A longitudinal assessment using data from electronic medical records (EMR) and claims data from individuals seen in 28 outpatient physical therapy clinics from November 2019 to March 2021 at Bellin Health, a large Midwest health system in the United States. The Bellin Health Institutional Review Board identified the study as exempt and consent deemed unnecessary because routinely collected de-identified health information was used. The Reporting of studies Conducted using Observation Routinely-collected health Data (RECORD) statement was used to guide reporting of this study.39

De-identified person-level data were extracted for all outpatient visits. The index date was the initial physical therapy evaluation. Self-reported outcomes completed on index date were collected. All medical visits that included International Classification for Disease-Tenth Revision (ICD-10) codes for the entire 1-year before the index date were abstracted, to include those related to the patient's initial MSP diagnosis (list of ICD-10 codes in the supplementary material).

Patients were included if they were ages 18–90 years and receiving a physical therapy evaluation for a new episode (defined as no physical therapy in the past year for the same body region) of either neck pain, low back pain, hip osteoarthritis (OA), or knee OA, (list of ICD-10 codes in the supplementary material), but the exact acuity of these conditions could not be determined. These musculoskeletal disorders were chosen as they represent common chronic MSP conditions.2,40,41 Patients seeking care for pain unrelated to a MSP condition were excluded (e.g., cancer, neurological disease) were excluded.

Patient demographics, including age, sex, race, marital status, and prior history of surgery (either total knee arthroplasty, total hip arthroplasty, and/or any spinal surgery), were identified.

The comorbidities of interest included anxiety, metabolic disorder, chronic pain diagnosis, depression, nicotine dependence, PTSD, and sleep disorders, all of which have been identified as potential influencers of MSP outcomes.6 For practical purposes, sleep apnea and insomnia were collapsed into a single category of sleep disorders. Comorbidities were identified as medical diagnoses (ICD-10) rendered by a licensed clinician after a care-seeking event in the year before the index date (supplementary material).

Covariates included age, marital status, race, and sex. Other covariates included body region and surgical history in the same body region. Marital status was grouped into three categories based on responses provided by participants at intake: married/cohabitation (married or domestic partner), single/unmarried (single, widowed, divorced, or separated), and unknown (declined or unknown). These demographic variables were chosen for the models because they are known to influence pain intensity or psychological distress.41–44 Specifically, marital status has been associated with emotional suffering in chronic pain.45 Pain reports can vary by body region.29 Diagnoses for pain in the cervical or lumbar region were collapsed into the classification of spinal region, and diagnoses for hip or knee pain consistent with OA were classified as lower extremity. Surgical history of knee or hip arthroplasty in the same extremity or spine surgery in the same region up to one year prior was identified from the EMR.43

Numeric Pain Rating Scale (NPRS): The NPRS measures pain intensity using an 11-point scale, ranging from 0 (no pain) to 10 (worst pain imaginable).42 Patients were asked to rate their average pain over the past week. The 11-point NPRS is considered a reliable, valid, and responsive self-report outcome for patients with MSP.43,44

Optimal Screening for Prediction of Referral and Outcome - Yellow Flags (OSPRO-YF 10): The OSPRO-YF 10 is a reliable and valid tool measuring a multidimensional pain-related psychological distress risk factor profile with 11 distinct flags across three domains.29,45 The three domains are: negative mood (depression, anxiety, and anger), negative coping (beliefs, catastrophizing, kinesiophobia, and pain-anxiety), and positive affect/coping (pain self-efficacy, rehabilitation self-efficacy, and chronic pain acceptance).29,45 The term yellow flag describes a psychological factor associated with the pain experience that carries a risk for poorer prognosis.31,46,47 However, yellow flags are not synonymous with mental health disorders.31 The 10-item version of the OSPRO-YF is standard of care in the clinics and has an 81% accuracy rate compared to the 17-item questionnaire.31

Scoring methods for the OSPRO-YF 10 include either a weighted average of the total yellow flag count (11 total possible), a simple summary score (3–53 points), or a total yellow flag domain count (three total possible).30 Scoring the OSPRO-YF tool using either the simple summary score method or yellow flag count has yielded similar predictive accuracy demonstrating good model performance and quality.30 The OSPRO-YF has demonstrated accuracy in predicting 12-month disability, the persistence of chronic pain, risk for subsequent surgery following an episode of physical therapy, and identification of pain and psychological distress phenotypes in patients with hip and knee OA.28,29,41

