Improved adherence to clinical guidelines for low back pain after implementation of the BetterBack model of care: A stepped cluster randomized controlled trial within a hybrid type 2 trial

ABSTRACT

Background

The BetterBack model of care (MoC) for low back pain (LBP) was recently developed in Swedish physiotherapy (PT) primary care.

Objective

To evaluate if PTs’ adherence to LBP clinical practice guidelines (CPGs) improves after implementation of the BetterBack MoC (intervention).

Methods

This was a stepped, single-blinded cluster randomized controlled trial. Patients nested in the three clusters were allocated to routine care (n = 222) or intervention (n = 278). The primary outcome was referral to specialist consultation. This was among five best practice recommendations divided into an assessment quality index (no referral to specialist consultation and no medical imaging) and a treatment quality index (use of educational interventions; use of exercise interventions; no use of non-evidence-based physiotherapy). For overall adherence, patients had to be treated with all five recommendations fulfilled. Logistic regression was used for between-group comparisons.

Results

The proportion of patients receiving referral to specialist consultation during the PT treatment period was low in both groups with no between-group differences. However, patients in the intervention group showed significantly higher assessment quality index, treatment quality index and overall adherence compared to routine care. Adherence to the separate recommendations showed improved stratified number of visits, use of exercise was maintained high, patient educational intervention increased and use of non-evidence-based physiotherapy decreased. A reduction of medical imaging during the physiotherapy treatment period was also observed.

Conclusions

The adoption of CPGs could be substantially improved by introducing a MoC through PT training and supportive materials.

KEYWORDS:

• Clinical practice guidelines
• physiotherapy
• low back pain
• primary health care
• adherence

Introduction

Implementation research in the field of low back pain (LBP) is scarce, and especially for evaluating the effects on health-care practitioner’s behavior change (Hartvigsen et al., 2018; Hodder et al., 2016). Evidence-based clinical practice guidelines (CPGs) are considered important tools to improve quality in health care (Suman et al., 2015). Current CPGs for the treatment of LBP recommend diagnostic triages, patient education and physical exercise as first-line interventions to enable patient’s ability to stay active and self-manage LBP (Hill et al., 2011; National Clinical Guideline Center, 2016; O’Connell, Cook, Wand, and Ward, 2016). Moreover, use of risk stratified primary care with fewer visits in groups with low risk of prolonged problems and more visits and use of behavioral medical interventions in high-risk groups may improve clinical outcomes and cost savings (Hill et al., 2011). CPGs also recommend against the use of non-evidence-based interventions such as passive physical modalities as well as routine use of medical imaging or referral to specialist consultation for benign LBP (O’Connell, Cook, Wand, and Ward, 2016). An evidence-to-practice gap for LBP care has been identified internationally, with poor uptake of these treatment recommendations (Hartvigsen et al., 2018).

In Sweden and in many other countries, patients with LBP have direct access to physiotherapists (PTs) in primary care. This patient management has been found to be cost-effective (Goodwin and Hendrick, 2016); safe; and well received by patients (Goodwin and Hendrick, 2016; Ludvigsson and Enthoven, 2012). However, there are still treatment interventions that Swedish PTs in primary care consider to be relevant despite lack of evidence for their effectiveness, such as advice on posture and transcutaneous electrical nerve stimulation (TENS) (Bernhardsson et al., 2015). Adherence can be defined as the extent a person’s behavior corresponds to recommendations (Word Health Organization, 2003). CPG adherent care has been found to be associated with improved outcomes and lower cost of treatment in patients with LBP (Fritz, Cleland, and Brennan, 2007; Hanney et al., 2016; Rutten et al., 2010). The content of recent guidelines for LBP treatment highlights a shift toward a more biopsychosocial management approach. However, the literature shows that PTs are hesitant to address these new biopsychosocial aspects in LBP and prefer dealing with the more mechanical aspects (Hall et al., 2018; Richmond et al., 2018; Synnott et al., 2015). Health-care practitioners using a biomedical orientation are less likely to adhere to treatment guidelines such as patient advice on early return to work and normal activity (Darlow et al., 2012; Gardner et al., 2017).

Bernhardsson et al. (2014) showed that 96% of a cohort of Swedish PTs considered guidelines important, but fewer than half reported using guidelines frequently. A review on implementation of various knowledge translation interventions into PT practice showed that attitudes and beliefs, skills and guideline awareness can be improved, but there were no consistent improvements in clinical practice and patient outcomes (Berube et al., 2018). In a scoping review, seven of the nine studies showed increased PT adherence to CPGs as a result of knowledge translation strategies (Stander, Grimmer, and Brink, 2018). However, studies often had short measurement intervals with 6 months as the most common interval, small sample sizes and different criteria for adherence. Furthermore, there is no consistent evidence on the most effective LBP guideline implementation strategy to improve professional behavior outcomes, except that multifaceted implementation significantly improves the use of active treatment advice (Suman et al., 2015).

