Do psychological factors affect outcomes in musculoskeletal shoulder disorders? A systematic review

Background Psychological factors may impact recovery in patients undergoing treatment for shoulder complaints. The aim of this review is to systematically analyse the evidence for the effect of modifiable psychological factors (MPF) on outcome, for patients with musculoskeletal shoulder disorders undergoing conservative or surgical treatment. MPF refers to factors that may change with intervention. Methods This is a systematic literature review. Five databases searched (MEDLINE, CINAHL, Cochrane Library, Embase and PsycInfo), for longitudinal studies investigating the influence of MPF on prognosis of patients with shoulder disorders, all diagnoses, undergoing clinical interventions (conservative or surgical). Level of evidence was determined using Scottish Intercollegiate Guidelines Network (SIGN) methodology. Moderate and high quality evidence was included. We extracted all MPF, categorized constructs into the following domains: beliefs (self-efficacy, expectation of recovery), coping (catastrophizing, avoidant coping), and affect (depression, anxiety). We evaluated constructs for its predictive value of at least one outcome. Outcomes were informed by this review. Evidence was classified into three categories: evidence for, inconclusive evidence, and evidence against. Results Of 1170 references, 40 distinct publications based on 35 datasets were included (intervention type: 20 surgical; 20 conservative). Overall, 22 studies (20 cohort studies and 2 RCTs) were classified as high quality and 18 studies (16 cohort studies, 2 RCTs) were classified as moderate quality. Outcomes reported included pain, disability/function, perceived recovery, physical and mental health, and work status. Based on the review, of the psychological constructs explored, these data would suggest that expectation of recovery, catastrophizing, avoidant coping, depression, and anxiety may predict outcome for patients managed surgically. In patients undergoing conservative intervention the evidence was either against (catastrophizing, depression, anxiety) or inconclusive (self-efficacy, expectation of recovery, avoidant coping) for the predictive value of psychological factors on outcome. Conclusions Five constructs were predictive of outcome for surgically managed patients. This suggests that implementing the biopsychosocial approach (i.e., preoperative screening, intervention by a trained clinician) may be advantageous for patients recommended for shoulder surgery,,. The same is not indicated for conservatively managed patients as no conclusive association of MPF with outcomes was noted. The importance of other MPF on outcome requires further investigation.


Introduction
Background Shoulder conditions are the third most common musculoskeletal complaint [1,2]. Only 50 % of patients with a new episode of shoulder pain experience complete recovery within 6 months and pain persists in 40% for more than 1 year [3]. In those who seek care, there is limited understanding of how to identify patients who may or may not respond to interventions [4]. Therefore, we need to understand barriers to and facilitators of recovery in patients with shoulder pain.
To improve treatment outcomes for shoulder complaints, modifiable factors that influence the prognosis should be identified. The focus of this review is on psychological factors. Modifiable psychological factors (MPF) are patient cognitions and emotions associated with health conditions that may impact recovery, and may respond to treatment [4,5]. Exploring the relationship between MPF and outcome is valuable, as effective management may improve outcomes [6,7]. MPF are different than psychological traits and refractory psychiatric diagnoses that are more difficult to manage, such as bipolar disorder and pervasive depressive disorder, and not considered in this review. Some MPF have been recognized as impacting recovery in other musculoskeletal conditions [1,2,[8][9][10][11]. Maladaptive pain beliefs, negative affective reactions and poor coping are indicators of psychological distress that may influence both the short and long-term outcomes of treatments in patients with spine, hip and knee conditions [5,[12][13][14].. Conversely, self-efficacy and positive expectation of recovery are coping resources that have been associated with better functional outcomes in patients with musculoskeletal disorders [10,11]. Kendall and Burton propose that in the absence of red flags suggestive of an emergent medical situation, all musculoskeletal conditions that limit activity may be treated like low back pain [15]. This treatment would include advice for self-care, education on expectation of a good recovery and instruction to continue with usual activity as tolerated. Despite compelling evidence to monitor and address MPF in patients with spine pain as part of routine care evidence to monitor and address MPF in patients with spine pain as part of routine clinical care [16,17], to date there is equivocal evidence to support the importance of MPF in MSD [11,[18][19][20][21][22]. As such, these factors typically are not part of routine clinical evaluation and treatment for patients with MSD [23,24].
Recent reviews explored psychological factors in various patient groups, including those receiving conservative and surgical care [18,25], conservatively managed patients only [19,20], patients with selective diagnoses, [21,[25][26][27][28], patients undergoing arthroplasty [29] or with conditions associated with chronic shoulder pain [11,25]. The heterogeneity of these diagnoses makes it difficult to compare the conclusions. In addition, methodologic limitations and variability of previous reviews was also noted [11,18,22,29]. Therefore, [11,25] the current reviews provide a limited perspective on the relationship between MPF and outcomes in patients with musculoskeletal shoulder disorders (MSD).
The aim of this literature review was to systematically summarize the current evidence on the importance of MPF on outcome in patients receiving care (conservative or surgical) for MSD. The MPF that may be found to be associated with outcome in MSD includes patient beliefs, coping and affect. Unlike previous systematic reviews that focused on some MPF and did not subcategorize studies based on intervention, our aim was to capture studies on all MPF in surgical and conservative studies to better identify those that predict outcomes. This review included all phases of shoulder disorders (acute, subacute, chronic) and all MPF referenced in the reviewed studies, to gain insights regarding the relationship between MPF and MSD. literature for all psychological factors found to be associated with shoulder pain and disability/function, and focused on those considered to be modifiable [31]. An updated search was conducted in December 2020.
The search was conducted with the help of a research librarian (MG). Two detailed search strategies are depicted in Appendix 1.
To ensure the completeness of the literature search, one reviewer (MW) conducted an electronic hand search of the four most often-retrieved journals and added all potentially eligible references not retrieved by the systematic search. In addition, two reviewers (MW, EB) examined bibliographies of included studies and review articles related to the research question, and relevant references were considered for full-text review (inclusion and exclusion criteria applied). We further searched clinical trials.gov for additional trials relevant to the topic and searched the grey literature after consulting with experts in the field. In potentially relevant studies with insufficient details for data extraction, we contacted the study authors for additional information.

