Patient Education Booklet to Support Evidence- Based Low Back Pain Care in Primary Care – A Cluster Randomized Controlled Trial


 Background

Inappropriate imaging and low-value care for low back pain (LBP) are common. A new patient-education booklet was created to overcome identified barriers to the delivery of recommended care, including the use of inappropriate imaging. Our aim was to assess the effectiveness of this booklet as part of primary care for LBP patients in comparison to usual care.
Methods

A cluster-randomized trial was performed. The intervention involved providing practitioners with the new patient-education booklet and a 30-minute training session on its use. The booklet was provided during the clinical consult to all consenting LBP patients in the intervention group. Primary outcomes were the proportion of patients presenting with LBP who underwent imaging examinations during the first three months of follow-up and PROMIS PF-20 (Patient-Reported Outcomes Measurement Information System, 20-item physical functioning short form) change between baseline and three-month follow-up. Secondary outcomes, including sick leave and imaging examinations at 12 months, were investigated. Logistic regression using GEE-estimation was used for dichotomous outcomes, Poisson regression using GEE-estimation for count outcomes, and linear mixed models for continuous outcomes.
Results

Using the patient education booklet appeared to substantially reduce the proportion of LBP patients who underwent an imaging examination at three months, but the result was not statistically significant (OR 0.57, 95% confidence interval (Cl) 0.27 to 1.22). At 12 months, the effect was slightly larger and statistically significant (OR 0.50, 95%Cl 0.30 to 0.83, p = 0.008). No difference was observed in the PROMIS PF-20 T-score change between baseline and 3 months or 12 months (p = 0.365 and p = 0.923, respectively). The number of sick leave days in the intervention group was less than that in the control group at 3 months (RR 0.47, 95%Cl 0.26 to 0.83, p = 0.010) and at 12 months (RR 0.36, 95%Cl 0.18 to 0.72, p = 0.004).
Conclusions

The booklet appeared to be effective in reducing the proportion of LBP patients who underwent imaging examinations over 12 months. The intervention had no discernible effect on the PROMIS PF20 T-score change. The number of sick leave days was substantially lower in the intervention group.
Trial registration

ISRCTN, ISRCTN14389368, Registered 4 April 2019 - Retrospectively registered


Results
Using the patient education booklet appeared to substantially reduce the proportion of LBP patients who underwent an imaging examination at three months, but the result was not statistically signi cant (OR 0.57, 95% con dence interval (Cl) 0.27 to 1.22). At 12 months, the effect was slightly larger and statistically signi cant (OR 0.50, 95%Cl 0.30 to 0.83, p = 0.008). No difference was observed in the PROMIS PF-20 T-score change between baseline and 3 months or 12 months (p = 0.365 and p = 0.923, respectively). The number of sick leave days in the intervention group was less than that in the control group at 3 months (RR 0.47, 95%Cl 0.26 to 0.83, p = 0.010) and at 12 months (RR 0.36, 95%Cl 0.18 to 0.72, p = 0.004).

Conclusions
The booklet appeared to be effective in reducing the proportion of LBP patients who underwent imaging examinations over 12 months. The intervention had no discernible effect on the PROMIS PF20 T-score change. The number of sick leave days was substantially lower in the intervention group.
Trial registration ISRCTN, ISRCTN14389368, Registered 4 April 2019 -Retrospectively registered Background Low back pain (LBP) has been estimated as the leading cause of global years lived with disability in many countries and is one of the most costly health problems worldwide ,, . Management of LBP is often inconsistent with guidelines, and low-value care, such as inappropriate imaging, is common , . For example, one third to one half of LBP patients undergo inappropriate imaging, which has been associated with increased health care costs, increased downstream health care utilization and increased disability ,,, . Increased compliance with guidelines has shown to reduce health care costs and worker's compensation claims.
The implementation of clinical guidelines for LBP is complicated and has several physician-and patient-reported barriers , .
Australian researchers have developed a new patient education booklet which focuses on both physician-and patient-related barriers and: 1) provides patient and practitioner education; 2) reminds practitioners of recommended care; 3) provides clinical decision support; and 4) facilitates practitioner-patient communication. The booklet has been translated into and validated in Finnish. Preliminary evaluation by patients and practitioners in Finland has suggested that the booklet may be helpful in LBP management and in decreasing the need for LBP imaging.

