What Are the Underlying Mental Health Constructs Associated With Level of Capability in People With Knee and Hip Osteoarthritis?

Abstract

Background

Mental health characteristics such as negative mood, fear avoidance, unhelpful thoughts regarding pain, and low self-efficacy are associated with symptom intensity and capability among patients with hip and knee osteoarthritis (OA). Knowledge gaps remain regarding the conceptual and statistical overlap of these constructs and which of these are most strongly associated with capability in people with OA. Further study of these underlying factors can inform us which mental health assessments to prioritize and how to incorporate them into whole-person, psychologically informed care.

Questions/purposes

(1) What are the distinct underlying factors that can be identified using statistical grouping of responses to a multidimensional mental health survey administered to patients with OA? (2) What are the associations between these distinct underlying factors and capability in knee OA (measured using the Knee Injury and Osteoarthritis Outcome Score, Joint Replacement [KOOS JR]) and hip OA (measured using Hip Disability and Osteoarthritis Outcome Score, Joint Replacement [HOOS JR]), accounting for sociodemographic and clinical factors?

Methods

We performed a retrospective cross-sectional analysis of adult patients who were referred to our program with a primary complaint of hip or knee pain secondary to OA between October 2017 and December 2020. Of the 2006 patients in the database, 38% (760) were excluded because they did not have a diagnosis of primary osteoarthritis, and 23% (292 of 1246) were excluded owing to missing data, leaving 954 patients available for analysis. Seventy-three percent (697) were women, with a mean age of 61 ± 10 years; 65% (623) of patients were White, and 52% (498) were insured under a commercial plan or via their employer. We analyzed demographic data, patient-reported outcome measures, and a multidimensional mental health survey (the 10-item Optimal Screening for Prediction of Referral and Outcome-Yellow Flag [OSPRO-YF] assessment tool), which are routinely collected for all patients at their baseline new-patient visit. To answer our first question about identifying underlying mental health factors, we performed an exploratory factor analysis of the OSPRO-YF score estimates. This technique helped identify statistically distinct underlying factors for the entire cohort based on extracting the maximum common variance among the variables of the OSPRO-YF. The exploratory factor analysis established how strongly different mental health characteristics were intercorrelated. A scree plot technique was then applied to reduce these factor groupings (based on Eigenvalues above 1.0) into a set of distinct factors. Predicted factor scores of these latent variables were generated and were subsequently used as explanatory variables in the multivariable analysis that identified variables associated with HOOS JR and KOOS JR scores.

Results

Two underlying mental health factors were identified using exploratory factor analysis and the scree plot; we labeled them “pain coping” and “mood.” For patients with knee OA, after accounting for confounders, worse mood and worse pain coping were associated with greater levels of incapability (KOOS JR) in separate models but when analyzed in a combined model, pain coping (regression coefficient -4.3 [95% confidence interval -5.4 to -3.2], partial R² 0.076; p < 0.001) had the strongest relationship, and mood was no longer associated. Similarly, for hip OA, pain coping (regression coefficient -5.4 [95% CI -7.8 to -3.1], partial R² 0.10; p < 0.001) had the strongest relationship, and mood was no longer associated.

Conclusion

This study simplifies the multitude of mental health assessments into two underlying factors: cognition (pain coping) and feelings (mood). When considered together, the association between capability and pain coping was dominant, signaling the importance of a mental health assessment in orthopaedic care to go beyond focusing on unhelpful feelings and mood (assessment of depression and anxiety) alone to include measures of pain coping, such as the Pain Catastrophizing Scale or Tampa Scale for Kinesiophobia, both of which have been used extensively in patients with musculoskeletal conditions.

Level of Evidence

Level III, prognostic study.

Introduction

Mental health concerns account for a substantial proportion of the observed variation in levels of symptom intensity and capability across musculoskeletal conditions [24‐26, 34, 35]. This finding is especially relevant in the management of osteoarthritis (OA), where the prevalence of such concerns is relatively high [2, 19, 20, 33, 45, 54]. Aspects of mental health include negative mood (feeling worry or despair and symptoms of anxiety or depression) [3, 16, 18, 45, 51, 53, 54], fear avoidance (tending to avoid movement or activities for fear of increased pain or reinjury) [47, 48], unhelpful thoughts and misinterpretations regarding painful symptoms (responding to actual or anticipated pain with maladaptive, catastrophic thoughts) [16, 18, 42, 46-48], and low self-efficacy (expressing low confidence in achieving activities of daily life and one’s goals because of pain) [34, 42, 50].