Inclusion and exclusion criteria
Included were all longitudinal studies (cohort studies, randomized controlled trials (RCT), and studies on registry data) investigating patients with shoulder complaints undergoing conservative or surgical treatment for the shoulder disorder. Studies were eligible when they included the influence of MPF on the prognosis or treatment outcome. Excluded were experimental studies (i.e., identification of genetic markers) in which clinical interventions were not used to modify outcome (i.e., pain, function), crosssectional studies, case series, epidemiological studies, and studies on patients younger than 18 years of age. Studies of personality traits and psychiatric conditions were excluded. Although we did not specifically exclude studies on joint arthroplasty, the search was not set up to identify all studies on total shoulder joint replacement. Therefore, excluded studies on joint arthroplasty for the current review.

Data collection and abstraction
Two reviewers (MW and ERB), a physician and a physical therapist with extensive clinical and research experience, screened all references independently by title and abstract. Disagreements were discussed and resolved by consensus or by third-party arbitration (SSW), a physical therapist. For any study where questions arose regarding psychological constructs or outcome measures, a psychologist (SW, co-author) was consulted. References with insufficient information in the title or abstract to assess eligibility, were included in the full text review. All full texts were then appraised by both reviewers independently (MW and ERB) for inclusion or exclusion. Alternative researchers with specific language proficiencies were used for non-English language references, with no language restrictions. In the case of several publications for the same cohort without change in outcome or follow-up duration, the most recent publication was chosen and missing information from the previous publication was added. Systematized criteria were defined to extract specific variables from each reference and were followed by each reviewer. All information needed to describe the study population and methodology was collected: study setting, study design, number of patients, age, proportion of women, intervention, and follow-up duration. In addition, the methods of assessment and information on the type of analysis of the prognostic, predictive or mediating factors were extracted. The inclusion/exclusion criteria guided this process.

Assessment of study quality
A quality rating was assigned based on the risk of bias, using the Scottish Intercollegiate Guidelines Network (SIGN) methodology checklist for cohort studies and randomized clinical trials and the overall quality was rated as high, moderate, or low [32]. The ratings were as follows: high quality (++), most (≥60%) of the criteria fulfilled; moderate quality (+), some criteria fulfilled (< 60%); and low quality (−), few or no criteria fulfilled. Two reviewers (MW and EBR) assessed each reference. Any discrepancies were resolved by another member of the research team (SSW). High and moderate quality studies were included in this review.