Trial objectives
The primary objective of this cluster randomized controlled trial was to assess the effectiveness of the "Understanding low back pain" patient-education booklet in addition to usual care in comparison to usual care alone, in reducing the proportion of patients presenting with LBP who undergo imaging examinations due to LBP over the rst three months of follow-up (individual participant-level data), and to improve physical functioning at three-month follow-up (individual participant-level data). We used cluster randomization to minimize contamination between the intervention and control groups.
The secondary objective was to assess the effectiveness of the intervention in reducing the proportion of patients presenting with LBP who undergo imaging examinations due to LBP, LBP-related sick leave days, health care appointments and disability, and to improve physical functioning and quality of life over 3-and 12-month follow-ups (individual participantlevel data).

Methods
We used a cluster-randomized trial design to assess the effectiveness of using the patient education booklet with LBP patients in primary care in comparison to usual care. Ethics approval was granted by the Ethics Committee of the University Hospital of Oulu. The study was reported in accordance with the CONSORT 2010 checklist with extension for Cluster Trials.
Trial design and participants Figure 1 presents a ow chart of the trial design and participants. Cluster randomization was performed at the health care region level rather than at the practitioner level, because patients would be evaluated by different health care professionals in the same health care region. We selected six public health care regions and two occupational health service organizations for clusters. The clusters were chosen and matched according to population size and type of health care provider (public or occupational) (Fig. 1). The managers of all the participating centres in the clusters were informed of the study and consented to participate before pairwise randomization into control or intervention groups was performed by an external statistician using a random number generator. Blinding was impossible because of the study design. All the physicians and physiotherapists in the study clusters were invited to participate in the study and recruit eligible patients for the study.
In all the centres (in both the intervention and control groups), we provided the professionals (physiotherapists, nurses and physicians) with a 30-minute training session, which covered the eligibility criteria and the overall study procedures (see Powerpoint slide show, Appendix 1). The physicians, physiotherapists and nurses identi ed eligible patients when they attended appointments for LBP. The professional explained the study to the patient and requested their consent. The professionals then sent the signed patient-consent forms to the research assistant (outside the centre). Whether a patient initially presented to a physiotherapist, physician or nurse did not change because of the study process. In Finland, LBP patients may initially present to a: 1) physician; 2) physiotherapist; 3) nurse together with a physician; or 4) nurse alone (the rst contacted professional), depending on local care policies, the education level of the professionals, the availability of appointments at the time, and the patient's symptoms (usually determined by the triage nurse over the phone). The rst professional each patient contacted and patient baseline information was identi ed from electronic patient record data. The research assistant sent repeated information and a reminder about the study to the physicians, physiotherapists and nurses every month. The email contained a patient information sheet, brief information on patient recruitment, and details on eligibility criteria.
The patients who were eligible for inclusion were aged 18 to 65 and presented for treatment for LBP with or without associated radicular pain, that had lasted for two weeks or less (low risk of prolonged pain) for the rst time. The exclusion criteria included a suspected serious cause of LBP (for example, cancer, fracture or infection); or LBP requiring urgent (emergency) care, such as cauda equina syndrome. Only patients who signed a consent form were included as participants in the study and patients were free to discontinue their participation in the study at any time without a reason.

Intervention
The main focus of the intervention was the use of the patient education booklet in addition to usual care as part of the appointment with the professional (individual patient-level intervention). The professionals (physicians, physiotherapists and nurses) in the intervention arm attended a 30-minute group training session on the booklet's content and how it should be used during appointments (see Powerpoint slide show, Appendix 2) (cluster-level intervention). The training sessions were held during the weekly clinic meetings to enhance participation. Smaller units could alternatively participate online. After the education session, the professionals were encouraged to use the booklet (printed Finnish version) in the manner that they preferred (e.g., to guide the whole interaction with the patient, or to emphasize key points at the end of the consultation, etc.) during the LBP patients' appointments. The English version of the booklet is available online at https://tinyurl.com/lowbackpaineducation. Other aspects of care were provided as usual. All the consenting LBP patients in the intervention health care units received the booklet from their physician, physiotherapist or nurse.

Control
The professionals in the control arm received a short training session on the study and recruitment procedure only. All the consenting LBP patients presenting to the control clinics received usual care without the booklet.

