Emotional distress was associated with persistent shoulder pain after physiotherapy: a prospective cohort study

The primary outcome in this study was the Numeric Pain Rating Scale (NPRS) [10]. The NPRS is an 11 point scale where 0 indicates no pain, and 10 indicates the worst imagined pain. In this study, the scale was used as a measure of pain intensity during the last week. The PSFS was used as a secondary outcome [9]. In this questionnaire, the patients wrote down up to three activities they found impossible or had difficulty doing because of their problem. The difficulty associated with each activity was rated on a scale from 0 (impossible to perform the activity) to 10 (no difficulty, or at the same level as before the pain occurred). The average rating of the activities was used in the analyses as a measure of self-reported disability [11, 12]. Only scores from activities that had been rated both before and after treatment were included in the calculation of the average score.

The set of variables considered as potential prognostic factors was obtained by the FysioPol-questionnaire package at study onset, including clinical characteristics identified with prognostic value in previous research, and demographic factors. The Hopkins Symptom Checklist-25 (HSCL-25) [13] was used as a measure of emotional distress. The questionnaire aims to assess symtomps of anxiety, depression and somatization. HSCL-25 is a shorter version of the Symptom Checklist 90 (SCL-90) and consists of 25 items that are rated from 1 (not at all) to 4 (very much). The total score was obtained by averaging the scores, and ranged between 1 and 4. A maximum of five missing items were accepted. A higher total score indicates a higher level of emotional distress. The Norwegian version of the HSCL-25 has been used in several studies of musculoskeletal pain [14‐17]. Evidence of psychometric properties of HSCL-25 in the population of patients with shoulder pain is to our knowledge lacking. Other clinical characteristics considered as potential prognostic factors included pre-treatment pain intensity measured by NPRS; pre-treatment disability measured by PSFS; duration of pain divided into 0–3 months, 4–12 months and more than 12 months; use of pain relieving drugs divided into less than every week and every week or more; concomitant neck pain; and number of pain sites divided into two pain sites or less and three pain sites or more. The demographic factors included age, sex, body mass index (BMI), level of education, work status, relationship status and smoking status. Level of education was divided into lower level (≤ 13 years) and college / university (> 13 years). Work status was divided into working and not working. Students were included in the working group, which consisted of both full time and part time working patients, while retirees were included in the not working group, which also included unemployed patients and patients on full time sick leave and disability pension.

Descriptive data are presented as number of patients and percentages, means and standard deviations or medians and interquartile range. Paired t-tests were used to identify changes from pre- to post- treatment measures of pain intensity and self-reported disability. Characteristics of the individuals who were lost to follow-up were compared to pre-treatment characteristics of the study sample. The groups were compared on pain intensity, disability, emotional distress and age using independent t-tests, and sex using a Chi-square test.

Simple linear regression analyses were performed to examine the relationship between each of the potential prognostic factors and the outcome (the NPRS and the PSFS). The variables with a statistically significant relationship with the outcome at the 20% level (P < 0.20) [18] were considered for the final multiple regression models. A backward manual elimination method was used to remove those variables with the highest P-value, one by one. The elimination was repeated until the remaining variables in the models were all statistically significant at the 5% level (P < 0.05). To prevent elimination of a variable at one step in the analysis process being crucial, the variables removed on backward elimination were all re-entered in the models one by one, and remained in the models if they were statistically significant at the 5% level. The multiple regression models were adjusted for age and sex.

Assumptions for the regression models were assessed. Correlation analyses were performed for all the independent variables and the correlation had to be less than 0.7 between the variables to be entered in the models [18]. An extreme value in the variable of BMI (> 39) was interpreted as an univariate outlier, due to a standardized score in excess of 3.29, disconnected from the other standardized scores [18]. Since a BMI-value of > 39 may involve other health problems than the ones investigated in this study, and the case therefore may not be a part of the population we intended to investigate, the case was excluded.

SPSS version 24 was used for all of the statistical analyses.

In a recent systematic review of prognostic factors for shoulder pain, strong evidence was found that high scores on the SPADI questionnaire, more shoulder pain, and a longer duration of complaints, were associated with persistent shoulder pain [6]. In contrast to the results of the present study, the systematic review did not find any evidence that psychological factors were associated with shoulder pain. However, the authors of the systematic review claimed that psychosocial factors might have been underestimated due to limited use of questionnaires containing these functions in shoulder pain populations [6].

