Predictors of response in PROMIS-global in a chronic low back pain specialty clinic: STarTBack and chronic overlapping pain conditions

1.Introduction

Low back pain (LBP) afflicts 8.2% of United States adults [1]. It is the leading global cause of disability [2]. While multiple treatments are available, their mean treatment effects are typically mild, especially for chronic low back pain [3]. Even the best treatments improve pain by only two points on an eleven-point Visual Analogue Scale (VAS) scale. LBP is a heterogeneous condition influenced by multiple factors including biological, psychological, and social factors, as well as age [4], gender [5], race [6], culture [7], and co-morbidities [8, 9]. Identifying specific subgroups that respond best to targeted treatments is a key objective for LBP research [10].

The Keele STarTBack Screening Tool (SBT) [11, 12] was developed in the United Kingdom (UK) and identifies three subgroups of low back pain patients in primary care, defined by their risk of progressing to persistent disability (low risk, medium risk, high risk). The SBT is easy to administer as it has only 9 items. Four items focus on psychosocial factors: catastrophizing, fear, anxiety, and depression [11]. These psychosocial factors were chosen as they were thought to be modifiable in the primary care setting. The SBT successfully predicts long-term disability in both UK and United States (US) primary care settings [13, 14]. In the UK population specifically, stratifying treatment based on the risk subgroup reduced disability [12]. As such, SBT is being increasingly used to guide primary care pathways for back pain [15].

While the SBT has been extensively evaluated in primary care, there is less evidence supporting its use in other settings. SBT classification at initial presentation to the Emergency Department did not predict outcome pain intensity and disability at 6 weeks or at 26 weeks [16]. Field and Newell found that SBT is not as successful in differentiating outcomes in LBP patients seeking chiropractic care in the UK [17]. In the physical therapy setting, Beneciuk found that SBT was able to predict 6-month disability outcomes but not pain in a US population of chronic low back pain patients [18]. Kendell and colleagues studied SBT in a more heterogeneous population of cLBP patients in Western Australia recruited from general community, physical therapy, psychology, and pain management clinics. They found that the SBT did predict disability but was unable to discriminate changes in pain or global perceived change on a 7-point Global Rating of Change Scale [13]. The ability of SBT to predict disability has not been studied in specialty care settings.

Other groups have looked for alternative ways to screen those at high risk for chronic disability. Some have used pre-existing surveys assessing “pain”, “distress”, “social-environment”, and “medical care-environment” [19]. However, the need to use multiple surveys to measure multiple domains can hamper the ease of use of these instruments. A potential solution, which would improve risk stratification with minimal impact on patient and provider burden, is to use information readily available in the electronic health record (EHR) alongside SBT to help predict outcome. Rodeghero and colleagues demonstrated the feasibility of such a risk stratification approach for low back pain patients in physical therapy practice. They found that insurance type, duration of symptoms, and past surgery are the strongest predictors of outcome [20].

The objective of our study was to determine what features available in the EHR at the time of initial visit, including SBT, can predict change in function in response to treatment in a low back pain specialty clinic.

Figure 1.

Patient flow.

2.Methods

3.Results

Baseline characteristics among all 241 patients and by SBT category are summarized in Table 1. Overall, the mean age (± SD) was 59.5 ± 16.4 years, 61.7% were Female, 8.5% were Hispanic, with 44.0% being Caucasian, 14.0% African American, 28.4% Asian and 13.6% other race. About one-third of patients (29.7%) were low risk on the SBT, 41.8% were medium risk, and 28.5% were high risk. Average pain at time of starting ISS was 6.7 ± 2.1 per the numerical rating scale (NRS). The average baseline PROMIS mental health score was 44.8 ± 9.4, and the average baseline PROMIS physical health score was 39.4 ± 8.6. The average Charlson Comorbidity Index (CCI) was 1.3 ± 1.3, and the average number of COPCs was 0.25 ± 0.50, with 21.6% having at least one overlapping pain condition in the previous 6 months. The most common COPCs were migraine (11.2%), irritable bowel syndrome (4.1%), fibromyalgia (2.9%), tension headaches (2.5%), and endometriosis (2.1%).

Compared to low-risk and medium-risk patients, high-risk SBT patients were more likely to have increased pain and worse mental and physical health function at baseline (Table 1). High-risk SBT patients also had higher mean BMI and were more likely to be on Medi-cal and to have used analgesics and anti-depressants in the 6 months prior to baseline.