Generalized linear models were used to determine the relationship between the presence of comorbidities and pain intensity, as well as pain-related psychological distress. Each model was adjusted for the covariates previously defined. The NPRS data were normally distributed and not zero-inflated; therefore, normal identity link was used in the generalized linear model to assess the relationship between pain-related comorbidities and NPRS values. Due to the overdispersion of the OSPRO-YF 10 total yellow flag count variable, a negative binomial family of distributions was used to assess the relationship between pain-related comorbidities and pain-related psychological distress. Negative binomial regression is commonly employed to analyze overdispersed count data by inference based on adjusted score equations for mean and median bias reduction.48 A priori significance was chosen at p < 0.05 and 95% Confidence Interval (CI) was reported. All variables were assessed for multicollinearity. Only variables with low concern for multicollinearity were retained in the model (based on the correlation matrix, variance inflation factors, and tolerance level estimates). Analyses were completed with SPSS version 28.0 (IBM Corp, Armonk, New York). The best model fit for each analysis was identified using the lowest Akaike's Information Criterion, Bayesian Information Criterion, and most appropriate goodness of fit with Omnibus testing. Variables without a normal distribution are reported with median and interquartile range (IQR).

As no standard consensus exists yet on the optimal way to clinically interpret the OSPRO-YF 10 results, sensitivity analyses were run using several variations of the pain-related psychological distress measures. In addition to the total count of yellow flags, a model using the OSPRO-YF 10 yellow flags domain count (0 to 3 domains) and another model using the simple summary score were also evaluated (supplementary material). Finally, as some have recommended treating the NPRS as a count variable due to its ordinal nature and the potential for zero-inflated scores,49 we re-ran the analysis using a Poisson distribution instead of the normal identity link.

Adjusting for all other factors, comorbid depression increased pain intensity by 0.7 points (95%CI: 0.2, 1.2). A diagnosis in the spine region compared to the lower extremity was associated with an increase in pain intensity of 0.6 points (95%CI: 0.1, 1.1). History of prior surgery was associated with an increase in pain intensity of 0.8 points (95%CI: 0.3, 1.4; Table 3). A sensitivity analysis using a Poisson distribution did not change the results except that spine and lower extremity regions were no longer significant predictors of pain intensity (supplementary material). The data were normally distributed which likely contributed to minimal differences in each model.

No pain-related comorbidities were associated with the number of total yellow flags (Table 4). The sensitivity analyses for OSPRO-YF 10 simple summary score and number of domains showed no associations with pain-related comorbidities. The one variable in the model associated with pain-related psychological distress was female sex, showing a small reduction in the total number of yellow flags (0.2 flags; 95%CI: −0.3, −0.02; Table 4). The prevalence of individuals positive for the negative coping domain was different by sex (males n = 170; 73.3% and females n = 199; 66.3%). Females had 28.1% decreased odds of scoring positive for negative coping compared to males; however, this was not statistically significant (odds ratio [OR]= 0.7, 95%CI: −0.5, 1.1, p = 0.09) – supplementary material).

Depression's association with increased pain intensity corroborates earlier studies investigating depression's impact on pain.28,36,37 Accounting for pain-related comorbidities associated with healthcare outcomes and utilization, our results maintained the association of depression with pain intensity while providing new evidence highlighting the relationship at the point of care during an initial physical therapy evaluation.6,16,50 Depression and higher pain intensity are prognostic factors in developing chronic low back pain.51 Likewise, depression is associated with pain intensity and multiple medical comorbidities in patients with knee OA.52 There is likely a bi-directional association between MSP and depression. For example, symptomatic knee OA, in addition to other MSP conditions, have been shown to be associated with a new onset of depression.52–55 The impact of not addressing depression throughout an episode of MSP may influence the long-term outcomes for pain intensity and should be considered in further research.

Despite the high prevalence of metabolic disorders in this cohort, a significant relationship between the presence of care-seeking for metabolic disorders and pain intensity was not found. A higher body fat percentage, particularly central adiposity, is associated with developing MSP and is weakly indicative of worsening joint pain.56,57 Our study relied on a documented diagnosis of obesity in the medical record, likely representing a high level of variance across the spectrum of obesity. Obesity measures using a continuous variable (e.g., body mass index) may have increased specificity of these relationships but were not available to the research team. We also did not analyze the total number of pain-related comorbidities cumulatively for each individual. It is plausible there is a compounding effect of multiple pain-related comorbidities on pain intensity.