To bridge the gap between research and clinical practice, CPGs can be delivered as a model of care (MoC) including support tools to facilitate implementation and de-implementation of target behaviors (Briggs et al., 2017; Wang, Norris, and Bero, 2018). Guideline implementation is challenging, repeated actions to stimulate practitioner’s behavioral change are of importance for implementation success (Mesner, Foster, and French, 2016; Stander, Grimmer, and Brink, 2018). Furthermore, targeting different organizational levels and addressing local barriers can also contribute to implementation success (Fixsen et al., 2005). To facilitate this, the BetterBack☺ MoC has been developed and implemented with a multifaceted strategy in Swedish primary care through a hybrid trial research program containing series of parallel sub-studies (Abbott et al., 2018). In line with the a-priori hypothesis for the implementation process sub-study, improvements in the practitioner confidence scale (PCS) and the attitudes and beliefs scale for PTs (PABS-PT) along with facilitating determinants of implementation behavior (DIBQ) were observed after implementation of the BetterBack☺ MoC (Schröder et al., 2020). For the patient outcome related sub-study, a-priori hypothesis was partially supported by the BetterBack☺ MoC showing superiority to routine LBP care for patient satisfaction at 3 months, meaningfully improved illness perception at 3 months and health-related quality of life at 3 and 6 months (Schröder et al., 2021). In a final sub-study investigating implementation outcomes, we hypothesized that, compared to routine LBP care, the BetterBack☺ MoC would result in increased PTs adherence to CPG recommendations for LBP. The aim of the present study was to evaluate if physiotherapists’ adherence to CPG recommendations for LBP improves after a multifaceted implementation of the BetterBack☺ MoC compared to routine care in Swedish primary care physiotherapy.

Methods

Trial design

This was a stepped single blinded cluster randomized controlled trial (Hemming et al., 2015), nested within a hybrid type 2 effectiveness-implementation trial (Abbott et al., 2018). The current sub-study concerns primary and secondary outcomes for the BetterBack☺ MoC implementation related to registry and therapist reported use of guideline consistent interventions as reported in the protocol (Abbott et al., 2018).

Participants and setting

Eligible participants were registered PTs who worked regularly with patients with LBP. All primary care PT clinics in the Östergötland public health-care region of Sweden were represented. The clinics formed three units, which became the three clusters based on municipal and geographical area and organizational structure in Östergötland’s health-care region. This cluster randomized multicenter design was applied due to the logistics involved in the implementation in the different geographical areas and to prevent cross-contamination between the individual physiotherapists. There were no organized joint educational activities between clusters during the study time period. Cluster 1 (West) comprised 5 clinics with 31 PTs; cluster 2 (Central) 4 clinics with 48 PTs; and cluster 3 (East) 4 clinics with 44 PTs. The three units (cluster level) were randomized to either routine care or the BetterBack☺ MoC intervention. The PTs in participating clusters (practitioner level) and their patients (patient level) were nested within these three clusters. Patients were therefore allocated to routine care or intervention according to a stepped cluster (Hemming et al., 2015) dogleg structure (Hooper and Bourke, 2014). With the dog-leg design, one group was assessed before and after the intervention, whereas two other groups provide respective, independent treatment comparisons in each period (Hooper and Bourke, 2014). As outlined in Table 1, the first cluster was assessed during the study after an initial implementation of the BetterBack☺ MoC. The second cluster was assessed during a period of routine care (control) and assessed again after the implementation of the MoC (intervention). The third cluster was assessed during the study delivering routine care (control).

Table 1. The stepped cluster structure (dog-leg design) of the trial.

2017											2018
Mar	Apr	May		Jun	Jul	Aug	Sep	Oct	Nov	Dec	Jan	Feb	Mar
PT Cluster random allocation	Patient recruitment during internal pilot phase			Patient recruitment during main trial phase
Cluster 1 West	MoC	MoC		MoC	MoC	MoC	MoC	MoC	MoC	MoC	MoC	MoC	MoC
MoC implementation
Cluster 2 Central	Routine	Routine		Routine	Routine	Routine	MoC	MoC	MoC	MoC	MoC	MoC	MoC
						MoC implementation
Cluster 3 East	Routine	Routine		Routine	Routine	Routine	Routine	Routine	Routine	Routine	Routine	Routine	Routine

PT = physiotherapist, MoC model of care = intervention group, Routine = control group.

All PTs working in public financed primary care in the region of Östergötland were invited to a mandatory 2-day BetterBack☺ MoC workshop between March 2017 and January 2018. The workshops were held at the largest clinic in each cluster. Patients with LBP were treated by the participating PTs during the study period. The inclusion criteria for patients were 18–65 years; fluent in Swedish; and accessing public primary care due to a first-time or recurrent episode of acute, subacute or chronic phase benign LBP with or without radiculopathy. Exclusion criteria: current diagnosis of a malignancy; spinal fracture; infection; cauda equina syndrome; ankylosing spondylitis or rheumatic disease; previous malignancy the last 5 years; current pregnancy or previous pregnancy up to 3 months; patients who fulfill the criteria for multimodal/multi-professional rehabilitation for complex long-standing pain, severe psychiatric diagnosis; and spinal surgery the last 2 years. Based on an a priori hypothesized small-medium effect size difference from the trial protocol (Abbott et al., 2018) and previously reported <50% CPG adherence in routine Swedish physiotherapy context (Bernhardsson et al., 2014) which has equivalence with an odds ratio = 2 (Chen, Cohen, and Chen, 2010), a one tailed p-value = 0.05 and 80% power, at least n = 112 patients are needed for logistic regression analyses.