Outcomes
The two primary outcomes were as follows: 1) The proportion of patients presenting with LBP who underwent imaging examinations due to LBP during the rst three months of follow-up; and 2) Change in PROMIS PF-20 T-score (Patient-Reported Outcomes Measurement Information System, 20-item physical functioning short form) from baseline to threemonth follow-up (individual patient level) , . Imaging data were collected from electronic patient record data to determine the proportion of patients presenting with LBP who had undergone imaging examinations due to LBP (yes/no). The patientreported outcomes were collected using web-based questionnaires.
The secondary outcomes were determined from the electronic patient record data on the basis of the following: the proportion of patients presenting with LBP who had undergone imaging examinations due to LBP over 12 months of followup; imaging examinations strati ed by the type of imaging [radiographs/ magnetic resonance imaging (MRI)/computer tomography (CT)] over 3-and 12-month follow-ups; the number of LBP-related sick leave days; and the number of LBPrelated health care professional appointments (primary care: physician, physiotherapist, nurse; secondary health: physiatrist and orthopaedist) during the 3-and 12-month follow-up.
The secondary patient reported outcomes were: change in Oswestry Disability Index (ODI) between baseline and 12-month follow-up; change in Roland Morris disability questionnaire between baseline and 12-month follow-up; change in PROMIS (Patient-Reported Outcomes Measurement Information System) (short form 20a) between baseline and 12-month follow-up; change in LBP intensity (0-10 Numerical Rating Scale (NRS)) during past week between baseline and 3-and 12-month follow-ups; change in leg pain intensity (0-10 NRS) during past week between baseline and 3-and 12-month follow-ups; and change in EQ-5D (EuroQol ve dimensions) between baseline and 12-month follow-ups and Frequency of LBP during last three months at baseline, 3-and 12-month follow-ups (daily LBP yes/no).

Electronic patient record data
In each health care region or organization of the study, all the primary health care professionals (physicians, physiotherapist, nurses, etc.) use the same electronic patient record data. The data are organized according to professions (e.g., one le for primary health care physicians, another le for physiotherapists, etc.). Imaging examinations are documented on separate les and contain lists of completed examinations and imaging reports. All health care professionals have access to both primary and secondary electronic health care data directly or through passcode-protected access to the nationwide database. The input of health data is obligatory for health care providers. Nationwide electronic patient record data are accessible to all health care providers in Finland. The data were collected manually from the organizational database by the rst author or one of two research assistants, using a paper checklist for each patient (Appendix 3). Initially, to ensure congruency, the checklist data for 10 patients were extracted independently by two people and compared. Electronic patient record data were collected from the date of informed signed consent up to the 3-and 12-month follow-ups.
Patient questionnaire data Patient-reported data were collected using online questionnaires sent to the patients via an email hyperlink. The baseline questionnaire was collected after the patient consented to participate and after the rst appointment; the follow-up questionnaire was collected 3 months and 12 months after the baseline questionnaire. If the patients did not complete the baseline and follow-up questionnaires within one week, the research assistant resent the hyperlink via email. If an email address was missing, the research assistant sent a paper version of the questionnaire. Further reminders were sent at two weeks and three weeks to the patients who had not responded to the baseline, 3-month or 12-month questionnaires by a text message containing a hyperlink to the related questionnaire.

Changes to the protocol
Changes were made to the original protocol because of unreliable data relating to one of the original secondary outcome measures (pain medication) and in order to respect the original purpose of the STarT Back Tool (SBT). We did not include the secondary outcome of pain medication as we could not accurately extract these data from the electronic records, and we only used SBT to describe baseline differences, not as a secondary outcome measure.

Sample size
At baseline, the imaging rate (radiographs, MRI or CT) was expected to be 30%. We used a web-based calculator to compare the proportion with a dichotomous outcome (imaging/nor imaging) of the two samples (http://www.samplesize.net/sample-size-proportions/). We estimated the Design Effect (DE) for unequal clusters in a cluster-randomized design to increase the statistical power of the cluster randomized study: DEunequal = 1+[(1 + CV 2 )x m -1]ρ; when ρ(ICC) = 0.01; coe cient of variation of cluster size (CV) was 0.208; CV = sd/m, (sd = standard deviation of mean cluster size; m = mean cluster size); sd = CSrange/4 (CS = cluster size) 21, . A sample size of 408 patients would enable the detection of an absolute 15% decrease in imaging proportion with 80% power and a type I error of 0.05.
The sample of 408 could provide more than adequate power for the other primary outcome of PROMIS based on the sample size a web-based calculator for Inference for Means in clustered samples: http://www.sample-size.net/means-samplesizeclustered/. The Minimal Important Difference (MID) for change in PROMIS PF-20 was about two points, and standard deviation (SD) was 3.66 . The type I error rate was 0.05, and the Intracluster Correlation Coe cient (ICC) 0.025. A sample size of 128 patients would enable the detection of a difference of two points in PROMIS PF-20 with 80% power.

