Importance of Outcome Domain for Patients With Musculoskeletal Pain: Characterizing Subgroups and Their Response to Treatment.

BACKGROUND: Patient-centered care models allow for the ability to tailor treatment to outcomes of importance to patients.
OBJECTIVE: The purposes of this study were to (1) define patient subgroups based on outcomes of importance; (2) determine whether patient subgroups based on outcomes of importance differed in demographic, clinical, and psychological measures; and (3) determine whether outcome domain subgroups differed in treatment responses.
DESIGN: This was a prospective, longitudinal observational study.
METHODS: This was a secondary analysis of the Optimal Screening for Prediction of Referral and Outcome study. Patients in the development phase completed the Patient Centered Outcome Questionnaire (PCOQ) and questionnaires related to negative mood, fear avoidance, and positive coping, as well as region-specific questionnaires. Patients in the validation cohort completed the PCOQ, measures of treatment satisfaction and region-specific questionnaires at 4 weeks, 6 months, and 1 year. A hierarchical agglomerative cluster analysis identified profiles based on importance outcomes as determined by the PCOQ. Analysis of variance and chi-squared assessed baseline subgroup differences in demographics, psychological factors, and clinical outcomes. Repeated-measure analysis of variance considered subgroup differences in outcomes longitudinally.
RESULTS: Cluster analysis identified 3 subgroups: (1) "Multiple Outcome Domains Important" subgroup characterized by high importance attached to improvement in all domains, (2) "Pain and Function Outcomes Important" subgroup characterized by high importance attached to improvement in pain and interference, and (3) "Pain Important" subgroup characterized by greatest importance attached to improvement in pain. LIMITATIONS: Our sample included patients from outpatient physical therapy and may not be representative of patients in other settings.
CONCLUSION: Patients can be characterized by the importance attached to improvement in outcome domains. The identified subgroups differed in baseline measures as well as response to treatment.

Nearly one-third of individuals in the United States develop chronic pain,, with annual costs exceeding those of cancer, heart disease, and diabetes. Despite advances in medical technology, the burden of chronic pain is not lessening, and pharmacological approaches and specifically opioid prescription have increased in well-intentioned efforts to manage chronic pain., However, these approaches have not only been ineffective but also carry risks, including addiction and death., Currently, nonpharmacological approaches are recommended as the initial treatment option for chronic pain. While inarguably safer than opioids, nonpharmacological treatment approaches have limitations as individual interventions have small to moderate effect sizes., A patient-centered delivery system focused on mutuality or shared decision-making may represent an effective nonpharmacological intervention for individuals experiencing pain by increasing treatment effects and maximizing outcomes through active patient involvement.

Patient-centered care is an integral part of a biopsychosocial approach to treatment accounting for the patient’s perspectives as the centerpiece in clinical decision-making. Both the practitioner and the patient are active participants in directing treatment in patient-centered care, contrasting the traditional model in which patients are expected to rely on the practitioner to drive the treatment decision-making. Patient-centered care focuses on “informed choice” rather than the conventional practice of informed consent and, when applied in such a manner, is associated with a number of positive outcomes, including improved patient knowledge, and decision-making,,, overall satisfaction with care,,,, and quality of life. Subsequently, a patient-centered approach to nonpharmacological pain management could result in larger treatment effects than have been observed to date from clinical trials.

Patient-centered care models include consideration of the patient’s goals for treatment; however, health care providers do not consistently follow these models. Historically, practitioners have often assumed patient goals for treatment are similar. However, many patients may have individualized preferences for treatment outcomes. For example, patients with low back pain who sought physical therapist intervention endorsed goals in both physical and psychological outcome domains, and their specified recovery goals were not related to conventional rehabilitation treatment targets such as pain, strength, and range of motion. Subsequently, traditional clinical outcome measures often favored by health care practitioners may not align with outcomes of importance to the patient., The traditional, one-size-fits-all approach to clinical outcomes is not ideal for personalizing treatment to specific outcomes valued by the patient. Subsequently, a better understanding of characterizing patients by outcome domains of importance may further inform the clinical decision-making process for delivering patient-centered care.