Intervention, randomization and allocation

The BetterBack☺ MoC was developed based upon two clinical guidelines for LBP from the Danish Health and Medicine Authorities and the English National Institute for Health and Excellence (National Clinical Guideline Center, 2016; Sundhedsstyrelsen, 2016a, 2016b). To support the development, implementation and evaluation of the MoC, an international framework (Briggs et al., 2016) and the Behavior Change Wheel were used (Michie and Johnston, 2004; Michie et al., 2008; Michie, van Stralen, and West, 2011). To optimize transparency, the study protocol was published with open access (Abbott et al., 2018). A multifaceted implementation strategy targeting management and clinical organizational levels was used (Fixsen et al., 2005; Nilsen, 2015). The implementation started with a top-down strategy (Nilsen, Ståhl, Roback, and Cairney, 2013) with a request to improve LBP care from the rehabilitation managers. Then, a bottom-up strategy (Nilsen, Ståhl, Roback, and Cairney, 2013) was used with the involvement of an MoC support team, which consisted of a researcher and six clinical champions (Fixsen et al., 2005). The six clinical champions were selected by the rehabilitation managers from each unit and were trusted clinicians with special skills in LBP diagnosis and treatment. The MoC support team adapted two international clinical guidelines (National Clinical Guideline Center, 2016; Sundhedsstyrelsen, 2016a, 2016b) into 11 CPG recommendations for the Swedish context using the Swedish National Board of Health and Welfare (2016) methods for guideline construction. Each of the 11 CPG recommendations was supported by a clinical priority ranking (Supplementary File 1). Of these, eight CPGs were focused on physiotherapy scope of practice.

The MoC support team designed the mandatory 2-day workshop (13½ hr) for the PTs. The managers supported the implementation by making participation in the workshop mandatory for all PTs and by organizing the workshop on two alternative occasions in each setting. The workshop started with a presentation of the local adaptation process together with the 11 guideline recommendations and thereafter presentation and practical use of clinical tools to support CPG adherent health-care delivery. The supporting clinical tools (Supplementary File 1) included a standardized care pathway; structured subjective and objective assessment proformas; STarT Back Screening Tool (Hill et al., 2008); clinical reasoning and process evaluation tool; standardized patient education brochure and material supporting group-based patient education; standardized tools supporting the design and progression of individualized home-based and/or group-based exercise program; web-based educational module; and chat forum to facilitate communication between clinicians and researchers (Abbott et al., 2018). The workshop also involved patient cases, where PTs worked in small groups to apply the BetterBack☺ MoC to the patient cases. The intervention delivery to patients and dosing was stratified based on PTs’ clinical reasoning regarding risk of pain persistence and progression toward patients’ goals. Random concealed allocation was performed by a blinded researcher (AA) selecting from three sealed envelopes of the cluster’s allocation order. A researcher (KS) informed the clinics in the three clusters of their allocation either to routine care or intervention study condition. The recruited patients were blinded as they were unaware of the difference between routine care and the BetterBack☺ MoC. PTs in the clusters working according to routine care were unaware of the content of the BetterBack☺ MoC. The statistician used for reviewing statistical analyses was blinded. Patients were consecutively recruited during the study period in the three clusters. An internal pilot was done to determine the PTs’ acceptability of the intervention and trial within the first cluster (Abbott et al., 2018). To ensure sustainable long-time use and facilitate uptake, the clinical champions were involved in the workshops. The clinical champions provided reminders during the study, trained new staff and were the local clinical contact person throughout the study. Furthermore, feedback on patient recruitment rates and reminders regarding the use of the BetterBack☺ MoC was provided to PTs and managers in the study clinics during 1–2 outreach visits from the MoC support team to support sustainability (O’Brien et al., 2007). All treatments were delivered at PTs’ local clinics. Description of the implementation strategy and components of the MoC can be found in the protocol (Abbott et al., 2018). The reporting on this study follows the CONSORT checklist for cluster randomized controlled studies and the TIDieR checklist for intervention studies (Campbell, Piaggio, Elbourne, and Altman, 2012; Hoffmann et al., 2014).

Data collection

Referral to specialist care, imaging and health-care visits were collected from patients’ medical record from the public health-care regional registry by the researchers (Table 2). Since this registry datum was complete, a larger cohort was used than the data collected from the therapists and patient reported questionnaires. These data sources also provide descriptive information for the study population, including age, gender, pain duration, pain intensity, Oswestry disability index, Euro Qol 5D, general health and SBT risk groups. PTs reported choice of treatment for each patient in a paper-based clinical reasoning and process evaluation tool (CRPE). The CRPE tool was completed by PTs at the first and last consultation for all patients presenting a new or recurrent episode of LBP during the study period.

Table 2. CPG assessment- and treatment-related quality indicator variables, definitions, source of data and target behaviors.

	CPG related indicator variables	Definition	Source of data	CPG related target behavior
Assessment interventions	1. Proportion of patients receiving referral to specialist consultation	Referral to pain clinic, orthopedic or neurosurgical care during PT treatment period	Public health care regional registry	Reduce referral for benign non-complex LBP
	2. Proportion of patients receiving medical imaging	X-ray, MRI, CT scan during PT treatment period	Public health care regional registry	Reduce imaging for benign non-complex LBP
Treatment interventions	1. Proportion of patients receiving stratified number of PT visits	The number of visits to PT within actual treatment period for patients sub-grouped in low or medium/high-risk of persistent symptoms	Public health care regional registry	Fewer visits in the SBT low risk subgroup (1–3 visits) and more in the SBT medium/high-risk subgroup (≥4 visits)
	2. Proportion of patients receiving educational interventions	Reassuring information/education on pain, pain management and relapse prevention applying behavioral medicine approaches	CRPE	Increase overall use and stratified use with higher proportions in higher risk groups
	3. Proportion of patients receiving exercise interventions	Supervised/monitored individualized home or group-based exercise program	CRPE	Increase overall use
	4. Proportion of patients receiving manual therapy	Joint and/or soft-tissue mobilization/manipulation	CRPE	Short-term use as a part of †multimodal treatment when LBP with segmental movement impairments are found
	5. Proportion of patients receiving acupuncture	Acupuncture	CRPE	Short-term use as an adjunct to treatment intervention 2&3 when additional short-term analgesic effect as goal
	6. Proportion of patients receiving non-evidence-based treatments	Passive physical modalities e.g. corset, static traction, electrotherapy, ultrasound, *non-evidence-based manual therapy, **non-evidence-based acupuncture	CRPE	Decrease use