Definition of terms
For this study, MPF are defined as those factors that may be expected to change with appropriate therapeutic intervention and are therefore states rather than traits. We utilized a framework of psychological domains [16] and modifiable constructs extracted from the included studies (Table 1) in order to synthesize the findings. It is important to note that there is no gold standard for the definition and classification of MPF. Therefore, for those constructs that may fall into more than one domain, we sought the guidance of a clinical mental health expert to inform the distinct classification based on the context in which the constructs were considered in the studies. This allowed for the classification of all constructs within one domain.
The term prognostic factor is used to describe a MPF that influences or predicts the course or outcome of a shoulder disorder. The prognostic value of a psychological factor is based on the reported results and conclusions of the primary studies. No predefined outcomes were identified for this review. Study outcome was extracted from each included reference based on the reported measure of assessment.
We classified studies based on patients' duration of pain as subacute (< 12 weeks), chronic (> 12 weeks) or a mixed duration of shoulder complaints.

Classification of evidence
All included studies were grouped based on the MPF addressed, time from onset and clinical intervention (conservative, surgical). We evaluated each construct based on the number of studies that reported it as a predictor of at least one outcome or not a predictor of any outcome. Outcomes were purposefully not predefined, as our objective was to identify all outcomes that have been included in studies on MPF in patients with MSD. If the number of studies with results showing that a construct was predictive of outcome was greater than the number of studies showing it was not predictive, we considered the construct predictive. If the opposite was true, then we considered the construct to not be a predictor of outcome. In those cases where an equal number of studies found evidence for and against the predictive value of the construct, the evidence was found inconclusive. Based on these criteria, the evidence was classified into three categories: Category 1) Evidence fora majority of the studies found the construct to be a predictor of outcome; Category 2) Inconclusive evidence -An equal number of studies found evidence for and against the predictive value of the construct, Category 3) Evidence against-a majority of studies did not find the construct to be a predictor of outcome.

Study selection
In the initial search 1140 references were screened, and 121 full-text articles assessed for eligibility. After excluding 86 publications, 35 publications based on 33 patient data sets were included for data extraction and analysis, hereafter referred to as 35 studies. The main reasons for exclusion were mixed patient populations without reporting specific results for subjects with shoulder complains (n = 31) and studies that did not assess MPF (n = 26, Fig. 1). In the updated search conducted on December 20, 2020 we identified 138 additional references. After title and abstract screen, an additional 19 references were read in full text. Finally, we included 5 additional publications (2 additional publications of previously included studies and 3 publications from 2 additional studies). In total, the narrative analysis reflects our review of 40 distinct publications based on 35 patient data sets, hereafter referred to as 40 studies.

Baseline characteristics
Of the 40 included studies, four were randomized clinical trials. There were 20 studies on conservative intervention and 20 on surgical intervention. Follow-up duration ranged from end-of-treatment to 12 months. The studies represented a broad spectrum of shoulder diagnoses, representative of a typical clinical population ( Table 2).

Study quality
Risk of bias in 40 studies was assessed using the SIGN method (Appendix 2A). In all tables, highquality studies included in this manuscript (Appendix 2A) are indicated by bold typeface. Twenty cohort studies were rated as high quality and 16 studies rated as moderate quality. Two randomized clinical trials were rated as high quality and two were rated as moderate. Overall, 20 (50%) of included studies were rated as high quality, 14 studies related to conservative care, and 6 studies related to surgical intervention. Most studies did not provide a formal sample size calculation. Six (30%) of the conservative studies reported a required sample size and met the requirement. Five (25%) surgical studies reported a required sample size; three studies met the required sample size, and two studies did not (150 instead of 360 patients, same data set for both studies).  Tables 3 and 4.

Clinical intervention and time from onset Conservative intervention
Among the 20 studies on conservative intervention, four addressed patients with subacute MSD, five addressed patients with chronic MSD and 11 did not specify time from onset or presented a mixed population. All six MPF were investigated (Table 5).