The objective of this secondary analysis of patients seeking outpatient physical therapy care for musculoskeletal pain was threefold: (1) to define patient subgroups based on importance of improvement for selected outcome domains; (2) to determine whether outcome domain subgroups differed in key demographic, clinical, and psychological measures; and (3) to determine whether these subgroups differed in treatment outcomes from a physical therapy episode.

Methods

This was a secondary analysis involving participants from the development phase (cross-sectional) and validation phase (longitudinal) of the Optimal Screening for Prediction of Referral and Outcome (OSPRO) cohort study., The OSPRO cohort study is a specific project within the Orthopaedic Physical Therapy Investigative Network designed to develop standardized tools to assist physical therapists in clinical assessments. The OSPRO cohort study was approved by the University of Florida Human Subjects Institutional Review Board (IRB-01), and all participants provided informed consent to participate in the study.

A convenience sample was recruited from patients with complaints of neck, shoulder, low back, or knee pain seeking care from outpatient physical therapist clinics, including 330 individuals from the development phase and 440 participants recruited from the validation phase. The OSPRO development cohort was recruited from 3 outpatient physical therapy clinics in the University of Florida Health system in Gainesville, Florida, and 8 outpatient physical therapy clinics in the Brooks Rehabilitation health system in Jacksonville, Florida. Sites were selected to represent different socioeconomic strata as well as urban and rural communities. The OSPRO validation was recruited from 9 outpatient physical therapy clinics representing multiple health systems from the Mideast, Southeast, Great Lakes, Rocky Mountain states, and Far West regions of the United States. The development and validation phases used the same inclusion criteria. Inclusion criteria included patients: (1) aged between 18 and 70 years old, (2) seeking outpatient physical therapy for musculoskeletal pain, and (3) able to read and understand the English language. Exclusion criteria were patients: (1) diagnosed with a chronic pain syndrome (eg, fibromyalgia or Crohn Disease), (2) who had neuropathic pain syndrome (eg, complex regional pain syndrome), (3) with psychiatric history (either under the care of a mental health provider or using psychiatric prescriptions), (4) who had active cancer, and (5) with neurological disorders (eg, Parkinson’s or spinal cord injury).

At the initial intake physical therapist session, patients completed a demographic questionnaire that included age, sex, race, ethnicity, employment, whether in litigation due to current situation, marital status, educational level, insurance provider, self-reported health status, and surgical history. Additionally, patients were instructed to complete questions on historical data based on current symptoms. These questions included symptom duration, anatomical location of pain, symptom onset, previous treatment intervention, and number of prior incidents of pain.

Patients completed the Patient Centered Outcome Questionnaire (PCOQ) (Appendix). The PCOQ is a 5-item questionnaire. Separate 101-point numeric rating scales (NRS) from 0 = none to 100 = worst imaginable assess usual, desired, successful, and expected levels of pain, fatigue, emotional distress, and interference. A 101-point NRS (0 = not at all important to 100 = most important) assesses the importance of improvement in each domain. The PCOQ has shown sufficient test-retest reliability over a 48-hour period with correlation coefficients of 0.84 to 0.90 for usual levels of pain, fatigue, distress, and interference with daily activities, as well as 0.62 to 0.82 for importance ratings for pain, emotional distress, and interference., Additionally, the PCOQ has demonstrated good concurrent validity with measures associated with pain and disability, including visual analog scale measures of pain intensity (r = 0.52–0.78) and pain unpleasantness (r = 0.64–0.73) as well as the Pain Disability Index (r = 0.75) and Roland Disability Questionnaire (r = 0.69).,

Patients in the developmental phase completed well-established and commonly used questionnaires related to depression (Patient Health Questionnaire [PHQ-9]), fear of movement (Fear Avoidance Belief Questionnaire [FABQ], Tampa Scale of Kinesiophobia [TSK-11], Pain Catastrophizing Scale [PCS], anxiety (Pain Anxiety Symptom Scale [PASS]), pain and rehabilitation self-efficacy (Pain Self Efficacy Questionnaire [PSE], Self-efficacy for Rehabilitation Outcome Scale [RSE]), and anger (State-Trait Anger Expression Inventory [STAXI]) as well as region specific questionnaires (ie, Oswestry Disability Index: low back; Neck Disability Index: cervical spine; Quick Disability of Arm Shoulder and Hand: shoulder; and International Knee Documentation Committee Subjective Knee Form: knee). Patients in the validation (longitudinal) cohort completed the same associated region-specific disability questionnaires (Tab. 1).