CPG Clinical practice guideline, PT physiotherapy, LBP Low Back Pain, MRI Magnetic Resonance Imaging, CT Computerized Tomography, CRPE clinical reasoning and process evaluation data from physiotherapists. SBT STarT Back Screening Tool, † multimodal treatment = exercise and/or educational interventions, *non-evidence manual therapy = manual therapy without reported segmental movement impairments or used as unimodal intervention not combined with exercise and/or educational interventions, **non-evidence acupuncture = acupuncture used as unimodal intervention not combined with exercise and educational interventions.

Outcome measures

The CRPE tool (Supplementary File 1) is a standardized assessment, which enables analysis of the PTs’ treatment protocol that was based on the International Classification of Functioning, Disability, and Health (ICF) brief core set for LBP (Cieza et al., 2004) and health-care intervention codes. STarT Back Screening Tool (SBT) (Hill et al., 2008) was reported by patients at baseline. An overall score ranging from 0 to 9 is calculated and dichotomized into low risk (score 0–3) and medium/high risk (score 4–9) patient subgroups based on prognostic risk for persistent LBP-related disability (Foster, Hill, O’Sullivan, and Hancock, 2013). The eight PT-related CPG assessment and treatment-related quality indicator variables, definitions, source of data and target behaviors are outlined in Table 2. The primary outcome was referrals to specialist consultation during PT treatment period. This was chosen due to the high number of referrals from primary to secondary care for patients with benign LBP within the health-care region.

A clinical practice quality index was thereafter created based on the primary outcome variable and CPG-related quality indicator variables with the highest clinical priority rankings identified by the MoC support team in the BetterBack☺ MoC development phase. The clinical practice quality index included an assessment quality index containing two criteria and a treatment quality index containing three criteria, as outlined in Table 3. For overall adherence, all five criteria had to be fulfilled.

Table 3. PT clinical practice quality index for benign LBP.

	PT clinical practice quality indicators forming the clinical practice quality index
Assessment quality index	Cohort with registry data 1. No referral to specialist consultation during PT treatment period
	2. No imaging during PT treatment period
Treatment quality index	Cohort with CRPE data 1. Use of patient education interventions
	2. Use of exercise interventions
	3. No use of non-evidence-based interventions
Overall adherence	Cohort with CRPE data All 5 quality indicators fulfilled

PT physiotherapy, LBP Low Back Pain, CRPE data clinical reasoning and process evaluation data from physiotherapists.

Statistical methods

The PTs’ and patients’ baseline characteristics are presented as means with standard deviations (SD) and proportions (%). Analysis was performed on a per protocol basis by comparing the proportion of adherence to CPG recommendations in the intervention versus control. In a secondary analysis, the highest clinical priority ranked CPG-related indicator variables were merged and analyzed using a clinical practice quality index. Odds ratios (OR) with 95% confidence intervals (95% CI), p-value and ICC (intracluster correlation coefficient) are presented from the multilevel logistic regression analyses for binary variables. Scores were corrected for the clustering effects, and ICC was calculated according to the three regions West, Central and East that formed the two groups intervention and control with equal groups assumed. ICC was calculated in an empty model with no predictors. Sub-analyses of adherence according to SBT risk groups divided into low and medium/high risk were also performed. These SBT sub-analyses were used to analyze two target behaviors outlined in Table 2: 1) Fewer visits in the SBT low-risk subgroup (1–3 visits) and more in the SBT medium/high-risk subgroup (≥4 visits); and 2) Increase overall use and stratified use of educational interventions with higher proportions in higher risk groups. The regression coefficients were transformed to odds ratios with 95% confidence intervals (CI), indicating the effect of the intervention. Statistical analysis was performed using SPSS statistical software for windows (SPSS V25, IBM Corporation, New York, USA) and with the statistical package R 3.6.0 (https://cran.r-project.org). The level of significance was 0.05.

Result

A total of 500 patients with LBP were recruited from April 3, 2017, to March 8, 2018, with 278 patients allocated to intervention and 222 to control. For the variables: imaging, number of visits and referral to specialist care, the study includes complete registry data including all 500 patients. For PT reported treatment (CRPE data) a smaller patient cohort of 388 patients was used. In this cohort, 44 PTs treated 223 patients in the trial clusters delivering care after the implementation of the BetterBack☺ MoC and 54 PTs treated 165 patients in the trial cluster periods delivering routine care. Because the second cluster had a period of routine care before the BetterBack☺ MoC was implemented, 12 PTs treated patients during the routine care period and during the period after MoC implementation. This study reached 48% of all 1034 patients seeking physiotherapy in public financed primary care in the region of Östergötland during the study period that met the inclusion criteria. The CONSORT flow diagram is presented in Figure 1. Baseline characteristics of the PTs (Table 4) and patients (Table 5) did not differ significantly between groups. The registered data showed that participating patients mean age was 45.2 (SD 12.2) and 55.8% were female. Pain duration of less than 12 weeks was slightly more common. The mean ODI was 31.0 (SD 15.8) and can be categorized as moderate disability (Fairbank and Pynsent, 2000). The mean EQ-5D index was 0.535 (SD 0.301). The largest SBT risk group was the medium risk group with half of the patients (50.1%), followed by the low-risk group (36.8%) and high-risk group (13.1%).