Surgical intervention
Among the 20 studies on surgical intervention, one addressed patients with subacute MSD, five addressed      ----Distress (HSCL-25) and self-efficiency for pain (single item question) not associated with SPADI and shoulder complaint as measured by Global Assessment Score at 6 weeks.
Engebretsen et al.
2010 [36] (++) Pain and disability (SPADI) Work status ( a ADI) ----Self-efficacy was significant in the univariate analysis but not in the final model for disability and not significant for return to work. Distress (Hopkins Symptoms Checklist) was not significant in the univariate analysis.
Engebretsen et al.

Modifiable psychological domains and constructs
In this sample, the domains of "coping" and "affect" were most investigated, 14 (40%) publications and 29 (73%) publications respectively, and the domain of "beliefs" was least investigated, 9 (23%) publications (Tables 3, 4 and 5). Of the six predefined constructs, depression (Domain: Affect) was the most studied construct, 27 (68%) publications, and self-efficacy (Domain: Beliefs), the least studied, two publication (5%). For surgical care, we found evidence for catastrophizing, avoidant coping, depression, anxiety, and expectation of recovery as predictors of outcome. In patients undergoing conservative intervention the evidence was either against (catastrophizing, depression, anxiety) or inconclusive (self-efficacy, expectation of recovery, avoidant coping) for the predictive value of psychological factors on outcome. The following provides details of the prognostic value of each MPF in patients with shoulder problems managed conservatively or surgically.

Domain: coping Catastrophizing
Catastrophizing as a predictor of outcome was explored in ten publications (five surgical, five conservative). In seven publications (five (100%) surgical [two high quality], two (40%) conservative [two high quality]) catastrophizing predicted at least one outcome. Therefore, based on this review, catastrophizing in surgical cases fell into Category 1, evidence for, while for conservative cases it was Category 3, evidence against. Domain: beliefs Self-efficacy Self-efficacy as a predictor of outcome was explored in two publications (two conservative [two high quality]). In one publication (50%) self-efficacy predicted at least one outcome. Therefore, based on this review, evidence for self-efficacy as a predictor in conservative cases fell into Category 2, inconclusive evidence.

Expectation of recovery
Expectation of recovery as a predictor of outcome was explored in eight publications (two surgical, six conservative). In five publications (two (100%) surgical [one high quality], three (50%) conservatives [three high quality]) expectation of recovery predicted at least one outcome. Therefore, based on this review, evidence for expectation of recovery as a predictor in surgical cases fell into Category 1, evidence for, while for conservative cases it was Category 2, inconclusive evidence.

Discussion
In this study we explored the relationship between MPF and outcomes in patients with shoulder disorders, within the context of management (conservative, surgical) and temporal framework (time from onset).
The main finding of this review is that psychological factors affect recovery in patients with shoulder pain managed surgically. However, MPF was not associated with outcome in patients receiving conservative care for shoulder disorders, regardless of duration of pain. This suggests that type of clinical management and time from onset are critically important variables to consider when defining the prognostic value of MPF on outcome in patients with MSD.

Previous systematic reviews
Previous systematic reviews [11,18,22] have explored the association between MPF and outcome in patients with shoulder conditions including those receiving conservative and surgical care [18,25], conservatively managed patients only [19,20], patients with selective diagnoses, [21,[25][26][27][28], patients undergoing arthroplasty [29] or with conditions associated with chronic shoulder pain [11,25]. However, they did not account for confounding factors that may impact this relationship, such as the approach to management (conservative, surgical) and time from onset. In addition, these reviews explored this topic through a narrow lens considering only several psychological factors or specific diagnoses. Therefore, previous reviews provide a limited perspective on the relationship between MPF and patients with shoulder conditions. The question that was addressed in this review was broad and included all reported diagnoses, time from onset, approaches to management, and did not predefine MPF or outcome. We classified studies based on conservative and surgical intervention and all diagnostic phases from acute through chronic. Furthermore, we did not predefine psychological factors or outcomes but rather extracted from the reviewed studies. In addition, we applied no language or publication timeframe restrictions in our search allowing for a broad body of literature from which this topic could be explored. Defining and focusing specifically on psychological factors that are modifiable is relevant as these factors are responsive to short-term intervention, as opposed to more refractory psychiatric diagnoses that are more difficult to manage [72]. For these reasons, the findings of this review may be clinically relevant in that they may guide the approach to preoperative care.