Table 1

Outcome Measures Questionnaires Completed in the Developmental Phase

Questionnaires	Description	Scale	Reliability
Negative mood questionnaires
Patient Health Questionnaire	Assesses level of depressive symptoms	0–27(Higher scores signify increased depressive symptoms)29	α = 0.8429
State-Trait Anger Expression Inventory	Assesses level of anxiety symptoms	20–80(Higher scores signify increased anxiety levels)31	α > 0.8930,31ICC = 0.86 to 0.9531,32
Fear avoidance questionnaires
Fear-Avoidance Belief Questionnaire	Assesses level of fear-avoidance beliefs2 Subscales Physical Activity Subscale (PA) 4 itemsWork Subscale(W) 7 items	0–24: FABQ-PA0–42: FABQ-W(Higher scores signify increased fear avoidance beliefs in both subscales)33	FABQ-PAα = 0.7033ICC = 0.7734FABQ-Wα = 0.8833ICC = 0.9034
Tampa Scale of Kinesiophobia	Assesses level of pain-related fear of movement	11 items(Higher scores signify increased pain-related fear of movement and fear of injury)35	α = 0.7935ICC = 0.8135
Pain Catastrophizing Scale	Assesses pain catastrophizing (“an exaggerated orientation towards pain”)	13 items(Higher scores signify higher levels of pain catastrophizing)36	α = 0.8736ICC = 0.9337
Pain anxiety Symptoms Scale	Assesses pain-related anxiety	20 items(Higher scores signify increased levels of pain-related anxiety)38,39	α = 0.9138,39
Positive affect/coping questionnaires
Pain Self-Efficacy Questionnaire	Assesses strength and generality of patient’s beliefs about their ability to accomplish activities despite pain	10 items(Higher scores signify stronger self-efficacy beliefs)40	α = 0.9340
Self-Efficacy for Rehabilitation Outcome Scale	Assesses self-efficacy associated with performing various rehabilitation tasks	12 items(Higher scores signify higher elevated levels of self-efficacy during rehabilitation41	α = 0.9441
Region-specific disability questionnaires
Oswestry Disability Index	Assesses low back dysfunction on individual’s pain and functional activities	10 items6-point rating scale(Higher scores increased low back dysfunction)44	α = 0.71–0.8342,43ICC = 0.84–0.9144–48
Neck Disability Index	Assesses disability in patients with neck pain	10 items6-point rating scale(Higher scores related to greater neck disability)50	α = 0.8049ICC = 0.5050
Quick Disability of Arm Shoulder and Hand	Assesses ability of individuals to perform upper extremity activities	11 items5-point rating scale(Higher scores signify greater level of upper extremity disability52	α = ≥ 0.9251ICC ≥ 0.9451
International Knee Documentation Committee Subjective Knee Form	Assesses an individual’s knee function	18 items(Higher scores signify increased knee function)55	ICC = 0.9453,54

Patients in the validation phase completed follow-up at 4 weeks, 6 months, and 1 year that included the same questionnaires. Furthermore, satisfaction with treatment was assessed at 6 months and 1 year through the following questions: (1) “If you had to spend the rest of your life with the symptoms you have right now, how would you feel about it?” (1, very dissatisfied to 5, very satisfied); (2) “Would you have the same physical therapy treatment again if you had the same condition?” (1, definitely not to 5, definitely yes); and (3) “How would you rate the overall results of your physical therapy treatment?” (1, terrible to 6, excellent). The first question is suggested to reflect satisfaction with treatment outcomes while the latter 2 questions reflect satisfaction with the treatment process.

All initial and follow-up data collection occurred online and was deidentified. All responses were self-report and completed electronically in a web-based electronic records database (REDCap; Vanderbilt University, Nashville, TN, USA) by the patients.

Data Analysis

All data analyses were performed using SPSS, version 24.0 (IBM Corp, Armonk, NY, USA).