Statistical methods
We analysed the between-group differences at baseline using independent-samples t-tests for normally distributed continuous variables, the Mann-Whitney U-tests for non-normally distributed continuous variables, and the Chi-square test or Fisher's exact tests for dichotomous variables.
The statistical methods used to analyse the differences between the intervention and control groups were selected according to the type of the outcome. We used logistic regression for dichotomous outcomes (imaging and LBP frequency), providing an estimated difference between the groups expressed as odds ratio (OR) with a 95% con dence interval (CI). Poisson regression was used for count outcomes (health care professional appointments and sick leave days), providing an estimated difference between the groups expressed as risk ratio (RR) with a 95% con dence interval. Logistic and Poisson models were analysed using generalized estimating equations (GEE) with an exchangeable working correlation to account for the clustered nature of the data. We analysed the continuous outcomes (PROMIS, ODI, NRS for pain, EQ-5D index and EQ VAS) using a linear mixed model with random effects for clusters, providing an estimated least squares mean difference with a 95% con dence interval. Age-adjusted analyses were also conducted for all outcomes because of baseline differences.
Post-hoc sensitivity analyses were conducted for the primary outcomes based on the rst contacted professional, as this was unbalanced between the groups and there was strong rationale for why the intervention may be more or less effective depending on the profession of the rst professional contacted. In the Finnish health care system, physiotherapists cannot directly refer patients for imaging, so which professional the patient rst contacted was particularly important for our imaging outcome. Statistical analyses were conducted using IBM SPSS Statistics for Windows, version 26 (IBM Corp., Armonk, NY). Two-sided tests were used, and p-values less than 0.05 were considered statistically signi cant.

Baseline characteristics
The number of professionals who attended the training and study information sessions and participated in recruiting the intervention group were: In total, 418 LBP patients consented to participate in the study. Recruitment was carried out between 13th April 2017 and 30th May 2020. The number of participants in each cluster, the missing data in the follow-up questionnaires, and the missing data in the electronic patient records data are shown in Fig. 1. The patients in the intervention and control groups had similar characteristics at baseline (Tables 1 and 2). Approximately half of the patients were actively working. The patients' mean age was 41.4 years in the intervention group and 44.6 years in the control group (p = 0.011). No differences were found between the PROMIS T-score of the intervention and control groups at baseline. However, the rst contacted professional who used the booklet during the appointment differed signi cantly (p = 0.001; Table 2). In the intervention group, the rst contacted professional was a physician for 44% and a physiotherapist for 43% of the patients, in comparison to 28% and 62% in the control group, respectively. Only a few patients met both a physician and a nurse (12% in the intervention and 9% in the control group). The rst contacting professional was seldom a nurse alone (0% in the intervention and 1% in the control group).  Contact with other professional before and after study consent given. Intervention results

Primary outcomes
The use of the patient education booklet appeared to substantially reduce the proportion of LBP patients who underwent an imaging examination at three months, but the result was not statistically signi cant (OR 0.57, 95% con dence intervals (Cl) 0.27 to 1.22, p = 0.147) ( Table 3). The proportion of patients presenting with LBP who had undergone imaging examinations due to LBP over three months was relatively low in both groups: 10.8% in the intervention and 16.7% in the control group.  1 Difference between intervention and control groups analysed with logistic regression using generailzed estimating equations with exchangeable working correlation matrix. 2 Difference between intervention and control groups analysed with Poisson regression using general estimating equations with exchangeable working correlation matrix.

*OR not possible due to zero frequency
The between-group difference in PROMIS PF-20 T-score change between baseline and three-month follow-up was not statistically signi cant (mean difference 1.0, 95% CI -1.5 to 3.5, p = 0.365) ( Table 4). The intraclass correlation of PROMIS tscore change at three months was 0.00005. Change 3m calculated from baseline to 3-month follow-up and change 12m from baseline to 12-month follow-up.