These aspects can be measured using a range of validated surveys, which have variably demonstrated a negative association between mental health and symptom intensity, capability, and recovery trajectories in the management of OA [3, 12, 15, 16, 18, 20, 21, 34, 45, 47]. As a result, there are a growing number of best-practice recommendations that encourage routine screening for mental health concerns as part of a more comprehensive approach to managing musculoskeletal conditions [1, 27, 41, 43]. Further, there is increasing awareness among the musculoskeletal community about building services and clinical competencies responsive to patients experiencing different mental health concerns [1-3, 22, 31].

Despite widespread evidence that underscores the importance of mental health in patients with OA, clinical knowledge gaps remain regarding the extent to which different characteristics of an individual’s mindset impact capability and which of these characteristics warrant particular attention [22]. Specifically, if we can simplify the multitude of mental health characteristics into conceptually and statistically distinct constructs, this may offer clinical and statistical benefits. For instance, if clinicians can understand which constructs are most strongly associated with capability, and therefore warrant particular attention, they can prioritize their approach to assessment and intervention, accounting for mindset. Statistically, methods such as factor analysis can group mental health characteristics into factors that can minimize redundancy, be incorporated in multivariable regression analyses, and address multicollinearity, where highly correlated variables entered into the same model may lead to false-negative results [23, 52].

We therefore asked: (1) What are the distinct underlying factors that can be identified using statistical grouping of responses to a multidimensional mental health survey administered to patients with OA? (2) What are the associations between these distinct underlying factors and capability in knee OA (measured using the Knee Injury and Osteoarthritis Outcome Score, Joint Replacement [KOOS JR]) and hip OA (measured using Hip Disability and Osteoarthritis Outcome Score, Joint Replacement [HOOS JR]), accounting for sociodemographic and clinical factors?

Patients and Methods

Study Design and Setting

We performed a retrospective cross-sectional analysis of data collected from a consecutive series of 954 adult patients who presented with knee and hip pain secondary to OA at a comprehensive outpatient OA management program (The Joint Health Program, Duke University) [23, 40]. The program is offered in a large academic health system providing orthopaedic specialty care, including care focused on lower extremity conditions, to urban and rural populations between October 2017 and December 2020. The program is delivered across 23 ambulatory clinics and led by 25 physical therapists working as primary OA care providers who develop comprehensive, individualized management plans involving a range of evidence-based nonoperative strategies [23, 42].

Patients

All adult patients 18 years or older attending a new-patient appointment at our program with a primary complaint of hip or knee pain secondary to OA (diagnosed by the referring orthopaedic surgeon) were eligible for inclusion. Of the 2006 patients in the database, 38% (760) were excluded because they did not have a diagnosis of primary OA and 23% (292 of 1246) were excluded because of missing data, leaving 954 patients available for analysis. Patients entering the program complete a registration questionnaire including basic demographics, patient-reported outcome measures, and mental health surveys before or on arrival to the clinic unless they have cognitive deficiencies or language barriers precluding completion. Patients completed surveys on a tablet device via a Health Insurance and Portability Accounting Act–compliant patient portal in the electronic health record (Epic) before their clinical review. To create an analytic dataset, we queried all new patients from program inception with complete measurement data (described below). Approximately 72% of all patients in the program have complete questionnaire and survey data at baseline. Reasons for not having complete data include patients skipping survey items during intake, patient refusal to complete certain items or surveys, or data that may not have been transferred from paper to electronic forms in the early phases of the program.

Descriptive Data

Of the 954 patients included in this study, 73% (697) were women, with a mean age of 61 ± 10 years (Table 1). A notable proportion of patients (65% [623]) associated with being of White race, were married or had a partner (58% [556]), were employed (49% [471]), and were insured under a commercial plan or via their employer (52% [498]). Seventy-nine percent (754) of patients had knee OA and 21% (200) had hip OA. Sixty-three percent (474 of 754) of patients with knee OA and 61% (121 of 200) of patients with hip OA had an advanced radiographic grade of OA (Kellgren-Lawrence [KL] Grade 3 or 4).