Identification and Characterization of Domain Importance Subgroups

Cluster analysis is a statistical approach allowing categorization of individuals into distinct groups. We conducted an exploratory hierarchical agglomerative cluster analysis using Ward clustering method and squared Euclidean distances to identify unique subgroups based on outcome domains of importance as identified by the PCOQ from the development phase of the OSPRO study. The cluster solution was identified based on statistical and theoretical criteria informed by our prior studies of the PCOQ., We examined the resultant dendrogram and schedule of agglomeration coefficients to determine the optimal solution between 2 and 4 clusters based on percent change between cluster solutions as well as the plot characteristics. Once the number of clusters was identified, we performed a one-way analysis of variance (ANOVA) with Bonferroni post hoc correction to descriptively identify the composition of the cluster groups from the domains of importance.

ANOVA models for continuous measures and chi-squared levels for categorical measures were used to assess subgroup-related baseline differences in demographic factors, type of medical insurance, psychological factors, expectations for treatment, success criteria for treatment, and usual levels of pain, fatigue, emotional distress, and interference.

Cluster Validation Across Samples

There are several ways statistically to determine the validity of a cluster solution when only 1 sample is available (eg, holdout, k-fold, or leave-one-out cross validation methods). For this study, we had access to 2 independent samples and had an a priori plan to reproduce cluster solutions across these 2 separate samples. Subsequently, we performed the same cluster analysis for PCOQ outcome domains of importance for the development and validation cohorts of the OSPRO study. The same analytical approach was used for both cohorts, but independent cluster solutions were determined for each cohort (ie, there was no attempt to control the cluster solution for the validation cohort based on the results of the development cohort). In this manner, we report on the validity of cluster solutions across different samples in this study rather than further investigating the validity of these cluster solutions within a given cohort.

Outcome Importance Subgroup Response to Treatment

Cluster solutions from the validation cohort were then used to create “outcome importance” subgroups. Repeated-measure ANOVA considered subgroup-related differences in clinical outcomes from baseline to 4 weeks, 6 months, and 12 months. Chi-square considered subgroup-related difference in treatment satisfaction at 6 months and 12 months.

Role of the Funding Source

The funder played no role in the design, conduct, or reporting of this study.

Results

Identification and Characterization of Subgroups

Subgroups. A total 330 patients enrolled in the OSPRO development cohort and completed the PCOQ. Demographic, psychological, insurance type, and clinical measures are presented in Table 2. A 3-cluster solution was observed for importance of outcome domains (Fig. 1).