Secondary outcomes
The proportion of patients who underwent any imaging over the 12-month follow-up was lower in the intervention group. The imaging rate over 12 months was 17.5% in the intervention and 29.7% in the control group (OR 0.50, 95% Cl 0.30 to 0.83, p = 0.008). Similar results were found when radiographs and MRI were analysed separately (at 12 months radiographs (OR 0.55, 95% Cl 0.34 to 0.88, p = 0.013) and MRI (OR 0.61, 95% Cl 0.44 to 0.84, p = 0.002)). The results were similar in the ageadjusted analyses ( Table 3).
The mean number of sick leave days was 3.8 days in the intervention group and 8.4 days in the control group for the rst three months, and 7.5 and 20.8 days, respectively, for the whole 12-month period. The differences in sick leave days were statistically signi cant over the 3-month (RR 0.47, 95% CI 0.26 to 0.83, p = 0.010) and 12-month follow-ups (RR 0.36, 95% Cl 0.18 to 0.72, p = 0.004; Table 3).
There were no statistically signi cant differences between the groups in the number of health care appointments (Table 3) or patient-reported secondary outcomes: physical functioning (PROMIS T-score), back or leg pain intensity, disability (ODI), quality of life (EQ-5D), self-rated health (EQ VAS) or frequency of daily LBP (Table 4).

Sensitivity analyses
Sensitivity analyses were conducted on the basis of the rst contacted professional (e.g., physician or physiotherapist), because of the large baseline differences and the likelihood that the effects of the intervention would be different across the different professionals (Table 5). When the rst contacted professional was a physician, 7% of the patients underwent imaging examinations over the rst three months in the intervention group, in comparison with 29% in the control group (OR 0.15, 95% CI 0.09 to 0.24, p < 0.001). Over 12 months, the proportion of imaging use was 15.7% in the intervention group and 42% in the control group (OR 0.23, 95% CI 0.16 to 0.33, p < 0.001). The intervention had no effect on PROMIS PF-20 change between baseline and 3-month or 12-month follow-up. In addition, when the rst contacted professional was a physician, the intervention reduced sick leave days to one third at both 3 months (RR 0.35, 95% Cl 0.20 to 0.64, p = 0.001) and 12 months (RR 0.34, 95% Cl 0.23 to 0.51, p < 0.001). The risk of further physician appointments decreased by one third at both 3 months (RR 0.65, 95% Cl 0.55 to 0.75, p < 0.001) and 12 months (RR 0.60, 95% Cl 0.49 to 0.72, p < 0.001). The number of secondary health care appointments was lower at both 3 months (RR 0.79, 95% Cl 0.75 to 0.82, p < 0.001) and 12 months (RR 0.91, 95% Cl 0.88 to 0.95, p < 0.001). There were no between-group differences in other secondary outcomes among the patients who rst saw a physiotherapist.
The combination of a physician and nurse as the rst contacted professional was uncommon (11% (n = 23) in the intervention group and 8% (n = 17) in the control group), and occurred in only two of the eight clusters. The sensitivity analyses of the effect of using the patient education booklet during an LBP patient appointment within this subgroup is presented in Appendix 4.

Discussion
Key ndings: The use of the patient education booklet to support evidence-based care of LBP patients in primary care appeared to substantially reduce the proportion of imaging at three months; however, this was only statistically signi cant when a physician was the rst contacted professional. The impact of using the education booklet on imaging rates was substantial at 12 months, with a reduction in imaging proportions of 50% in the intervention group compared to the control group. In the other primary outcome of change in PROMIS T-score, we found no difference between the intervention and control groups. In the secondary outcome of sick leave days, we observed a signi cant reduction both at 3 months (RR 0.47 95% CI 0.26 to 0.83, p = 0.010) and 12 months (RR 0.36 95% CI 0.18 to 0.72, p = 0.004).
Radiographic imaging rates were low in both the intervention and control groups (radiograph rate 8% in the intervention and 15% in the control group, MRI rate 13% and 21%, respectively) compared to a previous systematic review, in which 31% of patients underwent radiograph imaging due to LBP in primary care over one year, whereas 16% received MRI. This difference may be explained by the profession of the rst contacted professional. When the rst contacted professional was a physician, the radiographic imaging rate over 12 months was 8% in the intervention group and 24% in the control group. In contrast, when physiotherapists were the rst contacted professional, radiograph imaging use was 9% in the intervention group and 12% in the control group. In the Finnish health care system, physiotherapists cannot directly refer patients for imaging, which likely explains the lower imaging proportions and the stronger reduction in imaging in the sub-group of patients who contacted a physician rst (OR 0.15, 95% CI 0.09 to 0.24). The total (including all imaging modalities) imaging rate was lower in both groups at three months (11% in the intervention and 17% in the control group) than we expected in sample-size calculations (30%). This reduced the statistical power to detect signi cant differences at that time point. The results at three months indicated a trend towards less imaging in the intervention group. The effect sizes at 3 months and 12 months are very similar, despite differences in signi cance due to lower power at three months.
Comparison with previous literature: To our knowledge, the effectiveness of an intervention developed to address both physician-and patient-related barriers to reducing the use of imaging has not previously been tested. Previous studies have demonstrated reduced imaging examinations with the inclusion of epidemiologic data in lumbar spine MRI reports ,, . In our study, information on the usefulness of imaging was provided to all the patients and professionals, regardless of whether or not imaging was undertaken.
Interesting ndings from secondary outcomes include the reduction in sick leave days at 3-and 12-month follow-ups.
According to Ree et al. (2016), a workplace educational back pain intervention also reduced sick leave for up to six months. In addition, in an occupational setting, booklet information alone was cost-effective in comparison to no intervention among patients with mild LBP. Decreasing imaging may also enhance reductions in sick leaves. Previous research showed that decreased imaging for LBP was associated with decreased health care utilization and sick leave days , . In contrast to imaging, where an effect was observed only when a physician was the rst contact, sick leave days also decreased when a physiotherapist was the rst contacted professional. This is likely explained by physiotherapists' ability to grant short-term (up to ve days) sick leave due to LBP in Finland.