Table 1.

Patient demographics (n = 954)

Variable	Knee (n = 754)	Hip (n = 200)
Age in years	62 ± 10	61 ± 10
BMI in kg/m2	35 ± 8.9	32 ± 7.9
Men	25 (191)	33 (66)
Race
White	63 (478)	73 (145)
Marital statusa
Married or partner	58 (439)	59 (117)
Single	16 (118)	16 (32)
Divorced or separated	12 (91)	12 (24)
Widowed	9 (70)	6 (11)
Employment status
Employed	48 (363)	54 (108)
Disabled	37 (282)	34 (68)
Retired	8 (57)	4 (8)
Unemployed	7 (52)	8 (16)
Insurance type
Commercial or employer	52 (395)	52 (103)
Medicare	39 (297)	41 (82)
Safety-net	3 (26)	5 (9)
Medicaid	3 (24)	2 (4)
Other	2 (12)	1 (2)
Kellgren-Lawrence grade
1	28 (211)	26 (52)
2	9 (69)	14 (27)
3	27 (207)	25 (49)
4	35 (267)	36 (72)
Elixhauser comorbidity indexa	0 (-3 to 4)	0 (-1 to 5)
Primary total joint replacement	14 (102)	31 (61)
KOOS/HOOS JR at baselineb	49 ± 14	51 ± 15

Data presented as % (n), mean ± SD, or median (IQR).

^aThere were 36 missing values in the variable “marital status” in the knee group (4.8%) and 16 in the hip group (8.0%).

^bKOOS JR and HOOS JR range from 0 (worst joint health) to 100 (best joint health); Elixhauser Comorbidity Index ranges from -7 (lowest disease burden) to +12 (highest disease burden).

Measurements

In patients with hip OA, the HOOS JR [37] was collected upon initial evaluation (baseline) to evaluate pain, function, and activities of daily living after THA. In patients with knee OA, we collected the KOOS JR [36] at baseline to evaluate overall knee health after TKA, including levels of stiffness, pain, function, and activities of daily living. All items in these tools are scored based on a 7-day recall period [36]. The HOOS JR and KOOS JR instruments are scored from 0 (worst joint health) to 100 (best joint health). An anchor-based minimum clinically important difference has been defined as 18 for the HOOS JR and 14 for the KOOS JR [38].

We used the 10-item Optimal Screening for Prediction of Referral and Outcome-Yellow Flag (OSPRO-YF) [32] assessment tool to assess multiple mental health characteristics. It is best described as a calculator designed to estimate what a patient would score on 11 full-length psychologic (Patient Health Questionnaire-9, State-Trait Anxiety Inventory, State-Trait Anger Expression Inventory) and fear avoidance questionnaires (Fear-Avoidance Beliefs Questionnaire physical activity subscale, Fear-Avoidance Beliefs Questionnaire work subscale, Pain Catastrophizing Scale, Tampa Scale of Kinesiophobia, Pain Anxiety Symptom Scale, Pain Self-Efficacy Questionnaire, Self-Efficacy for Rehabilitation, and Chronic Pain Acceptance Questionnaire) [9]. Originally designed to reduce the substantial response burden associated with assessing many different mental health characteristics, and unlike most mental health assessment tools, the OSPRO-YF is scored to obtain 11 different score estimates (which are continuous metrics) using a series of mathematical equations [32]. The tool is shown to have good test-retest reliability and predict important outcomes such as pain intensity, magnitude of incapability, general health status, and use of surgery after nonoperative care among people with musculoskeletal conditions [7, 10, 13, 14, 19, 36]. In our institution, we use the 10-item version, which takes approximately 2 minutes to complete and shows similar predictive and concurrent validity to the longer 17-item version [14].

We also collected sociodemographic data (age, gender, race, marital status, employment status, and insurance type) and clinical data (clinical comorbidities to calculate the Elixhauser score). An institutionally developed machine-learning algorithm that was trained using input from two senior total joint arthroplasty surgeons (WJ and RCM) was used to rate the severity of OA according to KL grades from baseline radiographs (either sent from referring practices or taken during their initial review by the orthopaedic surgeon or physical therapist) [30]. The KL grading system is known to have moderate interobserver agreement [30, 53]. Doubtful joint space narrowing with possible osteophyte lipping is rated as Grade 1; definite osteophytes and possible joint space narrowing is rated as Grade 2; multiple osteophytes, definite joint space narrowing, and possible bony deformity is rated as Grade 3; and large osteophytes, marked joint space narrowing, severe subchondral sclerosis, and bony contour deformity on radiographs are assigned as Grade 4 [29].