Figure 1. CONSORT flow diagram.

* 12 PTs in cluster 2 crossed over from control to intervention phase after implementation of the BetterBack☺ MoC and therefore treated patients in both phases of the study

Table 4. Characteristics of PTs treating patients during control and intervention phases of the trial.

	PTs in intervention† n = 44	PTs in control† n = 54	P values
Age, mean (SD)	35.4 (11.8)	36.1 (11.1)	0.776§
Sex: female, n (%)	30 (68.2)	37 (68.5)	1.0#
Clinical experience, years, n (%)			0.681#
1–5	27 (61.4)	25 (48.1)
6–10	5 (11.4)	10 (19.2)
11–15	3 (6.8)	3 (5.8)
16–20	4 (9.1)	4 (7.7)
21–25	3 (6.8)	4 (7.7)
> 25	2 (4.5)	6 (11.5)
Clinical specialist qualification*, n (%)	3 (6.8)	1 (1.9)	0.330#
Highest educational level**, n (%)			0.937#
Bachelor’s degree	35 (85.4)	44 (86.3)
Post graduate major	4 (9.8)	5 (9.8)
Post graduate master	1 (2.4)	2 (3.9)
PhD	1 (2.4)	0 (0)

PT physiotherapist, †12 PTs treated patients in both intervention and control group, SD Standard deviation, n number of observations, PhD Doctor of Philosophy, *Clinical specialist qualification by the Swedish association of registered physiotherapists, **Education levels with European Tertiary Credit system, §Independent T-test, #Fisher's Exact Test.

Table 5. Baseline characteristics of the included patients.

	Original cohort with registry data		P values	Cohort with complete data		P values
	Intervention n = 278	Control n = 223		Intervention n = 223	Control n = 165
Age, mean (SD)	45.0 (12.1)	45.3 (12.4)	0.779§	45.2 (12.2)	45.1 (12.6)	0.962§
Sex: female, n (%)	157 (56.5)	122 (55.0)	0.734#	121 (54.3)	89 (53.9)	0.950#
Pain Duration, n (%)			0.562#			0.742#
< 12 weeks	149 (56.4)	120 (59.1)		118 (56.5)	85 (58.2)
> 12 weeks	115 (43.6)	83 (40.9)		91 (43.5)	61 (41.8)
NRS back pain, mean (SD)	6.4 (2.2)	6.1 (2.3)	0.135§	6.5 (2.2)	6.1 (2.3)	0.093§
NRS leg pain, mean (SD)	3.7 (3.3)	3.7 (3.3)	0.986§	3.8 (3.3)	3.4 (3.2)	0.277§
ODI, mean (SD)	30.8 (15.2)	31.3 (16.7)	0.714§	30.4 (15.0)	31.5 (17.1)	0.556§
EQ-5D index, mean (SD)	0.525 (0.303)	0.548 (0.299)	0.430§	0.532 (0.303)	0.557 (0.297)	0.451§
General health, mean (SD)	6.4 (2.1)	6.4 (2.1)	0.788§	6.4 (2.1)	6.5 (2.1)	0.707§
STB risk groups, n (%)			0.990#			0.914#
Low risk group	97 (36.7)	75 (36.9)		76 (36.4)	54 (37.0)
Medium risk group	132 (50.0)	102 (50.2)		104 (49.8)	74 (50.7)
High risk group	35 (13.3)	26 (12.8)		29 (13.9)	18 (12.3)

SD Standard deviation, n number of observations, NRS Numeric Rating Scale (0–10) higher score indicates higher pain intensity, SBT STarT Back Screening Tool, ODI Oswestry Disability Index (1–100) higher score indicates higher disability, EQ-5D index EuroQol (−0.594–1) higher score indicates better health, General health (0–10) higher score indicates better health. §Independent Student T-test, #Persons Chi-Square.

Clinical practice guidelines related indicator variables

CPG-related indicator variables with the proportion of adherence to CPG recommendations are presented in Table 6. The proportion of patients receiving referral to specialist consultation during the PT treatment period was very low in both groups with no statistically significant difference. However, the proportion of patients receiving medical imaging during the PT treatment period was significantly lower (OR 0.5; 95% CI 0.3 to 0.8; p = .011) in the intervention group (8.3%) compared to the control group (16.3%). The registry data showed a significant difference in the proportion of patients in the intervention group (54.7%) and control group (44.6%) receiving stratification of the number of PT visits based on the SBT (OR 1.5; 95% CI 1.0 to 2.3; p = .040). The proportion of patients receiving educational interventions was significantly higher (OR 5.1; 95% CI 3.3 to 8.0; p < .001) in the intervention group (73.5%) compared to in the control group (35.3%). When this analysis was sub-grouped according to the SBT, the proportional odds were largest for the medium/high risk SBT subgroup. The proportion of patients receiving an exercise intervention was similar in both groups with over 85% use in all LBP cases. The proportion of patients receiving manual therapy as a part of multimodal treatment with segmental movement impairments was 11.2% in the intervention group and 16.4% in the control group with no difference between groups (OR 0.7; 95% CI 0.4 to 1.2; p = .154). In only four cases, acupuncture was used as an adjunct intervention to exercise and educational interventions displaying no between-group differences. The proportion of patients receiving non-evidence-based treatments was significantly lower in the intervention group with 11.2% compared to 21.8% in control group (OR 0.4; 95% CI 0.2 to 0.8; p = .015).