During an average of 17.0 ± 7.5 months between the baseline and follow-up visits, mean change in PROMIS mental health (MH) score was +0.86 (SD = 8.11), and mean change in PROMIS physical health (PH) score was +2.39 (SD = 7.52). Unadjusted associations between change in PROMIS scores and demographics (Table 2), pain characteristics (Table 3), and care-seeking behavior (Table 4) are shown. Baseline EHR features associated with a change in MH at p-value < 0.10 in unadjusted models are shown in Table 5. Compared to patients identified as low risk on the SBT, high-risk patients had an unadjusted mean 4.51 points greater improvement in their MH score (p= 0.001). On average, each additional chronic overlapping pain condition resulted in 3.72 points less improvement in the MH score (p= 0.0003). Compared to patients who never smoked, everyday smokers had, on average, a 8.96 points greater improvement in their MH score (p= 0.008). Patients with pain down the knee had a mean 2.68 points greater improvement in their MH score compared to those without knee pain (p= 0.04). When including all measures in Table 4in the same model, two features remained significant: 1) compared to low-risk SBT patients, high-risk patients had an adjusted mean 4.35 points greater improvement in their MH score (p= 0.004); and 2) on average, each additional chronic overlapping pain condition resulted in 4.23 points less improvement in the MH score (p= 0.0001).

Baseline features associated with a change in PH at p-value < 0.10 in unadjusted models are shown in Table 6. When all features were included in the same model, all measures found to be significant in the unadjusted models remained significant in the multivariate models. Compared to patients identified as low risk on the SBT, medium-risk patients had an adjusted mean 3.82 points greater improvement (p= 0.002) and high-risk patients had a mean 3.54 points greater improvement (p= 0.006) in their PH score. On average, each additional chronic overlapping pain condition resulted in 4.02 points less improvement in the PH score (p< 0.0001). For each 1-point increase in the Charlson Comorbidity Index score, there was an adjusted mean 1.06 points less improvement in the PH score (p= 0.004). Patients who had injections within the 6 months prior to baseline had an adjusted mean 9.31 points greater improvement in their PH score than other patients (p= 0.008). Patients with pain down the buttock/thigh had an adjusted mean 2.15 points greater improvement in their PH score compared to those without this pain (p=A 0.03). Compared to patients who never smoked, everyday smokers had, on average, a 6.64 points greater improvement in their PH score (p= 0.04).

4.Discussion

Two baseline features predicted both physical and mental health disability at follow up in LBP patients in a specialty clinic setting- the SBT and COPCs. On average, compared to SBT low-risk patients, SBT medium-risk patients had a 3.82 point greater improvement in physical health (PH), and SBT high-risk patients had a 3.54 point greater improvement in physical health (PH) and 4.35 point greater improvement in their mental health (MH) score on PROMIS-global. Each additional chronic overlapping pain condition (COPCs) resulted in an average 4.02-point decrease in PH and a 4.23 point decrease in MH scores. Other EHR features associated with greater improvement in PH scores (but not MH scores) in multivariate analysis were everyday smoking status, injections in previous 6 months, lower CCI, and having pain down the buttock/thigh.

Prior studies have shown that SBT can predict disability in patients LBP in the primary care [11, 12, 14], physiotherapy [24], as well as chiropractic [25] population. However, our study is the first to look at the utility of SBT in LBP population within a specialty clinic. As compared to Suri’s study, also in the US but in the primary care setting, our patients at baseline had higher levels of pain (6.6 here on VAS as compared to 5.4 on Numerical Rating Scale) with 29% being considered high risks per SBT as compared to 21% in their group. Similarly, the physiotherapy population also has lower rates of high risk at 19% [24]. This may be expected given that patients may be more likely to be referred to a specialty clinic should they have worse pain and higher risks.

Our study focused on correlating SBT with change in PROMIS scores rather than with baseline or with follow-up scores. Clinically, there is a need to identify risk stratification tools that would predict response to a treatment such as participation in a service like ISS. Prior pragmatic randomized trials, including two performed in the United States, have also correlated SBT to change in outcome in the primary care setting. In both trials, the intensity of care was stratified based on SBT with higher acuity patients receiving more care, similar to our approach. Such stratified approach did not necessarily result in improved outcomes after treatment [26, 27]. Our study design does not allow for any conclusions about the clinical utility of SBT in cLBP specialty care, as ISS is a psychologically informed pathway that shifted during the duration of this study rather than a standardized intervention. However, patients with higher risk scores on SBT demonstrated improved outcomes after completing this treatment, perhaps because the ISS team flagged them needing more intensive care. These higher risk patients were discussed at monthly multidisciplinary case conferences so that progress was closely monitored, and treatment adjusted as indicated; this may partly explain the greater improvement seen in the high-risk patients. The overall improvement was small (0.86 for mental health and 2.39 for physical health) and did not meet the threshold for a meaningful clinical response established by Lapin and colleagues: 2.73 (SD 6.37) for mental health score and 4.76 for physical health score [28]. This finding is somewhat similar to that found by Hill and colleagues in the trial performed in the United Kingdom by Hill and colleagues where the high-risk individuals actually had improved pain score as compared to lower risk ones [12]. Perhaps the intensified care made a difference versus those with worse baseline scores had more room to improve. Further examination is warranted to see if a stratified approach as guided by SBT can predict positive response in a LBP specialty clinic.