Primary and Secondary Study Outcomes

Our primary study objective was to identify distinct underlying mental health factors using an exploratory factor analysis of the OSPRO-YF score estimates. An exploratory factor analysis can identify statistically distinct underlying factors for the entire cohort based on extracting the maximum common variance among the variables of the OSPRO-YF. Once exploratory factor analysis establishes factors based on intercorrelations between different mental health characteristics, a scree plot was used to select the subset of distinct factors. Subsequently, our secondary study objective was to identify associations between underlying mental health factors (latent variables) and capability in knee and hip OA after accounting for sociodemographic and clinical factors. Predicted factor scores of these latent (underlying) variables were generated and were used as explanatory variables in a multivariable analysis to identify variables associated with HOOS JR and KOOS JR scores.

Ethical Approval

This study, which used retrospective data, was deemed exempt by our institutional review board.

Statistical Analysis

We collected descriptive statistics of all participants. Continuous variables are presented as the mean ± standard deviation or median (interquartile range), depending on their distribution. We performed an exploratory factor analysis (Stata 13.0, StataCorp) of the 11 OSPRO-YF score estimates to identify statistically unique underlying factors for the entire cohort. This statistical technique extracts maximum common variance among variables, which demonstrates how strongly the mental health characteristics are intercorrelated. The strength of the intercorrelation between variables is expressed in factors that can range from -1 to 1, where -1 signifies a perfect negative correlation, 0 signifies no correlation, and 1 signifies perfect positive correlation. The derived factors may be correlated with each other; however, correlations between factors are generally weaker than correlations among the characteristics within factors, meaning that this process identifies factors that are statistically and often conceptually unique from one another. To determine the number of factors present, we generated a scree plot (a graph that is commonly used to determine the number of unique factors in an exploratory factor analysis by plotting the total amount of variance explained against the principal components [latent factors]) and considered all factors with Eigenvalues (the total amount of variance explained by the principal components) above 1.0 as unique factors. This is a common method of factor determination when using exploratory factor analysis. Predicted factor scores of these latent variables were generated and used as explanatory variables in the multivariable analysis. We report factor loadings and item-rest correlations (Cronbach α), which are measures that tell us how strongly each mental health characteristic is related to the common factors identified with the technique. The Cronbach α was calculated separately for items loading on to each factor.

To determine how strongly each factor was associated with baseline HOOS JR and KOOS JR scores, we conducted multivariable linear regression analyses. We decided a priori to develop separate models for the HOOS JR and KOOS JR. In these analyses, we wanted to control for other potentially important covariates such as age, sex, race, BMI, marital status, employment status, insurance type, comorbidity index, and KL grade. To determine which covariates to include in the models, we conducted bivariate analyses to gain a better understanding of how covariates could be independently associated with HOOS JR and KOOS JR. For bivariate analyses, we conducted t-tests or analysis of variance tests for categorical variables, and calculated Pearson correlation coefficients for continuous variables. Any covariate with a bivariate relationship of p less than 0.10 was included in the appropriate multivariable model. In the bivariate analysis, age, BMI, sex, race, marital status, employment status, insurance type, and KL grade were associated with KOOS JR score. In the bivariate analysis, only employment status and KL grade were associated with HOOS JR score (Supplemental Tables 1-6; http://links.lww.com/CORR/B274). Therefore, these covariates were added to their respective multivariable models.

We first conducted multivariable models with each of the derived factors entered into separate models, along with the appropriate covariates. For example, if we found two distinct factors in the exploratory factor analysis, we would conduct the same multivariable analysis for each of the two factors (that is, we would include only one factor at a time). This step would help us understand whether each of the derived factors alone explained meaningful variance in HOOS JR and KOOS JR scores after accounting for covariates. Then, in the final step, we developed a multivariable model that included all derived factors together, accounting for covariates. This step allowed us to identify which derived factor or factors had the strongest relationship with HOOS JR and KOOS JR scores when considering the influence of other derived factors. The result of this step would provide guidance on which mental health characteristics may be the most important to assess in clinic, because they would have the strongest relationship with self-reported level of capability.

