Introduction

Regional musculoskeletal (MSK) disorders, such as back, neck and shoulder pain, are some of the most common and disabling health conditions internationally, leading to substantial personal, social and economic burden [1, 2]. The high costs of disability and lost productive work time associated with these conditions demand significant improvements in health care strategies, especially in those aimed at helping patients return to work. Systematic reviews of health care interventions, such as physical conditioning or pain management programs, targeted to regional MSK disorders have indicated modest effectiveness [3–6]. Furthermore, the response of individual patients to these interventions is highly variable. Some patients benefit greatly, while others do not improve, or even experience a worsening of problems [7]. Improved methods for selecting the optimal interventions for individual patients (i.e. personalized rehabilitation) would be invaluable.

Clinical decision support (CDS) is a term that has been used to define the myriad of ways in which knowledge is represented in health information and/or management systems to assist health care providers and other stakeholders in patient management decisions [8]. CDS tools are devices, instruments, questionnaires or other diverse resources (including algorithms, continuums of care, and treatment models) that present knowledge to health care decision-makers, and are often designed as point-of-care resources that support decisions regarding optimal treatment choices. Research and development of CDS tools is a rapidly growing field. These tools are attractive options, given the widespread adoption of computer tablets and smart phones. Also, CDS tools can be an efficient and time-saving strategy for busy clinicians if treatment algorithms are evidence-based and present minimal risks. This technology has the potential to augment complex decisions such as those performed for regional MSK disorders. Computerized CDS has the potential to significantly improve human decisions by expediting information retrieval, identifying unique patient needs, triaging care, and matching patients to the most appropriate resources and treatments.

Some promising CDS tools have been developed specifically for use with patients that have regional pain disorders [9–11]. However, the effectiveness, utility and feasibility of CDS resources in the treatment of regional MSK disorders has been under investigated [12]. Previous systematic reviews of CDS tools have focused on the evaluation of medical management and included only randomized controlled trials from the health care literature [12–17]. However, CDS for the treatment of patients with regional MSK disorders is an emerging area that covers multiple disciplines (including health care, computing science, occupational health services and human resource management). The current literature is therefore diverse and fragmented [11, 18, 19] using inconsistent terminologies and methods. However, to date, no thorough synthesis and summary of these methods is available. In addition, the state of the science in terms of effectiveness, utility, and feasibility of CDS resources in the treatment of MSK disorders has not been summarized as a whole. Given the diversity of the literature and emerging nature of the field, a comprehensive scoping review is needed to map the scientific and grey literature on this topic [20].

The purpose of this project was therefore to conduct a scoping review of CDS tools designed to help decision-makers select interventions that are specifically intended to improve function and return to work in patients with pain-related MSK disorders. This review was also open to other patient related outcomes such as pain, and disability. Our study aims were to identify and inventory CDS tools for helping front-line staff select interventions. We considered both computer-based CDS and other available tools such as treatment algorithms, care pathways, prediction rules, and models. In addition, we aimed to summarize key concepts and terminology to provide criteria for future reporting, evaluate and synthesize evidence of the effectiveness and utility of the available tools, and recommend directions for future research and development in this area.

Methods

Design

This study was a scoping review, which is a methodology for rigorously collecting, synthesizing, appraising and presenting findings from existing research on a topic [20–22]. This approach is especially relevant when an area is emerging or diverse because it examines the extent, range and nature of the research activity [23]. Generally scoping reviews are referred to as ‘a mapping process’ since they summarize a range of evidence in order to convey the breadth and depth of a field [24]. Unlike systematic reviews, scoping reviews do not require appraisal of the quality of the included studies. However, the scoping process requires an analytical interpretation and inventory of the available literature. A scoping review is also useful for determining whether enough literature is available on a topic to conduct a formal systematic review or a meta-analysis or to identify gaps in the literature. In addition, scoping reviews can include a range of study designs and address complex and diverse questions that cannot typically be addressed with a systematic review. Our research area is both emerging and diverse. For these reasons, we chose to conduct a scoping review.

We adopted the scoping review framework proposed by Arksey and O’Malley [23]. This framework progresses through five stages: (1) identifying the research question; (2) identifying relevant studies; (3) selecting studies for analysis; (4) charting the data; and (5) collating, summarizing and reporting results. Each stage will be discussed in detail below.