Table 6. Strength of associations between interventions and adherence to CPG recommendations.

	CPG related indicator variables	Intervention group, n (%)	Control group, n (%)	Between-group comparison OR (95% CI); p-value; ICC
Assessment intervention	1. Proportion of patients receiving referral to specialist consultation Cohort with registry data	6/278 (2.2)	12/221 (5.4)	0.6 (0.3 to 1.4); p = .257; ICC = 0
	2. Proportion of patients receiving medical imaging Cohort with registry data	23/278(8.3)	36/221 (16.3)	0.5 (0.3 to 0.8); p = .011: ICC < 0.001
Treatment intervention	1. Proportion of patients receiving stratified number of PT visits Cohort with registry data	152/278 (54.7)	99/222 (44.6)	1.5 (1.0 to 2.3): p = .040; ICC = 0.010
	2. Proportion of patients receiving educational interventions Cohort with CRPE data	164/223 (73.5)	58/165 (35.2)	5.1 (3.3 to 8.0); p < .001; ICC = 0.148
	- SBT low risk subgroup	49/76 (64.5)	20/54 (37.0)	3.1 (1.1 to 8.6); p = .026; ICC = 0.091
	- SBT medium/high risk subgroup	107/133 (80.5)	31/92 (33.7)	8.1 (4.4 to 14.9); p < .001; ICC = 0.228
	3. Proportion of patients receiving exercise interventions Cohort with CRPE data	203/223 (91.0)	146/165 (88.5)	1.3 (0.7 to 2.6); p = .411; ICC < 0.001
	4. Proportion of patients receiving short-term use of manual therapy as a part of multimodal treatment when LBP with segmental movement impairments is found. Cohort with CRPE data	25/223 (11.2)	27/165 (16.4)	0.7 (0.4 to 1.2); p = .154; ICC < 0.001
	5. Proportion of patients receiving acupuncture as an adjunct to treatment interventions 2 & 3 Cohort with CRPE data	1/223 (0.4)	3/165 (1.8)	0.8 (0.3 to 2.4); p = .682; ICC < 0.001
	6. Proportion of patients receiving non-evidence-based treatments Cohort with CRPE data	25/223 (11.2)	36/165 (21.8)	0.4 (0.2 to 0.8); p = .015; ICC = 0.042

CPG Clinical Practice Guidelines, n number of observations, OR Odds ratio, CI Confidence Interval, CRPE data clinical reasoning and process evaluation data from physiotherapists, SBT STarT Back Screening Tool, ICC intracluster correlation coefficient, Bold text p < 0.05.

Clinical practice quality index

The proportion of clinical practice quality index fulfillment is presented in Table 7. Analyzing the registry data, the assessment quality index in both groups had a high proportion of fulfillment with 91.0% in the intervention group versus 82.9% in the control group with a statistically significant difference between groups (OR 2.0; 95% CI 1.2 to 3.4; p = .011). The proportion of fulfillment in the treatment quality index was significantly higher (OR 3.5; 95% CI 1.7 to 7.2; p < .001) for the intervention group (66.4%) compared to the control group (28.5%). The overall adherence to the clinical practice quality index was significantly higher in the intervention group, 58.7% versus 25.5% (OR 3.5; 95% CI 1.9 to 6.6; p < .001).

Table 7. The proportion of clinical practice quality index fulfillment.

	Number of criteria for adherence	Intervention group, n (%)	Control group, n (%)	Between group comparison OR (95% CI); p-value; ICC
Assessment quality index	Both criteria fulfilled Cohort with registry data	253/278 (91.0)	184/222 (82.9)	2.0 (1.2 to 3.4); p = .011; ICC < 0.001
Treatment quality index	All 3 criteria fulfilled Cohort with CRPE data	148/223 (66.4)	47/165 (28.5)	3.5 (1.7 to 7.2); p < .001; ICC = 0.182
Overall adherence	All 5 criteria fulfilled Cohort with CRPE data	131/223 (58.7)	42/165 (25.5)	3.5 (1.9 to 6.6); p < .001; ICC = 0.077

n number of observations, OR Odds ratio, CI Confidence Interval, CRPE data clinical reasoning and process evaluation data from physiotherapists, ICC intracluster correlation coefficient, Bold text p < 0.05.

Discussion

The observed between-group difference in implementation outcomes confirms our hypothesis that the multifaceted implementation of the BetterBack☺ MoC successfully improved adherence to CPG recommendations among PTs in primary care despite no statistical differences in the primary outcome referral to specialist consultation. These findings triangulate well with a qualitative sub-study, where patients in the BetterBack☺ MoC group experienced better knowledge about their LBP and received tools to better manage their health condition (Enthoven et al., 2021). Furthermore, other parallel sub-studies from the same research program indicate that improved quality of care can also translate to improved implementation processes and patient outcomes (Schröder et al., 2021, 2020). For example, improved practitioner confidence, attitudes and beliefs along with facilitating determinants of implementation behavior were observed after implementation of the BetterBack☺ MoC (Schröder et al., 2020). Moreover, patients receiving care from PTs trained in the BetterBack☺ MoC significantly improved in satisfaction with care and had a larger clinically meaningful improvement in health-related quality of life and illness perception compared to routine care (Schröder et al., 2021). However, there were no between-group differences in pain intensity, disability, patient enablement or global impression of change. Additional explorative analyze showed when PTs’ care was adherent to all five recommendations in the overall quality index, most patient reported outcomes improved compared to non-adherent care (Schröder et al., 2021).