Results

Underlying Mental Health Factors in Patients With Knee and Hip OA

We identified two underlying mental health factors we labeled “pain coping” and “mood” (Fig. 1) by examining the characteristics that loaded onto each of the factors (Table 2). Factor 1 included characteristics such as fear avoidance beliefs, catastrophizing, kinesiophobia, low self-efficacy, and poor acceptance; we called this factor “pain coping” because these characteristics reflect a person’s helpful and unhelpful thoughts and reactions toward pain. Factor 2 included characteristics such as anxiety, depression, and anger; we called this factor “mood” because these characteristics reflect feelings of hopelessness, worry, or despair, which are common in mood conditions.

Fig. 1.

This scree plot represents Eigenvalues by the number of factors.

Table 2.

Exploratory factor analysis of mental health questionnaires

	Factor loading
Question	1	2	Item rest correlationa
PHQ-9		0.77	0.83
STAI		0.82	0.92
STAXI		0.92	0.65
FABQ-PA	0.63		0.56
FABQ-W	0.85		0.63
PCS	0.67		0.80
TSK-11	0.88		0.84
PASS	0.82		0.89
PSEQ	-0.88		0.91
SER	-0.67		0.71
CPAQ	-0.84		0.85

Exploratory factor analysis identifies groupings of variables (factor loadings) that are statistically and often conceptually unique from one another. The technique extracts the maximum common variance among variables to demonstrate how strongly a group of characteristics are intercorrelated. Strength of the intercorrelation between variables expressed in factors ranges from -1 (perfect negative correlation), to 0 (no correlation), to +1 (perfect positive correlation). All items with a factor loading above 0.5 are shown.

^aCronbach α was calculated separately for items loading on to each factor. PHQ-9 = Patient Health Questionnaire-9; STAI = State-Trait Anxiety Inventory; STAXI = State-Trait Anger Expression Inventory; FABQ-PA = Fear-Avoidance Beliefs Questionnaire physical activity subscale; FABQ-W = Fear-Avoidance Beliefs Questionnaire work subscale; PCS = Pain Catastrophizing Scale; TSK-11 = Tampa Scale of Kinesiophobia; PASS = Pain Anxiety Symptom Scale; PSEQ = Pain Self-Efficacy Questionnaire; SER = Self-Efficacy for Rehabilitation; CPAQ = Chronic Pain Acceptance Questionnaire.

Associations Between Underlying Mental Health Factors and Level of Capability

For patients with knee OA, after accounting for confounders including age, BMI, gender, race, marital status, employment status, insurance type, grade of OA, and worse mood and worse pain coping were independently associated with greater levels of incapability (KOOS JR) in separate models. However, when analyzed in a combined model, pain coping (regression coefficient -4.3 [95% confidence interval -5.4 to -3.2], partial R² 0.076; p < 0.001) had the strongest relationship, and mood was no longer associated (Table 3). Similarly, for hip OA, after accounting for confounders including employment status and grade of OA, worse pain coping (regression coefficient -5.4 [95% CI -7.8 to -3.1], partial R² 0.10; p < 0.001) had the strongest relationship with level of incapability (HOOS JR), and mood was no longer associated (Table 4) (Supplemental Tables 1-6; http://links.lww.com/CORR/B274).

Table 3.

Multivariable linear regression analysis of factors associated with the KOOS JR at baseline, accounting for factor 1 pain coping and factor 2 mood