Identifying the Research Question

An iterative process was used in which we reflexively adapted our question, search terms, and strategy to ensure comprehensive coverage of the literature [23]. An initial question provided the scope for the review and contained several key concepts that guided the search terms used. However, the question was refined based on the broad spectrum of articles we obtained in the initial search. Initially we had included CDS tools for selecting interventions as well as making diagnoses and prognoses. However, due to the extensive breadth of the literature obtained and impracticality of reviewing all 3 research domains, we decided to focus on intervention tools only. This decision was made after consultation with all the researchers and knowledge users involved.

Our final research question was the following: “Do validated decision support tools (especially computer-based tools) exist for selecting appropriate interventions for improving function and return to work in patients with pain-related MSK disorders?”

Identifying Relevant Studies

Relevant studies were identified through online searches of health care, computing science and management databases. These searches were performed with the assistance of two experienced research librarians at the University of Alberta who had access to, and a thorough knowledge of all the necessary databases and search engines. Databases searched included Ovid MEDLINE, Ovid EMBASE, Scopus, CINAHL, Business Source Complete, ABI/INFORM Global, Social Science Research Network (SSRN), Web of Science, ACM Digital Library, IEEE Xplore, ACM Computing Reviews, Computing Research Repository (CoRR), NECI ResearchIndex (formerly CiteSeer) and Google Scholar. Our search strategies were adapted to the various databases as required with the assistance of the librarians. The search included all articles in all languages since the inception of the databases.

Keywords included musculoskeletal diseases; musculoskeletal disorders; back pain; neck pain; shoulder pain; disability evaluation; vocational rehabilitation; return to work; decision support techniques; decision support tools; decision making; clinical protocols; computer-assisted. An example of a search strategy performed in Medline is presented in “Appendix 1”.

Grey literature (unpublished documents from outside the peer-reviewed scientific literature) were also searched. We applied the Canadian Agency for Drugs and Technologies in Health’s Grey Matters search tool to search for relevant information and websites [25]. In addition, Google was searched to identify possible unpublished studies. Relevant articles from the study teams’ own research or libraries were also included.

Each CDS tool located was tracked in the Scopus database and Google Scholar to determine whether additional studies investigating the tool had been published.

Selecting Studies for Analysis

The following were the final set of inclusion/exclusion criteria for the review:

  • Topic of the article A CDS tool for selecting interventions.

  • Population Patients with any painful MSK disorder (e.g., regional pain disorders of the back, neck, knee, shoulder, etc.). Our review included all MSK conditions available in the literature and all terms referring to MSK conditions were included in the searches. We excluded articles on non-MSK disorders including metabolic/endocrine disorders (i.e. osteoporosis, diabetic ulcers), rheumatic disorders (i.e. ankylosing spondylitis, rheumatoid arthritis, fibromyalgia) and other general medical conditions.

  • Outcome Functional and work-related outcomes, including return to work, disability, performance, and absenteeism. Functional recovery is a crucial outcome in regional pain disorders. From the perspective of the various stakeholders involved (i.e. workers, insurers, employers and health care providers), recovery from pain is important; however, functional recovery—such that the patient can return to work and participate in normal daily living—is equally important and has important career and quality-of-life implications [26, 27]. Functional recovery is also often easier to measure. For these reasons, we focused primarily on interventions aimed at improving function or facilitating return to work and other activities of daily living.

  • Study type Any design describing or evaluating a CDS tool. Systematic reviews were excluded but references within those located were searched for further articles.

The titles and abstracts of articles obtained from the online databases were reviewed and appraised for relevance. Two independent researchers from the team read each title/abstract and judged whether they were relevant to the research question. When there were disagreements between reviewers, the principal researcher (DPG) offered additional consultation until a decision could be reached. If the relevance of a study was still unclear, then the full article was obtained. After selecting the relevant abstracts and titles, two independent researchers assessed the corresponding full versions of the studies to determine which articles should be included in the full review. If discipline-specific questions arose, the reviewers consulted with the team member with relevant expertise (i.e. computers, health care, human resource management, etc.) who could answer the question. We used a Microsoft Access (Redmond, Washington) database stored on an internal server at the University of Alberta that was securely accessible by team members for all stages of the review.