Evidence supporting MPF
In this review six distinct MPF were identified. However, for most of these factors, few studies have explored their relationship with outcome and not all were graded as high quality. For the purpose of this review, our conclusion regarding the effect of each factor on outcome was based on the preponderance of the included references. However, it should be noted that very small numbers of studies or nearly equivocal numbers of studies supporting or refuting the findings were used to determine our conclusions. This was particularly true in the review of those studies on conservatively managed cases. This can be highlighted by examining the findings for individual MPF. Depression was the most widely studied construct with 27 studies (12 conservative and 15 surgical) management. In the case of conservative management, four predicted outcome and eight did not, a clear conclusion. In the case of surgical management, nine predicted and six did not, also a clear finding. In contrast, self-efficacy was far less studied with only two studies for conservatively managed cases and none for surgical. In one study self-efficacy predicted outcome and in one it did not, and therefore the conclusion must be weighed carefully. Therefore, it is important to consider the total number of studies reviewed when interpreting the relationships between each individual MPF and outcome (Table 5).

Approach to management PPROACH TO MANAGEMENT
The implications of this review suggest that MPF are important considerations for those patients with MSD who are managed surgically. Our findings show that there is evidence for the predictive value of expectation of recovery, catastrophizing, avoidant coping, depression, and anxiety in patients receiving surgically intervention. In this group, there was no evidence that self-efficacy affected outcome. The results suggest the importance of assessment of these MPF as a part of routine surgical care for patient with shoulder disorders. In contrast, for those patients managed conservatively, the evidence for selfefficacy, expectation of recovery, and avoidant coping was equivocal and requires further study. However, there was no evidence for catastrophizing, depression and anxiety affecting outcome in this group. When evaluating the findings for each construct as it relates to management it is important to consider not only the number of studies but also the quality of studies informing the conclusion, as more highquality studies were noted for conservative management (Table 5).

Time from onset
In this review we explored the temporal influence, represented as time from onset, on the relationship between MPF and outcome. Time from onset of shoulder pain was not defined in 60% of the included references (14 of the surgical studies [82%] and ten of the conservative studies [50%]). When interpreting the findings, it is important to recognize that typically surgical intervention occurs during the chronic phase, after failed conservative management often recommended during earlier phases [73]. Therefore, it may be reasonable to conclude that in the absence of trauma, the majority of patients undergoing surgical intervention were likely in the chronic phase [74]. Although less than 20% of the surgical studies reported time from onset, in those that did, a relationship between MPF and outcome was found.
In the case of conservative intervention, it is difficult to draw conclusions regarding the temporal impact of MPF on outcome. This is because among those studies that did report time from onset, the findings were either inconclusive or against the predictive value of MPF on outcome. Therefore, we believe time from onset deserves further study in this group.

Limitations
Many included studies were small and may therefore not have sufficient power to capture a clinically relevant influence of the subgroups we have defined for this review. None of the included studies investigated all predefined constructs and therefore the full impact of these variables cannot be completely described. In addition, not all MPF were equally explored. Furthermore, some psychological constructs are complex, such as catastrophizing, which may be considered a belief or a coping strategy. For example, two studies that used the Pain Coping Scale designated catastrophizing as a coping strategy [42,50]. Yet most studies used the Pain Coping Scale to assess the impact of beliefs on expectation of outcome [46,[55][56][57]69]. There is no gold standard for the definition and classification psychological constructs. In this review, catastrophizing was assigned to the coping domain based on the opinion of a clinical mental health expert. However, future studies need to clarify the difference between beliefs and coping strategies and their impact on treatment outcome. In one study, the Orebro, a composite measure for MPF and other variables associated with outcome, was used to assess MPF [37]. Due to its composite nature, it was not possible to include the findings for specific MPF in this review. However, composite instruments may allow for the assessment of several domains simultaneously and may have clinical utility, compared to the methods in this review that explored each MPF individually. The impact of treatment for the MPF (i.e., medication, psychological interventions) on shoulder outcomes was not addressed in this review. The limitations of our review reflect the lack of a strong literature base, including the heterogeneity of study populations, which precluded the possibility of a meta-analysis [22,25]. Future studies need to address these methodological shortcomings.