Variable	Regression coefficient (95% CI)	Standard error	Partial R2	p value
Factor 1: Pain coping	-4.3 (-5.4 to -3.2)	0.56	0.076	< 0.001
Factor 2: Mood	-0.31 (-1.4 to 0.74)	0.53	0.00048	0.56
Kellgren-Lawrence grade	-0.87 (-1.6 to -0.11)	0.38	0.0073	0.02
Age	-0.0099 (-0.14 to 0.13)	0.069	0.000030	0.89
BMI	-0.30 (-0.42 to -0.19)	0.059	0.037	< 0.001
Gender
Women	Reference value
Men	2.3 (0.21 to 4.3)	1.0	0.0067	0.03
Race
White	Reference value
Other	-3.3 (-5.3 to -1.3)	1.0	0.015	0.001
Marital status
Married or partner	Reference value
Divorced or separated	0.66 (-2.1 to 3.5)	1.4	0.00031	0.64
Single	3.4 (0.86 to 6.0)	1.3	0.0097	0.009
Widowed	2.4 (-0.71 to 5.6)	1.6	0.0033	0.13
Employment status
Employed	Reference value
Unemployed	-0.51 (-4.2 to 3.2)	1.9	0.00011	0.79
Retired	1.1 (-1.6 to 3.8)	1.4	0.00095	0.42
Disabled	-2.9 (-7.0 to 1.2)	2.1	0.0028	0.16
Insurance type
Commercial or employer	Reference value
Medicare	-3.1 (-5.7 to -0.39)	1.4	0.0072	0.03
Safety-net	-5.0 (-11 to 0.47)	2.8	0.0046	0.07
Medicaid	3.1 (-2.7 to 8.8)	2.9	0.0016	0.29
Other	-2.8 (-10 to 4.5)	3.7	0.00083	0.45
Primary total joint replacement
No	Reference value
Yes	-3.6 (-6.2 to -0.97)	1.3	0.010	0.007

Adjusted R² = 0.25; that is, this combination of variables explains 25% of the variance in KOOS JR at baseline. KOOS JR = Knee Disability and Osteoarthritis Outcome Score for Joint Replacement.

Table 4.

Multivariable linear regression analysis of factors associated with the HOOS JR at baseline, accounting for factor 1 pain coping and factor 2 mood

Variable	Regression coefficient (95% CI)	Standard error	Partial R2	p value
Factor 1: Pain coping	-5.4 (-7.8 to -3.1)	1.2	0.10	< 0.001
Factor 2: Mood	0.51 (-1.8 to 2.9)	1.2	0.0010	0.67
Kellgren-Lawrence grade	-0.41 (-2.0 to 1.2)	0.82	0.0013	0.61
Employment status
Employed	Reference value
Unemployed	2.5 (-7.4 to 12)	5.0	0.0013	0.62
Retired	4.0 (-0.16 to 8.1)	2.1	0.018	0.06
Disabled	-9.4 (-17 to -1.6)	4.0	0.029	0.02
Primary total joint replacement
Yes	Reference value
No	-5.3 (-9.6 to -1.0)	2.2	0.030	0.02

Adjusted R² = 0.23; that is, this combination of variables explains 23% of the variance in HOOS JR at baseline.

Discussion

Mental health needs may be present in up to 50% of people seeking care for musculoskeletal conditions, including OA [5, 7, 8]. Although a range of validated mental health surveys have been used to better understand the associations between an individual’s mindset and levels of comfort and capability, knowledge gaps remain regarding which specific mental health needs should be prioritized and addressed in clinical practice. In this study, we collected information on numerous aspects of mental health using a single multidimensional mental health survey, performed statistical groupings of these variables into several factors, and simplified these groupings into two conceptually and statistically distinct underlying mental health factors: unhelpful thoughts regarding pain (or pain coping) and symptoms of anxiety and depression (or mood). After controlling for relevant and important sociodemographic variables, we found that pain coping was dominant. This suggests that musculoskeletal clinicians should pay greater attention to improving coping strategies when treating patients to improve mental health [55], perhaps by adding measures of unhelpful thinking, such as the Pain Catastrophizing Scale or Tampa Scale for Kinesiophobia, in evaluating new patients and those considering surgery.