Consultation with Knowledge Users

The consultation process for this study incorporated the development of an advisory committee that included knowledge users who were representatives of local rehabilitation clinics in the Edmonton area, national networks of health care clinics that provide rehabilitation to injured workers, and experienced compensation case managers. We held meetings with knowledge users at two key stages of the review: selecting studies for analysis, and summarizing and reporting results. Knowledge users were asked whether they knew of any CDS tools currently in use or relevant articles. Feedback from knowledge users during these consultative meetings highlighted the importance of: (1) including functional and return-to-work outcomes as search terms; (2) considering not only papers describing specific CDS tools, but also theoretical or conceptual papers dealing with models or algorithms describing treatment selection approaches for patients with MSK disorders; (3) having reviewers consider workplace-based interventions (i.e. accommodations, modifications, etc.) and work-related outcomes (i.e. productivity, absenteeism, etc.) during abstract and title screening; and lastly, (4) considering the importance of feasibility, time of tool administration, cost, and ease of interpretation in addition to scientific validation when considering the utility of any CDS tools located. Before charting the data, the knowledge users were consulted to determine whether the number of articles selected was appropriate and whether the search terms should be altered.

Data Analysis

Charting the Data

Reviewers extracted relevant information from the articles and entered it into an electronic data chart created with the Microsoft Access program. This form included data for authors, year of publication, article title, discipline of the lead authors, geographic location of the study, type and brief description of the CDS tool (including a list of factors included in the tool’s algorithm such as age, sex, pain level, etc.), cost of the tool, study population, study design and goals, methods used, outcome measures used, important results and any economic data recorded. For computer-based tools, we extracted additional information using categories taken from a previously published CDS taxonomy [28]. These charting methods provided a standard and systematic approach to summarize the papers and extract all relevant information.

Collating, Summarizing and Reporting Results

During this stage, we created an overview of all research located. Initially, we presented a basic numerical summary of the studies, including the extent, nature and distribution of the articles. Then, we summarized articles according to the types of tools described or evaluated, research methods used, populations studied, and study results/outcomes.

As mentioned earlier, the scoping review methodology was intended to summarize both the breadth and depth of the literature. We reported the number of articles for each CDS tool as well as some descriptive information about the articles. Since this was a scoping review, we did not undertake a critical appraisal of quality. However, we attempted to map the diversity observed and inventory the various study designs and methods used. This procedure allowed us to draw conclusions about the nature of research in this area and provide recommendations for future studies.

Several clinical prediction rules were designed to identify those individuals likely to respond positively to a particular treatment or intervention. These types of tools have been summarized in other reviews [29, 30], but we created an updated table to establish the range of tools in this category and to examine the strengths and limitations of these rules.

The various CDS tools identified in the articles were also categorized, and key concepts and terminology used in the articles were summarized in tables.

Guidelines developed by Terwee et al. [31] were used to define quality of measurement properties of the CDS tools. Briefly, quality of measurement included internal (internal consistency, relevance of items and representativeness of items of the scale-content validity) as well as external components of validity (the relationship with other tests in a manner that is consistent with theoretically derived hypotheses-construct validity). Intra and inter-rater reliability (i.e. repeatability of measurements taken by the same tester at different times and repeatability of measurements taken by different testers, respectively) were also considered. Definitions of psychometric properties for this review are provided in “Appendix 2”.

Results

The initial search considering all online databases identified 4605 potentially relevant articles. From these, 189 unique studies were included for the second stage; screening full texts. After screening full texts, 133 articles were selected. From these, 34 articles were removed since they were systematic reviews or considered irrelevant for the purposes of the study. However, their references were searched. From the reference search of these studies and stakeholder meetings, we obtained 24 additional articles. Thus, 123 studies were included for data extraction. The reliability of the screening process of titles and abstracts was high with an average agreement percentage of 92.3 % between reviewers. Figure 1 shows the flow chart of our article search and relevance selection process. A search of the grey literature obtained no new documents or websites specific to MSK disorders.