Care-seeking individuals for spine pain were slightly more likely to have increased pain intensity than those care-seeking for lower extremity conditions. To our knowledge, no direct comparison of pain intensity between body regions has been made. We offer two potential explanations for this finding. First, individuals with spine pain were younger than those with lower extremity pain. Higher pain intensity is associated with younger age in patients with hip/knee OA.58,59 While we adjusted for age in our models, the spine cohort was much larger, making for an unbalanced comparison, and a lack of diversity of ages across both cohorts could limit conclusions. Second, previous studies have suggested that a spatial summation, or how widespread the pain symptoms present, may be a significant determinant of pain intensity.60,61 The referral patterns or indications of spatial summation were not included in this analysis, but future studies exploring relationships between pain-related comorbidities, nociplastic pain, and somatic referral patterns are warranted. However, of noteworthy consideration is the lack of difference in pain intensity across body regions in the sensitivity analysis. Finally, the relationship between a previous history of surgery and pain is consistent with other studies that suggest the possibility of chronic post-surgical pain.62–64

The pain-related comorbidities investigated in this study were not associated with pain-related psychological distress. One rationale for this finding may be overlapping characteristics between the items of psychological distress. For example, depression and anxiety occurring concurrently may magnify the level of psychological distress for the individual, but not to the extent necessary to score within the top quartile for the legacy measure required to generate a yellow flag.31

Specific factors, including socioeconomic status, type of health insurance, and education level, were controlled for in the original derivation study of the OSPRO YF.31 We hypothesize that controlling for these additional factors may have provided more advantageous scores for pain catastrophizing in this cohort. An observable distinction between the OSPRO-YF scores by sex in this study identified possible differences in the negative coping domain but was not statistically significant (p = 0.08).

Our finding that female sex decreased the extent of pain-related psychological distress differs from reports in previous studies.65 For example, female sex was associated with severe psychological distress in patients receiving orthopedic care for shoulder pain.66 Our study controlled for surgical history, whereas the aforementioned study included patients undergoing rehabilitation and operative management. In this case, the differences in patient characteristics and intervention strategy may influence the generalizability of the results. Further investigation into sex-specific differences in the OSPRO-YF 10 negative coping items may reveal a protective association with yellow flag count.45 Furthermore, analysis of items in the fear-avoidance and catastrophizing domains of the OSPRO-YF 17 would enhance the understanding of different responses between sexes.

Sensitivity analyses identified that previous surgical history (hip/knee arthroplasty or spine) influenced the OSPRO-YF 10 simple summary score, which to our knowledge, has not been investigated and may provide early evidence of an association between surgery and psychological distress. The variance in the association of the covariates may represent subtle differences in the scoring methods. In addition, some comorbidities were excluded from the models due to such few cases (i.e., PTSD in only five individuals).

Understanding the associations of pain-related comorbidities with pain intensity, pain-related psychological distress, and other relevant outcomes remain important for the future of healthcare. Medical professionals often synthesize a long list of potentially relevant factors influencing pain outcomes. It has been estimated that 75–85% of depression goes untreated67 and improving primary care providers’ screening of depression has increased the odds of subsequent diagnosis and treatment.68 Additionally, physical therapists have identified that comorbid depression may be highly influential in patient management, but they only formally screen for it during 18% of physical therapy evaluations.22,69 Our study may help fill this knowledge gap for clinicians and inform subsequent psychologically informed management for pain-related outcomes.31,70 The results highlight the need to further explore multidisciplinary collaboration as a potential to influence the trajectory of pain and healthcare costs.5,71

There are limitations to this study. First, the results may lack generalizability in some settings as 95.3% of this sample was white with a median age of 59.0 years, and only individuals with spinal pain or hip/knee OA were included. Studies conducted with a larger diversity of race have demonstrated associations between pain-related comorbidities and pain intensity.38 Some comorbidities affect different populations disproportionately,72 and these results should be validated in more diverse cohorts. Second, all data collected were from one health system, which may reduce generalizability. Third, the stage and severity of the MSP disorders and comorbidities were not available and their impact on outcomes may vary. Fourth, the differing results on the OSPRO-YF 10 analysis between total yellow flag count and the sensitivity analyses for simple summary score and yellow flag domain count scoring indicate the results should be interpreted with caution as the findings were not consistent across the three methods (thus our multiple sensitivity analyses). Consensus on the most appropriate interpretation of the OSPRO-YF 10 is still lacking.30,65 Fifth, no causal claims can be made about these relationships, and further validation using a design to assess prospective changes in pain is recommended for more definitive conclusions.