Limitations

First, we performed this study at a single institution in the setting of a comprehensive OA management program that is not yet commonplace in current musculoskeletal care in the United States. Although this may affect the generalizability of our findings, such clinical settings are also well-suited to perform this type of work (that is, where mental health surveys and patient-reported outcome measures are routinely administered as a standard of care, with supportive therapies and services at hand to manage elevated symptoms of distress). Second, our analysis was retrospective and only included patients who were sufficiently concerned about their symptoms to seek care, rather than the general OA population. Nevertheless, our findings remain highly relevant to clinicians treating a diverse patient population with joint pain accessing specialty care. Third, patients may not be forthright about their responses to questions concerning distress, especially when receiving multiple questions related to different forms of distress, leading to partial survey completion or floor effects. Other nuances related to the subjective measurement of mental health include patients perceiving truthful responses to questions about distress, precluding them from receiving treatments such as joint injections or surgery. We anticipate these issues are relatively minor in this dataset because patients are primed before entering the comprehensive OA management program about the importance of completing patient-reported outcome measures as a valuable way for the clinical team to understand the impact of the patient’s condition in the context of their daily lives. Notably, using the OSPRO-YF—a single, brief, multidimensional survey that requires a response for all items—also reduces the risk of partial completion.

In addition, the exploratory factor analysis is one statistical approach to consolidating multiple measures of mental health and mitigating the risk of multicollinearity. Underlying factors generated using this approach, which involves a level of subjectivity in selecting the number of models, may exhibit some degree of collinearity. Future work involving techniques such as ridge regression modeling (a model-tuning method used to analyze data prone to multicollinearity) may further limit statistical overlap. Fifth, our data were cross-sectional, and no inferences can be made regarding the directionality of associations or causality. Future studies could test the efficacy of mental health interventions in the comprehensive care of patients with hip and knee OA. Finally, our sample size for the hip OA population was substantially smaller than the knee OA subset, and we did not conduct separate exploratory factor analyses for hip and knee OA. We deem it unlikely that these subsets of patients have notable differences in underlying mental health factors, but future studies could test this hypothesis. A combined cohort also allows for generalizability across many lower extremity orthopaedic practices. In the same vein, we included patients in the analysis regardless of treatment modality to provide a generalizable sample. Treatment modality was included as an explanatory variable in our models because symptom severity is known to guide operative and nonoperative treatment decisions [56], with treatment modality also serving as a good gauge of effect size in relation to the associations highlighted in this study.

Underlying Mental Health Factors in Patients With Knee and Hip OA

Most studies involving the assessment of mental health factors in musculoskeletal populations focus on symptoms of depression and anxiety [49]. The exploratory factor analysis allowed us to simplify multiple mental health characteristics beyond these symptoms, using the OSPRO-YF tool to generate two statistically and conceptually distinct factors: mood (including symptoms of depression or anxiety) and pain coping (thoughts and misconceptions regarding pain). Our findings are aligned with other studies involved in the discernment of variability in mental health among patients with OA [20, 31, 44]. Lentz et al. [34] identified four statistical mental health groups in a large population with OA using latent class analysis: a group defined by high overall distress (aligned with our negative mood group), one defined by misconceptions about pain (aligned with our negative pain coping group), one showing very low levels of psychologic distress, and one representing individuals with greater self-efficacy and low pain acceptance. Although that study used distinctly different methods to identify statistical groupings (as opposed to distinct factors that characterize a group of questionnaire items), we observed some divergence from their results in our study. Specifically, our analysis found that positively framed surveys (including the Pain Self-Efficacy Questionnaire, Self-Efficacy for Rehabilitation, and the Chronic Pain Acceptance Questionnaire) were grouped with negatively framed surveys (such as the Fear Avoidance Beliefs Questionnaire physical activity subscale, Fear Avoidance Beliefs Questionnaire work subscale, Pain Catastrophizing Scale, and Tampa Scale for Kinesiophobia) related to negative thoughts regarding pain. Conversely, in the study by Lentz et al. [34], a small but distinct statistical grouping, defined primarily by low scores on positively framed surveys that measured pain self-efficacy and acceptance, was identified [34]. These slight differences in results underscore the conceptual difference between using exploratory factor analysis to identify discrete domains measured by questionnaire items and using subgrouping techniques to identify common groups of patients with similar qualities based on their questionnaire answers (such as phenotyping). The former is helpful for understanding what a questionnaire is measuring (that is, the underlying domains, as identified in the current study). The latter subgrouping approach is more informative for understanding how questionnaires can be used to statistically group patients. Ultimately, statistically distinct constructs that load onto separate factors, such as in this study, are important to distinguish and translate into clinical practice.

