Management of low back pain by primary care physiotherapists using the pain and disability drivers management model: An improver analysis

Clinical practice guidelines consistently recommend the rehabilitation management of LBP as effective first-line intervention (Longtin, Décary, Cook, & Tousignant-Laflamme, 2021; National Guideline Centre (UK), 2016; Qaseem et al., 2017). Worth noting, it is well documented that patients with LBP respond differently to rehabilitation management approaches (Fillingim, 2017). Classifying patients into subgroups based on various characteristics to select the appropriate treatment has been proposed as a potential solution to the variability in response to treatment (Fourney et al., 2011; Fritz et al., 2007). Nonetheless, subgroup classification systems have shown limited evidence of effectiveness when compared to generic intervention (Tagliaferri et al., 2021). One potential limitation of classification systems is that these often focus on pathoanatomic features, failing to combine biopsychosocial factors, which might better explain the patients' pain and disability (Rabey et al., 2015; Rabey et al., 2017). Identifying patient characteristics that exhibit a clinically meaningful response following treatment for LBP because of the treatment they have received is recognized as a priority by many organisations (Costa et al., 2013; Deyo et al., 2015; Ostelo et al., 2008). Recently, Tousignant-Laflamme and colleagues proposed the Low Back Pain and Disability Drivers Management (PDDM) model, a framework that integrates a true biopsychosocial approach to obtain a complete profile of the patient by identifying unique domains driving pain and disability (Tousignant-Laflamme et al., 2019; Tousignant-Laflamme et al., 2017). In a recent pilot trial, the authors compared the effectiveness of the PDDM model compared to the recent guidelines' recommendations for the management of LBP and found improved outcomes in patients treated with a PDDM approach (Longtin et al., submitted). The purpose of this study is to report additional findings of the PDDM trial: (1) the proportion of improvers, defined as patients in the intervention group that achieved a predefined level of improvement on a composite outcome; and (2) identification of baseline characteristics that predicted improvement at 12 weeks following physiotherapy based on the PDDM model. This study was a secondary improver analysis of a pragmatic pilot cluster nonrandomised controlled trial that determined feasibility and reported comparative outcomes between a PDDM- and guideline-oriented approaches (Longtin et al., submitted). Details on the rationale, design, and methods for the primary study are available in a prior publication (Longtin et al., submitted). We only included patients in the intervention group because our objectives were to determine the proportion of improvers and baseline predictors of good outcomes following 12 weeks of physiotherapy guided by the PDDM model. Nine primary care clinics from the private and public healthcare in the province of Quebec, Canada were included for a total of 28 physiotherapists. Patients were recruited during initial consultation between October 2021 and February 2022. Inclusion criteria involved being 18 years or older, care-seeking for a primary complaint of LBP, ability to understand and read French, and a working email address. Patients were excluded if the presence of red flags (i.e., cancer, fracture) was suspected following initial assessment. Sociodemographic characteristics of age, sex, LBP duration, education, household income, marital status, employment, medication, service payer, smoking status, previous back surgery and concomitant treatment were collected. These variables were dichotomised, when possible, to present the proportion of improvers in different subgroups. The clinical outcome measures were selected based on core outcome measures used in the LBP literature (A Chiarotto, Boers, Deyo, & et al., 2018; Alessandro Chiarotto, Terwee, & Ostelo, 2016) and related to the domains of the PDDM model (Table 1). They were used to measure the effectiveness of the PDDM model-guided approach in the primary trial. The presence of emotional distress at baseline was also documented via the 4-item Patient Health Questionnaire, which represents a valid screening tool for detecting anxiety and depressive disorders (Kroenke et al., 2009). Methodological approaches such as responder analyses purport to identify study participants who experience a ‘clinically meaningful’ improvement with a specific treatment. Despite a broad endorsement, these designs have several methodological shortcomings (e.g., natural history, regression to the mean, and changes related to common factors), and are unable to determine if the change is attributable to the treatment or something else (Stull et al., 2009). Accordingly, most responder analyses are more accurately identified as ‘improver analyses.’ We anchored our definition of an improver on documented clinically meaningful changes on relevant outcome measures. Composite-based methods are commonly used to define a clinically meaningful change in responder analyses (Henschke, van Enst, Froud, & Wg Ostelo, 2014; Pham et al., 2004). These methods incorporate multiple meaningful thresholds from outcomes of importance recommended by the literature into an amalgamated score (Henschke et al., 2014; Pham et al., 2004). It offers the advantage to provide a multidimensional definition of an improver to an intervention, not limited to pain intensity or physical function (Henschke et al., 2014; Pham et al., 2004). We developed a decision tree (Figure 1) to identify improvement following treatment based on validated and multidimensional outcome measures (Dougados et al., 2000; Pham et al., 2004). Decision tree to identify improvers based on biopsychosocial criteria. The decision tree is utilised for each patient. If there is a high improvement in pain severity or pain interference, the patient is considered an improver. If there is not a high improvement in these outcomes, the patient must have a moderate improvement in 3 or 4 of the domains identified to be classified as an improver. Pain and physical function are commonly used for the definition of response to an intervention (Henschke et al., 2014). For these measures, the BPI-severity and BPI-interference subscales were used (Tan et al., 2004). As there is no known minimally importance change scores documented for the BPI-severity subscale, we retained a ≥20% reduction as the criteria for a significant improvement since this threshold was previously documented for pain severity on a numerical rating scale (Cruz et al., 2020; Dworkin et al., 2008; Ostelo et al., 2008). For the BPI-interference subscale, a ≥20% reduction was documented as a minimally important change (Dworkin et al., 2009; Pham et al., 2004). The Patient global impression of change (PGIC) scale has served to identify responders to physiotherapy treatment for spinal pain (Domingues et al., 2019) and to determine prognostic factors of poor outcomes in patients with LBP (Cruz et al., 2020). This measure offers the advantage of letting the patients decide what is most important to them in terms of outcomes (Hurst & Bolton, 2004; Kamper et al., 2009). A score of ≥3 (considerably improved) on the PGIC scale at the 12-week follow-up was retained to identify improvers (Bourgault et al., 2015; Dworkin et al., 2008). Furthermore, we included fear-avoidance beliefs about work and physical activity (Monticone et al., 2020; Waddell et al., 1993). A reduction of 4 points for the FABQ-PA subscale and 7 points for the FABQ-W subscale were identified as meaningful changes in a similar sample of LBP patients (Monticone et al., 2020), and were thus retained for identifying improvers. These four outcome measures and their cut-off scores for clinically meaningful change are presented in Figure 1. We decided to add a large improvement (≥50%) in pain (BPI-severity) or function (BPI-interference) as a criterion to identify improvers based solely on these two outcomes (Dougados et al., 2000; Pham et al., 2004). All statistics were carried out using SPSS version 28.0 (Armonk, NY: IBM Corp). Descriptive statistics were used to present the participants' baseline characteristics. We measured the proportion of improvers based on the predefined criteria in the whole sample and in subgroups formed by categorical baseline characteristics. Bivariate logistic regression models were used to assess the association between each baseline sociodemographic and clinical variables and improvement following treatment. Models' assumptions were all validated visually with their respective diagnostic plots. Possible predictors of improvement were identified based on univariate p value of <0.1 in concordance with the exploratory nature of the analyses (Cruz et al., 2020; Hosmer & Lemeshow, 2000; Roseen et al., 2021). An odds ratio with its 95% confidence interval was calculated for each predictor. Sensitivity analyses were performed for predictors of improvement to treatment with a PGIC score of ≥3 as the only criteria. Of the 72 patients that provided data at baseline, complete data was available for 44 (61%) patients at the 12-week follow-up. One participant had missing data on the pain duration variable at baseline. Patients' characteristics at baseline of those with complete data (n = 44) were similar to those with missing data at follow-up (n = 28), except for age and marital status with patients with complete data being slightly older (mean 51.09 ± 13.24 vs. 44.29 ± 13.49, p = 0.048) and more likely to be married (p = 0.028). The reasons for missing data were exclusively due to the impossibility of reaching the participants and were considered lost to follow-up. The patients' sociodemographic and clinical characteristics are presented in Table 2. The mean age (±SD) was 51.1 (±13.24) and most were female (56%) with persistent pain lasting more than 3 months (86%). Most patients (52.4%) were classified as high-risk for developing long-term disability according to the STarT back screening tool (SBST). Overall, 59% (26/44) experienced a clinically important change at 12 weeks and were classified as being significatively improved following the PDDM intervention. Proportion of improvers by baseline sociodemographic characteristics and clinical variables with their associated OR are presented in Table 3. In the bivariate analyses, no baseline variables were found to be predictors of improvement following the intervention (all p-values >0.1). Although not statistically significant, we observed a large difference in the proportion of patients classified as improvers in the moderate- (64.3%) and high-risk (60.9%) groups compared to the low-risk group (42.9%). The sensitivity analyses identified the PROMIS-satisfaction with social roles and activities and the PAINdetect baseline scores as the only statistically significant predictors of improvement (p-values <0.05). Higher PAINdetect score at baseline (OR 0.91; 95% CI 0.82–0.91) was associated with lower chances of a successful outcome on the PGIC measure, but higher PROMIS-satisfaction with social roles and activities score at baseline (OR 1.16; 95% CI 1.03–1.30) was associated with higher chances of improvement. This is the first study which evaluated proportions of improvers and predictors of improvement for patients with LBP managed using the PDDM approach. More than half of