Interpretation of ndings
These results are consistent with the ndings of the preliminary evaluation: the intervention reminded the professionals of the imaging guidelines and made it easier to follow them, and patient-related barriers were addressed by helping patients understand and receive an explanation for their pain 15 . It is possible that the shared-decision making and the written information supported decisions related to imaging use, and sustained, or even strengthened, the results over the 12-month follow-up. Common physician-reported reasons for using imaging include feeling that there is no alternative to imaging to offer the patient, and the lack of time to have a conversation with patients about diagnosis and why a scan is not needed. The booklet provides practitioners with an alternative way to discuss the use of imaging with patients. It was easy to carry out, made sense to both the professionals and the patients, and supported evidence-based care.
The use of the patient education booklet in addition to usual care had no apparent effect on patients' pain, physical functioning, disability or quality of life in comparison to usual care. This may have needed a more intensive and probably a more individualized intervention.

Implications of the ndings
The booklet appeared effective in reducing imaging and sick leave days among LBP patients in primary care and is suitable for implementation in similar primary care settings for this purpose. The patient education booklet is easy to use, inexpensive and seems to have no negative side effects, which endorses its implementation in practice. More research is needed to evaluate its effectiveness in different health care settings.

Strengths and limitations
One strength of this study is that we used a cluster randomized design, and assessed the effectiveness of the booklet in an everyday clinical environment in the Finnish health care system. Several elements enhanced the generalization of the results and the possibility to use the intervention in different primary care organizations in future studies: the low number of exclusion criteria for LBP patients; inviting all physicians and physiotherapists in the study health care regions or organizations to participate; easy implementation procedure; low costs of the intervention; and minimal or no suspected harm related to the intervention.
A limitation of the current study was that some patients saw a physician rst and others saw a physiotherapist rst. We therefore conducted sensitivity analyses, which found a stronger effect on imaging rates when the booklet was used by a physician. In this subgroup, it additionally reduced subsequent LBP-related physician appointments in primary care by one third and LBP-related secondary health care appointments by about 10%. A systematic review and meta-analysis has earlier shown education for patients with acute or subacute LBP to reduce health care use (NNT 17 to prevent one primary health care physician visit due to LBP).
The low number of clusters (eight) is also a limitation of the study. The health care policies of the clusters might have differed, even though we used suitable statistical methods for cluster randomized studies. For example, physicians' sickleave practices have shown to vary signi cantly. In addition, no information was recorded on the LBP patients who declined to participate in the study, or on whether the professionals neglected to invite some eligible patients to the study, both of which could have introduced selection bias. The loss to follow-up for the patient-reported outcomes is a major limitation in estimating PROMIS PF-20 change between baseline and three-month follow-up, one of the two primary outcomes, due to its potential to bias the results of this study. Loss to follow-up for the other primary outcome, imaging at three months, was relatively low.

Conclusions
The booklet appeared effective in reducing the proportion of LBP patients who undergo imaging examinations when delivered by a physician. The intervention appeared to have no effect on PROMIS PF20 T-score change. The number of sick leave days appeared substantially lower in the intervention group.

Declarations
Ethics approvaland consent to participate All protocols was carried out in accordance with the Declaration of Helsinki and the study was approved by the Ethics Committee of the University Hospital of Oulu, and all participants provided informed written consent. Personal information was kept in secured repositories and replaced with ID codes in all analyses. ASS, JK and research assistants have access to the nal trial dataset using a secured username and password.

Consent for publication
The manuscript does not include any individual person's data.