Aside from SBT, COPCs were also predictive for changes in PH and MH. There are several plausible explanations for why COPCs also predicted both changes in physical and mental health scores. COPCs refer a variety of pain disorders that are distinct conditions, such as headache, low back pain, fibromyalgia, temporomandibular disorders, and irritable bowel syndrome, but share common comorbidities and risk factors such as female sex, increased pain sensitivity, and genetic variants [29]. While the impact of the presence of COPCs has not been well studied in musculoskeletal pain, the presence of COPCs has been associated with worse outcomes for the treatment of other pain conditions, such as chronic migraines [30]. It has to be explored whether the presence of COPCs is also associated portends worse outcomes for LBP, as our study suggests. Our findings suggest that screening for the presence of COPCs at the time of referral to specialty care may be helpful.

Other EHR features associated with greater improvement in PROMIS were everyday smoking status, injections in previous 6 months, lower CCI, and pain down the buttock/thigh. The observed association between smoking and improved outcomes is counterintuitive but may be spurious as only 2.5% of our cohort were smokers. In general, ISS providers advise smoking cessation and possibly these individuals were motivated to stop smoking or particularly benefitted from the intensive intervention. The association between injections in the past six months and the improved outcome may represent confounding by indication. In other words, subjects who were selected for treatment with injections were also those that were likely to have the best outcomes with other treatments, perhaps due to more focal pathology. Regarding comorbidities, various studies have demonstrated the comorbidities and poor health in general are associated together and that LBP patients have increased number of comorbidities. An increased comorbidity burden has been associated with worsened outcomes for cLBP [31]. The reason for this association is unknown, but it is possible that co-morbidities lead to a sedentary lifestyle and social isolation, which may lead to worse LBP outcomes [31, 32, 33].

Our study has several limitations. The study stemmed from a retrospective quality improvement effort with non-standardized follow-up intervals, with only 241 of the 357 patients (67%) who went through ISS during the study timeframe included. This may reflect selection bias as follow-up EMR features are prone to classification bias. However, to account for this, we did perform a baseline comparison for the 241 ppts in our analysis versus the 116 ppts not in our analysis (lost to follow up) and found no differences in age, pain, PROMIS scores, or the proportion in each of the STarTBack risk categories (Supplemental Table 1). Higher risk patients per SBT had worse NRS and PROMIS scores at baseline. It is possible the larger improvements at follow up in the higher-risk patients is due to regression to the mean. We also only had access to NRS and PROMIS outcomes and would benefit from including other more sensitive outcome scales. Our study design does not allow for any causal conclusions about the clinical utility of SBT in cLBP specialty care. From here, a randomized controlled trial that includes risk stratification and standardized targeted treatments would be needed to determine the clinical utility of the SBT, with or without supplemental features, in a specialty clinic setting.

5.Conclusion

In conclusion, we found that SBT and the presence of COPC, which can be assessed at initial presentation to a LBP specialty clinic, predicts change in PROMIS following treatment. As identifying subgroups that will respond to targeted treatments is a crucial objective for LBP research, more effort is needed to identify other factors, even if nonmodifiable, that can help risk stratify patients. Phenotyping is the first step of developing more tailored care.

Author contributions

PZ, AT, CON collected data. PZ, CON, TH, SE, TP and ATE helped with data cleaning and analysis. All authors helped with manuscript writing and revisions.

Data availability statement

Data is available from the corresponding author upon request.

Ethical approval

The study received institutional review board approval from the University of California, San Francisco (#20-31733).

Funding

The research reported in this publication was supported by the National Institute Of Arthritis And Musculoskeletal And Skin Diseases of the National Institutes of Health (Award number U19AR076737). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Additional funding was provided by the Foundation of PM&R.

Informed consent

Per the institutional review board approval from the University of California, San Francisco (#20-31733), no informed consent was necessary as data was collected as part of a quality initiative project.

Supplementary data

The supplementary files are available to download from http://dx.doi.org/10.3233/BMR-230067.