Associations Between Underlying Mental Health Factors and Level of Capability

Factors reflecting mood and thoughts regarding pain (or pain coping) have a strong basis in the evidence for independently explaining a substantial proportion of the variation in levels of capability among patients seeking care for knee and hip OA [8, 12, 21]. In our study, when these factors were considered together, the relationship between capability and pain coping dominated. Recognizing pain coping as a major underlying mental health factor associated with the magnitude of incapability in people seeking care for OA further reinforces the need to incorporate this aspect of mindset into whole-person models of musculoskeletal care [2, 4, 29, 34]. Moreover, this finding emphasizes the importance of routinely incorporating measures of pain coping in orthopaedic care, which commonly focuses on assessing symptoms of depression and anxiety alone. The Pain Catastrophizing Scale or Tampa Scale for Kinesiophobia are validated instruments that have been extensively used in musculoskeletal studies. One of these tools, or the OSPRO-YF itself, can be administered at the initial visit to identify and respond to an individual’s level of pain coping. These tools are easy to use and can be programmed into the electronic health record for scoring and interpretation. One benefit of the OSPRO-YF is that in addition to score estimates for each of the 11 full-length psychologic questionnaires, it indicates which of the score estimates is in the highest quartile of population scores for those questionnaires. This function flags high scores, essentially notifying busy clinicians which psychologic constructs (including measures of coping) are of the highest concern [52].

The recommendation to screen for pain coping is supported by our prior work showing that foregoing an assessment of pain coping could result in failure to identify nearly one-fourth of patients with OA who have high levels of pain-related psychologic distress [34]. Based on the findings in this study, instruments measuring pain coping should augment rather than replace those screening for symptoms of depression and anxiety, which are still strongly associated with levels of capability and may preclude orthopaedic treatment, warrant referral to a psychologist, or even necessitate urgent care in patients with a high suicide risk [5, 6, 17, 28, 34]. Pain coping needs could be addressed through referral to behavioral health specialists for engagement in cognitive behavioral-based treatments. These patients may also be good candidates for psychologically focused treatments delivered by nonpsychologists, such as physical therapists who supplement their care with cognitive behavioral-based strategies (psychologically informed care) or through effective technology-enhanced delivery modes such as application-based cognitive behavioral therapy or mindfulness [11, 42]. There is detailed practical guidance on the implementation and integration of multidimensional psychologic screening tools in clinical practice [11, 39, 40, 51].

Our work also provides some direction about implementing a set of self-reported measures of mental health while minimizing survey fatigue and avoiding measurement redundancy. Based on the high levels of multicollinearity and statistical overlap observed among mental health measures, it may be of limited clinical benefit—and indeed inadvisable—to include multiple pain coping constructs such as catastrophizing, fear avoidance, and kinesiophobia. Because mood constructs such as depression and anxiety are statistically and conceptually distinct from pain coping, we recommend a separate assessment of these characteristics. The development of shorter measures of pain coping that maintain precision is required to further reduce the responder burden and mental health assessment.

Although thoughts and feelings are distinct entities, unhelpful feelings may be a byproduct of unhelpful thoughts and vice versa. This interaction warrants further study. Cremers et al. [10] showed that aspects of negative mood accentuated the role of negative pain coping in the relationship between pain intensity and level of capability. In other words, symptoms of depression or anxiety amplify the interaction between misconceptions about pain with the magnitude of capability. Such work further reinforces the need for an assessment of both pain coping and mood.

Conclusion

Our current study underscores the presence of two statistically and conceptually distinct mental health factors among people seeking comprehensive care for OA: a pain coping factor defined by characteristics such as fear avoidance, kinesiophobia, and self-efficacy, and a mood factor defined by characteristics such as symptoms of depression and anxiety. Musculoskeletal clinicians should assess pain coping at the time of the initial examination using validated tools such as the OSPRO-YF, Pain Catastrophizing Scale, or Tampa Scale for Kinesiophobia, because this mental health factor is dominantly associated with worse capability among people seeking care for OA. Surgeons should respond to patients who display unhelpful thinking on these validated tools by offering evidence-supported options to improve coping, which may include referrals to a psychologist, a specially trained physical therapist who delivers cognitive behavior-based treatments, or self-directed, application-based mental health–focused interventions. Future research should study the predictive validity of these factors and work to establish the causal relationship among aspects of mindset and capability.

Supplementary Material

abjs-482-633-s001.docx ^{(99.1KB, docx)}