Table 2

Characteristics of Clusters Derived From OSPRO Developmental Study[

Characteristic	OSPRO Developmental Study (n = 330)	Cluster 1: PFI (n = 93)	Cluster 2: MDI (n = 155)	Cluster 3: PI (n = 82)
	Mean ± (SD)	Mean ± (SD)	Mean ± (SD)	Mean ± (SD)
Affected region
Neck	21.5%	20.4%	21.9%	26.8%
Low back	27.5%	28.0%	25.2%	31.7%
Shoulder	25%	21.5%	27.1%	19.5%
Knee	26%	30.1%	25.8%	22.0%
Duration (d)	382.50 (993.63)	381.01 (962.47)	297.41 (724.74)	499.04 (1210.78)
Surgical status (% surgical)	24%	20.7%	28.4%	25.6%
Age	44.62 (15.54)	42.34 (16.05)	46.45 (14.65)	44.74 (15.06)
Sex (% female)	60.5	55.9	61.3	56.1
Race
American Indian or Alaska Native	0.5%	0.0%	0.7%	1.2%
Asian	3.1%	3.3%	2.6%	1.2%
Black or African American	21.7%	14.1%	22.5%	26.8%
White	74.7	82.6%	74.2%	70.7%
Education
Less than high school	4.7%	2.2%	5.2%	7.3%
Graduated from high school	13.3%	16.1%	15.5%	7.3%
Some college	32.6%	22.6%	36.1%	34.1%
Graduated from college	26.0%	24.7%	22.6%	29.3%
Some post graduate course work	7.0%	8.6%	5.2%	6.1%
Completed post graduate degree	16.4%	25.8%	15.5%	15.9%
Income
<$20,000	23.4%	21.3%	20.3%	28.0%
$20,000–$35,000	13.7%	10.1%	17.0%	13.4%
$35,001–$50,000	11.1%	15.7%	9.8%	11.0%
$50,001–$70,000	14.0%	6.7%	18.3%	14.6%
>$70,000	37.8%	46.1%	34.6%	32.9%
Insurance
Private	60.1%	77%	54.1%	60.2%
Medicare	14.2%	9.8%	14.5%	16.3%
Medicaid	11.8%	1.6%	13.4%	15.3%
Workers Compensation	3.6%	4.9%	4.1%	2.0%
Disability	0.9%	0.0%	1.2%	1.0%
Uninsured	1.5%	1.6%	1.7%	1.0%
Other	7.9%	4.9%	11.0%	4.1%
PHQ-9	4.96 (5.53)	3.59 (4.36)	6.04 (5.92)	4.36 (5.95)
FABQ-pa	13.86 (6.09)	13.68 (5.80)	14.64 (5.90)	12.24 (6.06)
FABQ-w	10.98 (11.94)	9.29 (11.30)	12.83 (12.84)	8.90 (11.04)
PCS	13.06 (12.32)	10.81 (10.89)	14.89 (13.14)	11.33 (13.11)
TSK-11	22.38 (6.70)	22.10 (6.36)	23.19 (6.59)	21.46 (7.39)
PASS	24.44 (20.41)	22.13 (17.21)	27.59 (22.02)	20.26 (20.80)
PSEQ	43.23 (14.26)	45.38 (13.48)	39.99 (14.82)	45.98 (15.18)
SER	104.71 (20.30)	106.82 (18.08)	102.56 (21.52)	107.06 (20.22)
STAXI	15.31 (4.81)	15.58 (4.89)	15.83 (4.98)	14.83 (5.29)
PCOQ usual
Pain	43.68 (28.36)	46.25 (26.87)	45.51 (28.35)	37.09 (29.28)
Fatigue	33.32 (28.57)	31.91 (28.06)	38.77 (29.14)	24.33 (26.24)
Emotional distress	22.28 (28.00)	17.87 (25.46)	28.15 (30.36)	15.59 (23.91)
Interference	35.89 (31.92)	36.42 (28.28)	41.21 (32.88)	25.67 (32.13)
PCOQ % Success
Pain	61.41 (68.43)	66.08 (28.49)	57.52 (92.98)	63.12 (40.31)
Fatigue	53.19 (56.76)	54.01 (64.68)	48.07 (46.89)	54.78 (45.86)
Emotional distress	49.23 (85.23)	50.95 (98.61)	43.97 (47.72)	47.27 (66.77)
Interference	63.26 (41.77)	64.29 (43.46)	61.85 (42.56)	65.63 (38.03)
PCOQ % Expected
Pain	65.02 (67.38)	71.40 (33.96)	67.75 (39.10)	52.25 (120.08)
Fatigue	61.62 (44.10)	62.29 (43.22)	59.09 (48.80)	61.94 (43.58)
Emotional distress
Interference	68.59 (49.91)	62.50 (68.20)	70.69 (38.39)	75.64 (34.41)

MDI = Multiple Outcome Domain Important; OSPRO = Optimal Screening for Prediction of Referral and Outcome; PFI = Pain and Function Outcome Important; PI = Pain Outcome Important.

Figure 1

Cluster ratings of importance for different domains of pain from the OSPRO developmental study. Y axis = importance attributed to improvement in each domain of pain with 0 (not at all important) to 100 (most important). MDI = multiple domains important cluster; PFI = pain and function important cluster; PI = pain important cluster.

Cluster 1 (n = 93) was labeled Pain and Function Outcomes Important (PFI) and was characterized by high scores for the importance of improvement in the domains of pain and interference. Cluster 2 (n = 155) was labeled Multiple Outcome Domains Important (MDI) and was characterized by high scores for the importance of improvement in all domains. Cluster 3 (n = 82) was labeled Pain Outcome Important (PI) and was characterized by generally low scores for the importance of improvement in all domains with the highest importance attached to pain (Fig. 2).

Figure 2

PCOQ importance of outcome clusters for patients seeking outpatient physical therapy.