Fig. 1
figure 1

Flow chart of article search and relevance selection process

Full size image

General Description of Articles

Of the 123 relevant articles located, most originated in North America (n = 70, 57 %), were published between 2006 and 2014 (n = 101, 82 %), and discussed a clinical prediction rule or a classification system (n = 79, 64 %). Twenty-one articles (17 %) discussed treatment algorithms or models, 15 (12 %) discussed questionnaires, while only 8 (7 %) discussed computer-based tools. Most of the articles pertained to low back pain (n = 69, 56 %), followed by neck, shoulder or arm pain (n = 21, 17 %), and general MSK disorders (n = 17, 14 %). When the article described an original study (n = 75), designs of these studies varied greatly from randomized controlled trials (RCTs) to case series and reports. The majority of the original studies were observational in nature, most commonly cohort studies (n = 31, 41 %). Table 1 displays more details about the characteristics of located studies.

Table 1 Descriptive characteristics of included articles (n = 123)
Full size table

Overall, there were 43 CDS tools located. After reviewing the tools and identified articles, these were classified into three main areas: (1) specific computer-based tools or questionnaires (n = 8, 19 %); (2) clinical prediction rules/classification systems aimed at categorizing patients into various treatment groups (n = 21, 49 %); and (3) theoretical or algorithmic approaches to selecting treatments (treatment algorithms/models) (n = 14, 33 %). Each of these areas and the tools located will be described.

Computer-Based Tools/Questionnaires

Table 2 provides an inventory of the 8 computer-based devices or questionnaires located for selecting interventions for patients with pain-related MSK disorders. Table 3 provides a summary of the original studies evaluating these tools. Twenty-two manuscripts including three theses [32–53] looked at these 8 tools. Three questionnaire-based tools were included: Keele STarT Back Screening Tool (SBST), the Pain Recovery Inventory of Concerns and Expectations (PRICE) questionnaire, and the Orebro Musculoskeletal Pain Questionnaire (OMPQ). We also located 5 tools incorporating computer technology: Repetitive Strain Injury (RSI) QuickScan intervention program, the Pain Management Advisor (PMA), the Decision Support System (DSS) for helping ergonomists better match workers with the work environment, the Soft Tissue Injury Continuum of Care Model with computerized prompts for case managers, and the Work Assessment Triage Tool (WATT). Three of these tools were aimed at workers with low back pain (SBST, PRICE and OMPQ), 2 were aimed at work-related upper extremity disorders (RSI QuickScan and DSS), 2 were aimed at general work-related MSK injuries (WATT and Soft Tissue Continuum of Care), and 1 aimed at assisting physicians in managing patients with chronic pain (PMA).

Table 2 Inventory of computer-based tools and questionnaires located
Full size table
Table 3 Summary table of original studies evaluating computer-based tools or questionnaires for selecting interventions for patients with musculoskeletal disorders
Full size table

Most of these tools had some psychometric testing in the way of validation of items, concurrent validity, acceptability of the tool, accuracy of the classification as well as testing the effectiveness of the tool compared to standard treatment. Nevertheless, this testing has been preliminary, and a more exhaustive validation process involving randomized controlled trials at multiple sites and settings is needed for all of the tools. Four of the tools [34, 35, 37, 38] showed positive preliminary results regarding the use of the tool to determine appropriate treatments for managing some MSK conditions. However, one tool (RSI QuickScan) did not prove to be effective for reducing work disability or cost-effective [40, 52], and two studies did not have clear positive or negative results regarding the tools evaluated (SBST and WATT) [33, 42]. One student thesis evaluated the utility of the OMPQ as a clinical decision support tool for workers’ compensation claimants, with negative results [39]. However, the OMPQ was initially developed as a screening or prognostic tool, not explicitly as a CDS tool. In addition, one protocol of a cluster RCT that attempts to use the OMPQ as a CDS tool for selecting interventions for patients with LBP was found [44]. The results of this RCT are still unpublished, so it is unknown how well the OMPQ functions as a CDS tool. Three other studies [32, 36, 41] only looked at the development phase of the tools (WATT and PRICE). Thus, evidence is limited regarding validity evidence of these CDS tools. For details of the measurement properties of the CDS tools found, see Table 4.