The primary outcome regarding lower referral to specialist consultation did not show any significant difference between groups. This can be explained by the low number of referrals in both groups, leaving little room for improvement. Our study results suggest that these referrals are most likely from other HCPs such as general practitioners with little influence from PTs. Other literature suggests that patients with musculoskeletal disorders treated by PTs were less often referred to secondary care compared to patients treated by GPs (Goodwin and Hendrick, 2016). A study evaluating referral to specialist consultation within a period of 12 weeks in a general practice context showed a referral rate of 5% in a multifaceted implementation group compared to 10.5% in a passive implementation group (Riis et al., 2016). To change this outcome, an involvement of other HCPs who are gatekeepers for referrals may be needed (Briggs et al., 2016).

In the current study, a higher proportion of patient cases in the intervention group (58.7%) compared to routine care (25.5%) fulfilled all five criteria in the overall quality index. The overall quality index represents the quality indicator variables with the highest clinical priority rankings by the MoC support team in the development phase of which BetterBack☺ MoC was designed to have the strongest focus on and this may explain the large between-group differences (Abbott et al., 2018). In a review by Briggs et al. (2016) adherence to LBP CPGs were ranging from 30% to 40% in physiotherapy contexts. Most of the studies in this review had defined CPG adherence as the use of active treatment approaches (e.g. therapeutic exercise) and a limitation of passive treatments. Earlier research has found a small increase of 3.1–12% regarding adherence to LBP CPG after PTs have received CPG education (Bekkering et al., 2005a; Evans et al., 2010; Rutten et al., 2013; van Dulmen et al., 2014). In comparison, the present study had a larger increase (33.2%) of PTs’ overall adherence to LBP CPG after a multifaceted implementation of the BetterBack☺ MoC. However, varying ways of measuring overall adherence to CPG for LBP have been used in previous studies, making exact comparisons of our study results difficult.

In the present study, the CPG-related indicator variable with the largest difference between the intervention and control groups was the significantly larger proportion of patients receiving educational interventions in the intervention group (73.5%) versus (35.2%) in routine care. A tendency to stratified use of educational interventions was found where patients with low risk were 3.1 times more likely and patients with medium/high risk of persistent LBP were 8.1 times more likely to receive educational interventions compared to routine care. An increased use of patient educational interventions after CPG implementation has been shown by Bekkering et al. (2005b), but this study is to our knowledge the first to show stratification of patient educational interventions with a larger use in higher risk groups after implementation of a MoC. A recent qualitative review showed that PTs feel unprepared to identify and treat cognitive, psychological and social factors in LBP (Synnott et al., 2015). Furthermore, the authors recommended use of the SBT to identify these factors as well as education and additional training of the PTs to treat the multidimensional pain presentations seen in LBP. This is also supported by a review of five randomized controlled trials of PT delivered cognitive-behavioral interventions for LBP concluding that without training and resources, successful translation and implementation remain unlikely (Hall et al., 2018). To facilitate implementation of patient education interventions, the BetterBack☺ MoC contains support tools for pain education including a brochure and for the group-based care a supporting PowerPoint presentation with a detailed manuscript for the presenting PT (Supplementary File 1). Easy access to support tools for psychologically informed LBP physiotherapy has previously been lacking in Sweden, and lack of time has been shown as the number one barrier for Swedish PTs using guidelines (Bernhardsson et al., 2014). The CPG-related indicator variable with target behavior regarding higher use of exercise interventions had already very high adherence (>88%) in routine care with no significant changes after the implementation of the BetterBack☺ MoC. This is in line with another Swedish primary care study, which reported 82% use of exercise for LBP (Bernhardsson et al., 2015).

A successful implementation needs to include both increased use of evidence-based interventions and decreased use of non-evidence-based treatments. In this study, the use of non-evidence-based treatments was significantly higher in the routine care group, 21.8% compared to 11.2% use after implementation of the BetterBack☺ MoC. This indicates that the BetterBack☺ MoC significantly contributes to the de-implementation of non-evidence-based treatment.

In the current study, there was a significant difference in the proportion of patients receiving stratification of the number of PT visits. Two studies have evaluated a similar guideline recommendation with a maximum of three treatment sessions for patients with less than 3 weeks of LBP resulting in only 17–20% adherence to this recommendation (Bekkering et al., 2005a; Swinkels et al., 2005). Fritz, Cleland, and Brennan (2007) have shown similar number of visits with a mean of 5.2, but in a sample of only acute LBP (<90 days). Higher number of PT visits has been shown by others, wherein a large study involving 122 723 patients the mean was 7.1 visits (Childs et al., 2015). Consensus concerning the optimal number of visits is, however, difficult to establish and may require adaptation to the specific health-care system context.