Emotional distress was investigated in two previous studies on shoulder pain, which found no association with outcome [16, 17]. One of the studies comprised patients with diagnosed subacromial pain in secondary care [16], the other study included patients with diagnosed rotator tendinosis in primary care [17]. The inconsistencies between these studies and the present findings might be explained by differences in the study populations; the present study was a cohort study comprising patients with self-reported shoulder or arm pain.

Furthermore, the present findings are not consistent with those in a previous study on patients presenting new episodes of shoulder- or low back pain to their general practitioner [19]. Interestingly, the cohort study found that for the shoulder pain patients, no psychological factors were associated with persistent symptoms or disability after three months, with the exception of catastrophizing, which in patients with a long duration of pain at study onset (≥ 3 months) was associated with persistent symptoms [19]. However, it is worth noting that the patients in this study had no distress at baseline, measured by a subscale of the Four-Dimensional Symptom Questionnaire [20], while the patients in our study had an average level of 1.6 (SD 0.5) on the HSCL-25. This might explain the difference in findings. Another possible reason for different results might be that factors associated with persistent pain after an exercise therapy intervention differ from factors associated with pain after other types of treatment.

Nevertheless, a recent cohort study from the UK reported that other psychological factors than distress, such as patient expectations of recovery and pain self-efficacy, were associated with the level of pain and disability after physiotherapy in patients with shoulder pain [7]. Although the study did not identify any association between anxiety and depression and outcome, the authors suggested that this could be due to a low number of included patients with extreme anxiety and depression [7]. Based on the findings of the study, the authors concluded that when assessing people with musculoskeletal shoulder pain and considering referral to physiotherapy services, psychosocial and medical information should be considered [7].

Our data showed no association between emotional distress at study onset and self-reported disability after treatment. The factors associated with higher post-treatment disability were higher pre-treatment disability, duration of pain for more than 12 months in comparison to 0 to 3 months, concomitant neck pain and a lower level of education. The results indicate that patients with a history of chronic shoulder pain and disability may have a poor outcome in terms of disability regardless of emotional distress.

Our findings that emotional distress was associated with pain, but not with disability, are difficult to explain. One explanation might be that mental functions are more directly associated with the experience of pain, than with disability. However, a cross-sectional study on people with chronic shoulder pain found that psychological distress was correlated with disability, but not with pain [21], which indicate that the relationship between distress, disability and pain may be complex.

This study has some limitations that should be considered. Firstly, since the inclusion of patients was based on self-reported shoulder- or arm pain, we were not able to discriminate between localised shoulder pain and pain related to the shoulder, arm and hand. Secondly, a number of patients who met the inclusion criteria were excluded from the analyses due to loss to follow-up or missing values in some of the variables. However, the patients lost to follow-up did not differ from the study sample on characteristics such as sex, age, pain intensity, disability or emotional distress at study onset. A third concern involves the duration of treatment. There was a variety in duration of treatment, with a median of 5 weeks of physiotherapy (IQR 3 to 6). A period of 12 weeks of physiotherapy is often suggested for patients with shoulder pain [22]. Future research could establish whether the results after a longer follow-up differ from the results in this study.

The results of this study showed that emotional distress at study onset was associated with the intensity of shoulder pain after physiotherapy. Psychological factors in general are little emphasised in the examination of shoulder pain, and it is rather the structural and biomechanical aspects of the condition that are usually considered in clinical decision-making. The findings of this study support, however, that emotional distress should be considered in the initial physiotherapy examination of shoulder pain.

Emotional distress may be a cause or a consequence of shoulder pain, and whether the treatment should be directed towards reducing distress is not possible to tell based on the present findings. Nevertheless, findings of a meta-analysis on neck and back pain indicate that psychological distress mediates the relationship between pain and disability [23]. Future research is needed to investigate the relationship between emotional distress, pain and disability. The research should identify whether targeting emotional distress in shoulder pain rehabilitation is likely to improve the outcome for shoulder patients with a high degree of emotional distress, to identify whether and how to best individualise the treatment for these patients.