Table 4 Summary of the quality of measurement properties of the computer-based tools or questionnaires located
Full size table

Treatment Algorithms/Decision-Models

Of the 22 articles [54–75] discussing treatment algorithms/models, there were 15 original studies evaluating 14 different algorithms or decision-models (theoretical or empirical) for selecting interventions for patients with MSK disorders. Details of these algorithms/models and the studies can be found in Table 5. Nine of the studies [54–61, 76] looked at low back pain, 2 discussed knee disorders [62, 63], 2 discussed shoulder disorders [64, 65] and 2 examined other body regions (wrist and neck) [66, 67]. Research designs used in these studies varied greatly, with the observational cohort study being the most common among them. Methodologies and types of algorithms were also wide-ranging, making the analysis of these studies challenging. Most of the algorithms or decision models were developed to determine possible treatment paths. Nevertheless, most testing of these algorithms/decision models was preliminary or exploratory (e.g., small sample size, secondary analyses of previous collected data, lack of replication or validation samples, use of research designs that are prone to bias including case series, cross sectional or cohort studies rather than randomized controlled trials). Nine of these algorithms [54, 58, 62–67, 76] seemed to lead to positive results when deciding on intervention strategies. Nevertheless, one study [56] found that the use of the algorithm would not result in better outcomes for patients with low back pain. Four studies [55, 57, 59, 61] only looked at the development of an algorithm/model without associated testing of it.

Table 5 Summary table of original studies describing or evaluating algorithms or decision-models (theoretical or empirical) for selecting interventions for patients with musculoskeletal disorders
Full size table

Clinical Prediction Rules/Classification Systems

The remaining 79 articles [76–154] looked at 21 unique clinical prediction rules or classification systems. Four articles described the clinical prediction rules in general. Most of these studies targeted rules for identifying responders to interventions for low back pain (8 rules) followed by neck pain (6 rules), patellofemoral pain (3 rules), lateral epicondylagia (2 rules), ankle sprain (1 rule) and thoracolumbar injury (1 rule). For details on the clinical prediction rules and classification systems found in this scoping review, see Table 6. The rules were developed to determine response to specific treatments that included spinal manipulation, stabilization exercises, McKenzie approach, mechanical traction, Pilates-based exercise, foot orthoses, patellar taping, or general classification models.

Table 6 Summary of clinical prediction rules and classification systems for painful musculoskeletal conditions
Full size table

From the rules looking at back pain (8 rules in total involving 47 articles [77–121, 150, 152]), three rules (rules for manipulation and stabilization exercise, and the treatment-based classification system) have been the most commonly studied. Confirmatory evaluation of these rules has shown mixed or unsuccessful results. The remaining 5 rules (rules for the McKenzie approach, mechanical traction, Pilates, and the CBI Health classification system) have been developed empirically or theoretically but no confirmatory testing has been conducted. Thus, it is unknown if the results from these studies would provide clarification regarding management of patients with back pain.

Six rules discussed in 18 articles targeted neck pain [122–139]. From these rules, only one (treatment-based classification system) showed positive results when tested in case series, pilot and cohort studies. However, this rule has not been tested in a randomized controlled trial. The remaining 5 rules for neck pain were either unsuccessful (rule for thoracic manipulation) or had no confirmatory testing evidence. From the rules developed for patellofemoral pain, 2 rules discussed in 5 articles [140–143, 146] (1 rule for patellar taping and 1 for foot orthoses) were not tested further and 1 rule for lumbopelvic manipulation obtained unsuccessful results when tested in a separate sample. The remaining rules developed for lateral epicondylalgia (2 rules in 2 articles [147, 149]), ankle sprain (1 rule in 1 article [148]) and thoracolumbar injury (1 rule in 2 articles [144, 145]) did not have additional testing. Of note, there were 3 interventions where two separate rules were generated for the same condition (traction for low back pain, manipulation for neck pain, and foot orthoses for patellofemoral pain), but results indicated the rules were not consistent and the rules were formed of different variables.