The proportions of patients receiving manual therapy or acupuncture according to the guideline recommendations were low and similar in both groups, which may be explained by a more prominent focus on active strategies to improve patient’s self-management. Considering that the results represent CPG recommendations, which had the lowest clinical priority ranking compared to the other CPG recommendations, it may be less likely that changes would occur. The proportion of patients receiving medical imaging was significantly lower in the intervention group (8.3%) compared to routine care (16.3%). In Swedish primary care referrals to medical imaging are the responsibilities of general practitioners but potentially influenced by communications by PTs. The implementation may have influenced patients that routine imaging is not necessary during the PT treatment period.

The de-implementation of both assessment and treatment interventions identified “don’t do recommendations” in CPGs is as important as “do recommendations” to improve clinical practice. During decades of guideline recommendations with advice to decrease imaging for benign LBP, the prevalence remains constant with an imaging rate of 25% of LBP cases in primary care (Downie et al., 2020). Educational interventions have been suggested to reduce the proportion of patients that believe that routine imaging is necessary (Chou, Fu, Carrino, and Deyo, 2009). Previous literature shows a tendency to change from mainly passive strategies in the early 1990 to predominantly active interventions in the present (Groenendijk et al., 2007). This is also supported by the results of the current study where the active strategies exercise and educational interventions are the most used treatment interventions and passive strategies with acupuncture and manual therapy mobilization are less used when treating patients with LBP in primary care.

Strengths and limitations

This study with a strong randomized design is among the first to show improvement in physiotherapy guideline adherent treatment after implementation of a model of care. We contribute with well-defined quality indicators and clarity concerning previous recommendations already in line with existing guideline recommendations and those in need of a further focus. This study has several strengths including the robust but pragmatic study design and the large practitioner and patient samples. The stepped-wedge design with a dogleg cluster structure with two assessments in routine care allowed for an internal pilot phase in cluster 1 to confirm PTs acceptability of the intervention and monitor patient recruitment to provide information on the optimal cross forward time for cluster 2, resulting in two balanced groups. The transition from routine care to intervention did not have a washout period; however, PTs were advised to not introduce the BetterBack☺ MoC to patients that had been included during the routine care period. Self-reported treatment may be somewhat inaccurate due to recall bias and misconception, but the inaccuracy is likely to be similar across the groups. Most studies evaluating adherence to physiotherapy guidelines used clinical vignettes as a proxy measure of clinical practice (Maas et al., 2015; Rutten et al., 2013; Rutten, Kremers, Rutten, and Harting, 2009; van Dulmen et al., 2014). In the current study, we used data from the public health-care regional registry and data from PTs’ reported treatment choice for each patient, which offered authentic assessment material that better reflects clinical practice (Scholte et al., 2016). Another choice of primary outcome in the study context such as use of a locally adapted overall adherence may have been more relevant than referral to specialist consultation. The present study contained different practice sizes from 1 to 4 PTs up to large practices with more than 20 PTs, this mix with smaller and larger practices in both rural and urban areas was similar in the intervention and control groups. Patient characteristics were similar to those reported in other primary care settings (Bier et al., 2018) and an earlier study from the same region (Enthoven, Skargren, and Öberg, 2004). The result can therefore be generalizable for different practice sizes. Studies have showed that private and publicly financed PTs have different treatment orientations with a more biopsychosocial orientation for publicly financed PT care (Derghazarian and Simmonds, 2011). Since this study contained publicly financed health care, the results are more generalizable for PTs working in this context. About 5% of the Swedish population lives in the region of Östergötland and has income, educational and health levels that are comparable with the whole country (Folkhälsomyndigheten) making our findings generalizable to Swedish PT care in primary health care and to similar health-care systems internationally. But generalizability must be interpreted with caution since implementation success depends on a high degree of contextual factors (Moore et al., 2015). To enable a national and international spread, the rather resource intensive 2-day workshop used in the multifaceted approach could be adapted to new digital possibilities. However, important facets such as support from managers and clinical champions are less resource intensive and can be easily utilized to fit different health-care contexts (Schröder et al., 2020). Further development of a multi-professional MoC and testing in other contexts is needed to increase the generalizability.

Conclusion

After a multi-facetted implementation strategy of BetterBack☺ MoC, the primary outcome referral to specialist consultation during the PT treatment period was very low and with no differences compared to routine care. However, an improved assessment, treatment and overall adherence to evidence-based guidelines could be seen including both implementation and de-implementation behaviors. Adherence with recommendations concerning stratified number of visits improved, use of exercise was maintained high, patient educational interventions increased especially in medium/high risk patient groups and use of non-evidence-based recommendations decreased. A reduction of medical imaging during the physiotherapy treatment period was also observed. In summary, a top-down implementation strategy with manager support and a bottom-up strategy with clinician involvement in both the development and implementation phases seems to be effective in increasing guideline adherent PT care for LBP in primary care.

Acknowledgments

Thanks to the Swedish Research council, the Research council in Southeast Sweden and the Region of Östergötland for supporting this study. The authors wish to thank participating PTs for data provision and engaged PTs in the MoC support team and managers in the implementation forum. Thanks also to our collaboration partners in Odense Denmark for advice and input on the intervention materials. This work was supported by the Research Council in Southeast Sweden [grant number FORSS*660371], Swedish Research Council [grant number 2017*01444] and Region of Östergötland [grant number LIO-817751, LIO-922501, LIO-938197].

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed on the publisher’s website

Additional information

Funding

This work was supported by the Region of Östergötland [LIO-817751, LIO-938197, LIO-922501]; Research Council in Southest Sweden [660371]; Swedish Research Council [2017-01444].