patients (59%) met the predefined ‘improvers’ criteria, which is a similar percentage of patients (61%) who were involved in a comparative effectiveness study of cognitive functional therapy (CFT) versus manual therapy plus exercise intervention for chronic LBP at a 3-year follow-up (Vibe Fersum et al., 2019). Both CFT and the PDDM-guided approach aim to identify factors influencing pain and disability unique to the individual in a true biopsychosocial perspective (O’Sullivan, 2018; Tousignant-Laflamme et al., 2017). A landmark study on the effectiveness of stratified primary care management for LBP based on the SBST compared to best practice documented a similar response rate (71% for medium-risk and 68% for high-risk patients) at 4-month follow-up based on a 30% reduction in Roland Morris Disability Questionnaire score (Hill et al., 2011). These similar results are not surprising since the SBST is included within the PDDM model. Similar interventions may partly explain the comparable and relatively high response rates of studies involving chronic LBP when psychosocial factors are important drivers of treatment (Chou & Shekelle, 2010; Taylor et al., 2014). We did not identify predictors of improvement based on a biopsychosocial composite outcome. This absence of findings is not consistent with other studies that identified several sociodemographic (i.e., higher income, higher education) and clinical characteristics (i.e., high pain self-efficacy, low baseline disability and pain intensity and low fear-avoidance beliefs at work) as predictors of good outcomes in a sample of patients with LBP (Beneciuk et al., 2018; Cruz et al., 2020; Roseen et al., 2021; Trinderup et al., 2018a). Possible explanations for these inconclusive findings include a different and untested set of criteria for improvement and a small study sample (n = 44) with a majority of patients presenting with persistent pain and a high-risk of prolonged disability. A greater proportion of patients classified as improvers to the PDDM intervention in the moderate- (64.3%) and high-risk (60.9%) groups for prolonged disability at baseline. In addition to its predictive abilities, the SBST aims to target the best treatment option based on the patient's risk of poor outcomes to ‘beat’ the initial prognosis (Hill et al., 2008). As the SBST is embedded in the PDDM model, the physiotherapists could use its prognostic information to tailor treatment. The PDDM model seems to be especially helpful for clinicians to identify and treat patients presenting with a significant contribution of psychosocial factors. This trend is concordant with a previous study on the acceptability of the PDDM model in which clinicians mentioned that the model helped them assess and manage such factors by giving them a more structured approach (Longtin, Décary, Cook, Martel, et al., 2021). Limitations include a small sample size, a higher-than-expected attrition rate, and a composite score that has not been independently validated. Strengths include our prespecified definition of improvers based on established clinically meaningful changes aligned with recommendations from various research organisations and the collection of a wide range of baseline characteristics (Deyo et al., 2015; Dworkin et al., 2009), where we combined relevant outcome measures thresholds into a single composite score (Dougados et al., 2000; Pham et al., 2004). To our knowledge, this is the first study to use this method. We are optimistic that our method could be used to identify treatment-effect moderators in future studies, which are baseline characteristics predictive of treatment response of an individual and identifying which patients' response best to an intervention such as the PDDM model (Deyo et al., 2015; Dworkin et al., 2009). Our findings demonstrate that a high proportion of patients were identified as improvers to a 12-week PDDM-based physiotherapy intervention for the management of LBP. No predictors were identified when using a composite score, but sensitivity analyses identified two predictors of improvement when using the PGIC score to identify improvers. This improver analysis is based on pilot data; thus, interpretation of the study findings is limited. Christian Longtin: Conceptualisation; Methodology; Data collection; Investigation; Formal analysis; Data curation; Writing – original draft; Writing – review and editing; Project administration. Yannick Tousignant-Laflamme: Conceptualisation; Methodology; Investigation; Formal analysis; Writing – original draft; Writing – review and editing; Supervision; Project administration. Michel Tousignant: Conceptualisation; Methodology; Investigation; Writing – review and editing; Supervision; Project administration. Chad Cook: Conceptualisation; Methodology; Writing – review and editing. Anaïs Lacasse: Conceptualisation; Methodology; Writing – review and editing. All authors reviewed and approved the final version of the manuscript. We thank the physiotherapy professionals of all the participating clinics and the CIUSSSE ̶ Centre Hospitalier Universitaire de Sherbrooke for their support and participation in this research project. We would also like to thank the Fonds de Recherche du Québec—Santé (FQRS) for their financial support to the main author (CL) for its doctoral studies. Doctoral training scholarship from the Fonds de Recherche du Québec—Santé (FQRS). None of the authors have a conflict of interest to disclose. Ethic approval was approved by the Ethics Review Board of the CIUSSSE ̶Centre Hospitalier Universitaire de Sherbrooke (CHUS) (project number: 2021–3524). All participants provided signed informed consent. Clinicaltrial.gov: NCT04893369. The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.