Discussion

The number of CDS tools relevant to MSK disorders is small but it appears that this field is rapidly expanding. Results of this scoping review identify that although there are several publications around CDS tools, with the majority (82 %) published since 2006, few correspond to formal and validated tools to help with the management of MSK conditions. Furthermore, the tools, models and classification systems we identified are intended for use by health care providers. One tool, the RSI Quickscan, is intended for use by ergonomists for identifying appropriate management strategies for workers with upper extremity disorders, including job or equipment modifications where appropriate. However, we were unable to locate any decision support systems for human resource managers or other employer agents who develop return-to-work processes and identify appropriate job modifications.

There was a wide range of literature including treatment algorithms/decision-models and several publications related to clinical prediction rules applied in the context of MSK disorders, most commonly low back pain. The included articles were rather diverse and most of this information was exploratory or developmental in nature, particularly with regard to use of research designs that are prone to bias including case series, cross sectional or cohort studies rather than randomized controlled trials, secondary analyses of previous collected data, and lack of replication or testing in validation samples. It appears that research in this area is starting to develop and would benefit from an internationally coordinated effort. Consequently, more studies regarding feasibility, usability, and effectiveness of these tools as well as psychometric testing would benefit the area of CDS tools applied to health care specifically to the area of MSK disorders.

Computer-Based Tools or Questionnaires

Our review located 3 questionnaires and 5 computer-based tools that were used to select interventions for patients with MSK disorders. Most of these tools were at initial stages of development or validation. However, we were not able to locate or get further information from the authors of 2 tools (DSS and PMA) indicating the developers likely did not pursue further development. The majority of the studies we reviewed were non-experimental in design, focusing on early stages of questionnaire development and testing or focused mainly on process measures, such as clinician ratings of system acceptability and usability. Of the located tools, six had some validity evidence (WATT, SBST, OMPQ, RSI QuickScan, PMA). The tool that appears to have been most evaluated has been the SBST. This tool has been translated into several languages and has demonstrated good discriminative validity when compared with widely accepted questionnaires such as the Roland Morris Disability Questionnaire and Tampa Scale of Kinesiophobia, among others (AUC ranged from 0.79 to 0.91 [155], and 0.75–0.89 [156]). Although this information is promising, this tool has not been examined through a clinical trial outside the United Kingdom. Thus, the validation studies for these tools overall have not provided strong evidence for use of these tools in clinical or workplace settings. Of note, the OMPQ was not explicitly developed as a CDS tool but as a screening/prognostic tool, which may explain the negative results in a validation study [39]. Thus, none of them are ready for widespread implementation in clinical practice since more testing is necessary.

Since the research obtained regarding CDS tools in MSK disorders is at the early stages, information about user preferences regarding the presentation of computer output, including content, formatting (e.g., color, graphics), and length, have not been conducted to date. Similarly, there are no published data concerning technical difficulties (e.g., type and number of system crashes or touch-screen calibration problems) encountered by computer-based CDS tool users. Both issues have important implications for future system refinements and implementation strategies. In addition, there is a lack of information regarding contextual circumstances or the processes used to integrate the CDS into the existing clinical workflow, as well as testing in different populations and settings. Most of the studies found have tested the CDS tools in one single group of patients. In addition, some limitations of the existing CDS tools for treating MSK conditions were lack of integration with computer and/or mobile devices, the reduced use of web-based interfaces, and infrequent use of data directly entered by patients. Some of the tools were even questionnaires administered by paper and pencil, which was also highlighted by the recent review performed by Pombo et al. [157].

Research of the effectiveness of CDS tools to improve patient outcomes is still fairly sparse. Only 3 of these tools (SBST, RSI QuickScan, Continuum of Care) have tested patient outcomes such as patients’ recovery, disability, cost, and quality of life. Results from these studies are inconsistent, and more replication with variable settings and population sampling strategies is needed. Other major patient outcomes of interest for policy makers have not been examined, such as health care utilization, health care costs, and communication with health care providers. Similar results have been obtained in early systematic reviews of computerized decision-support systems for chronic pain management in primary care and CDS tools targeted to healthcare professionals, especially for medical conditions [158, 159].

Further validation of these tools with larger samples and with stronger designs are needed. It is necessary that larger randomized controlled trials testing the effectiveness of CDS tools against standard care be performed to determine clearly if these systems are worth being implemented in clinical practice.

