg, pain, fear, and pain catastrophising) and those that cannot be modified with treatment (eg, number of previous recurrences, age, and previous surgery). This distinction becomes important when designing prognostic tools and prediction models because both types of prognostic factors have different roles and implications for clinical practice. The inclusion of treatment modifiable factors has the benefit that it gives clinicians some signals about potential treatment targets. By contrast, nonmodifiable factors, although they do offer useful prognostic information, are not helpful for identifying the specific target of treatments but could have the advantage, by their very nature, of being more stable than modifiable factors in their predictive abilities across different settings. An example to illustrate that prediction by modifiable factors can be unstable is the finding that the STarT Back Screening Tool (which only includes modifiable factors) had poor predictive performance in emergency care 61 and among patients with a very short LBP duration (