Future studies
There is mention of the importance of assessing psychological factors in clinical practice guidelines for managing shoulder pain [6]. However, this does not seem to be a routine part of clinical practice as is apparent from the limited number of studies found for this review. To gain deeper insight into how to explore the role of psychological factors as predictors of outcome, it is informative to look to the spine literature. Compared to the management of shoulder disorders, an extensive literature base drives clinical management of psychological factors associated with low back pain. Consistent evidence supports the role of these psychological factors on prognosis [17] and the relationship with outcome for patients with low back pain [75,76]. However, there are limitations in generalizing the findings to other musculoskeletal disorders such as shoulder pain. While the overall relationship of low back pain with physical functioning and MPF has been described, it is unclear if the same relationship may exist for other musculoskeletal conditions.
One consideration is the relationship between psychological factors and the natural history/tissue healing associated with various musculoskeletal conditions. For instance, in patients with low back pain, fear of pain is a strong predictor of outcome [75,76]. The concept that pain does not equal damage, an important message to patients with spinal pain, may not be relevant for patients with shoulder conditions. Furthermore, while studies on back pain may inform the methodologies and research questions for shoulder pain populations, researchers must be prudent in recognizing the limitations of transposing these ideas. For example, many of the tools used to measure psychological constructs have not been validated for shoulder complaints [20]. Finally, other psychological responses to pain, such as anger, have been studied in other musculoskeletal conditions, yet are not addressed in the shoulder literature [77]. Future studies should focus on developing shoulderspecific instruments, clinical management, time from onset and all relevant psychological factors that are potentially modifiable as they relate to outcome.

Conclusions
Based on this review, expectation of recovery, catastrophizing, avoidant coping style, depression, and anxiety were the MPF most predictive of outcome in surgically managed patients with shoulder complaints. This provides sufficient evidence to suggest that implementing a biopsychosocial care paradigm to this population may be advantageous. In patients undergoing conservative intervention the evidence was either against (catastrophizing, depression, anxiety) or inconclusive (self-efficacy, expectation of recovery, avoidant coping) for the predictive value of psychological factors on outcome. However, future highquality comparative investigations and those assessing understudied constructs may shed more light on the prognostic value of MPF on outcome in this population. There is clearly a place for the study of psychological factors associated with shoulder disorders. Further investigation of all psychological factors may provide deeper insight into understanding patients with shoulder MSD, and best approaches to clinical management. Cho   -, low quality, few or no criteria fulfilled. Columns 1-14 in Appendix 2A are in response to the following questions: 1. Study question focused? 2. Included groups selected from source population that are comparable. 3. The study indicate how many who were asked to take part did so. 4. The likelihood that some eligible subjects might have the outcome at the time of the enrolment is assessed and taken into consideration. 5. What are the percentage of individuals recruited that dropped out before the study was completed. 6. Comparison is made between full participants and those lost to follow-up. 7. Outcomes clearly defined. 8. The assessment of outcome is made blind to exposure status. 9. Where blinding was not possible, there is some recognition that knowledge of the exposure status could have influenced the assessment of outcome. 10. The method of assessment of exposure is reliable? 11. Evidence from other sources is used to demonstrate that the method of outcome assessment is valid and reliable. 12. Exposure or prognostic factor assessed more than once? 13. Main potential confounders identified and taken into account in analysis. 14. Have confidence intervals been provided. 15. Overall assessment of risk of bias (++/+/−/0). 2 B: Randomized clinical trials. Columns 1-10 reflect the SIGN questions listed in the legend below the Appendix.
-, low quality, few or no criteria fulfilled. Columns 1-10 in Appendix 2B are in response to the following questions: 1. Clearly and focused question. 2. The assignment of subjects to treatment groups are randomized? 3. An adequate concealment method is used? 4. The design keeps subjects and investigators 'blind' about treatment allocation? 5. The groups are similar at start of the trials? 6. The only difference between the groups is the treatment under investigation? 7. All relevant outcomes are measured in a standard, valid and reliable way? 8. What percentage of the subjects recruited into each treatment arm dropped out before the study was completed? 9. All the subjects are analysed in the groups to which they were allocated (Intention to treat analysis)? 10. Where the study is carried out at more than one site, results are comparable for all sites? Overall quality of the study? (++/+/−/0). ++, high quality, most (≥60%) of the criteria fulfilled (if < 60% fulfilled, the conclusions of the study are very unlikely to alter the findings) +, moderate quality: some criteria fulfilled (< 60%)