Clinical Prediction Rules/Classification Systems

Clinical Prediction Rules and classification systems that aim to identify which patients would benefit from a specific treatment have attracted the attention of many researchers regarding their effectiveness and validity. Several narrative and two systematic reviews have been conducted [29, 30]. Our scoping review adds to this literature by attempting to inventory all clinical prediction rules developed for a wide variety of MSK conditions and comment on the status of the research in this area. We located 21 clinical prediction rules that have been developed for MSK disorders, however studies evaluating effectiveness of these rules have been inconsistent. Most of the rules lack external validation in different samples using strong methods such as RCTs, but validation studies that have been conducted by separate research groups have largely been unsuccessful. We also found that rules developed for the same treatment for the same condition by different research groups were inconsistent in terms of the clinical variables in the final rules. These results are not surprising based on the results obtained from different systematic reviews focusing on rules in low back pain and the physical therapy area [29, 30, 160]. According to Beneciuk et al. [29] there are several clinical prediction rules in physical therapy that have not been validated in external samples. In addition, recently Patel et al. [161] examined the quality of the validation studies for clinical prediction rules in subjects with back pain. They found that the evidence from randomized trials validating rules for non-specific back pain is weak. These results were also in agreement with those of May et al. [30] Haskin et al. [160] and Patel et al. [161] Thus, based on the current evidence, more widespread use of clinical prediction rules for identifying responders to various interventions in clinical practice is not recommended at this point.

If clinical prediction rules are well designed and validated in appropriate populations, they could have the potential to identify patients most likely to benefit from a particular treatment. This in turn would help improve clinical decision-making and practice. However, the current evidence, especially the lack of cross-validation and replication, does not support large-scale implementation of clinical prediction rules to improve disability outcomes [160]. At present, it is unknown if the unsatisfactory performance of rules in clinical trials is because inappropriate rules have been tested, the trials have been poorly designed, underpowered, or simply that it is impossible to develop rules that are fit for all conditions, subjects and settings [29]. Thus more research is needed to elucidate all of these questions.

Treatment Algorithms/Decision-Models

The literature around treatment algorithms and models was diverse, which made the analysis of these studies challenging. Most of these algorithms or decision models have been developed for determining an appropriate treatment path without formal and rigorous testing. Sample sizes have been relatively small in most cases. The results from these studies also are inconsistent. Thus, no clear conclusions extracted from these algorithms or models can be made at this point.

Strengths and Limitations

This study represents the first attempt to inventory available CDS tools for MSK disorders, and comment on the status of the research literature. Strengths of our project include the involvement of a large international group or researchers and stakeholders with diverse backgrounds who provided input on the project. Additionally, we conducted a very comprehensive literature search (all languages and years since inception of databases) across health, computer science, and management databases with the assistance of research librarians as well as a search of grey literature using validated methods. The methodology used in this project was that of a scoping review, which summarizes the state of the science in a given area, but does not synthesize evidence on specific outcomes (e.g., patient outcomes, cost-effectiveness) across studies. This represents a limitation of the scoping review methodology, but it was appropriate in this case due to the diversity of methods and literature encompassed by the review. Also, scoping review methods do not require detailed critical appraisal and, therefore, study quality likely varied in the articles we identified. Additionally, while we sought to be as comprehensive as possible in our literature search, it is possible that there are other CDS tools under development that we failed to identify. As the various CDS tools are tested in different settings and using consistent methodology, more definitive conclusions about the impact of these tools on clinicians’ performance or patients outcomes may be drawn.

Conclusions

The potential for CDS tools, especially those employing rapidly advancing computer technologies, has sparked great interest among health care providers, case management organizations and funders of care. Our literature review identified 5 computer-based tools, 3 questionnaires, 14 algorithms or decision-models, as well as 21 clinical prediction rules or classification systems. However, currently none of these tools, models or systems appears ready for widespread use in clinical practice to select interventions for patients with MSK disorders. More research is needed examining more advanced levels of validity of existing tools, including impact on patient outcome, or developing new evidence-based CDS tools to help guide clinical and workplace practice for managing patients with MSK disorders.