The clusters differed significantly in terms of importance attached to changes in the domain of pain (P < .01), with PFI (P < .01) and MDI (P < .01) both indicating this as significantly more important than PI. The clusters differed significantly in terms of importance attached to changes in the domain of fatigue (P < .01), with PFI rating this significantly more important than PI (P < .01) and MDI rating this as significantly more important than PFI (P < .01) and PI (P < .01). The clusters differed significantly in terms of importance attached to changes in the domain of emotional distress (P < .01), with PFI rating this significantly more important than PI (P < .01) and MDI rating this as significantly more important than PFI (P < .01) and PI (P < .01). The clusters differed significantly in terms of importance attached to changes in the domain of interference (P < .01), with PFI rating this significantly more important than PI (P < .01) and MDI rating this as significantly more important than PFI (P < .01) and PI (P < .01).

Characterization of Subgroups by Demographic and Psychological Factors

The clusters did not differ in terms of demographic and type of insurance variables. Significant differences were observed in PHQ-9 (P < .01), with MDI reporting significantly higher levels than PFI (P < .01); FABQ-pa (P = .01) with MDI reporting significantly higher levels than PI (P = .01); PCS (P = .02) with MDI reporting significantly higher levels than PFI (0.04); PASS (P = .02) with MDI reporting significantly higher levels than PI (P = .03); and PSEQ (P < .01) with MDI significantly lower than PFI (P = .02) and PI (P = .01).

A total 440 participants enrolled in the OSPRO validation cohort completed the PCOQ. Demographic and clinical measures are presented in Table 3.

Table 3

Characteristics of Clusters From OSPRO Longitudinal Study

Characteristic	OSPRO Longitudinal Study (n = 440) Mean ± (SD)	Cluster 1: PFI (n = 165) Mean ± (SD)	Cluster 2: MDI (n = 178) Mean ± (SD)	Cluster 3: PI (n = 97) Mean ± (SD)
Affected region
Neck	26.8%	26.7%	23.0%	34.0%
Low back	22.3%	17.6%	25.3%	24.7%
Shoulder	24.3%	27.9%	23.6%	19.6%
Knee	26.6%	27.9%	28.1%	21.6%
Duration (days)	398.58 (1715.80)	238.38 (440.05)	379.62 (1200.57)	713.32 (3242.97)
Surgical status (% surgical)	18.9%	13.9%	21.9%	21.6%
Age	45.06 (15.82)	43.60 (15.55)	47.53 (15.42)	42.95 (16.58)
Sex (female)	62.6%	56.7%	70.2%	58.8%
Race
American Indian or Alaska Native	0.7%	1.3%	0.6%	0.0%
Asian	5.8%	6.3%	5.1%	6.2%
Black or African American	14.3%	5.6%	19.3%	19.6%
White	79.2%	86.9%	75%	74.2%
Education
Less than high school	2.5%	1.2%	3.4%	3.1%
Graduated from high school	8.8%	4.3%	10.3%	13.5%
Some college	25.8%	21.5%	26.9%	31.3%
Graduated from college	27.6%	27.6%	29.1%	25.0%
Some post graduate course work	12.9%	17.8%	10.3%	9.4%
Completed post graduate degree	22.4%	27.6%	20.0%	17.7%
Income
<$20,000	15.8%	10.1%	18.2%	22.1%
$20,000–$35,000	14.2%	12.1%	12.8%	20.8%
$35,001–$50,000	13.4%	16.1%	11.5%	11.7%
$50,001–$70,000	15.0%	16.8%	14.2%	13.0%
>$70,000	41.7%	45.0%	43.2%	32.5%
Insurance
Private	65.9%	79.0%	58.3%	57.3%
Medicare	12.6%	10.8%	12.5%	15.7%
Medicaid	4.6%	0.6%	8.9%	3.4%
Workers compensation	3.4%	0.6%	4.2%	6.7%
Disability	1.0%	0.6%	1.2%	1.1%
Uninsured	1.7%	1.3%	2.4%	1.1%
Other	10.9%	7.0%	12.5%	14.6%
PCOQ usual
Pain	43.54 (28.56)	42.09 (24.85)	50.81 (30.57)	32.66 (27.01)
Fatigue	36.21 (28.84)	36.17 (25.69)	42.86 (31.70)	24.09 (24.37)
Emotional distress	24.63 (28.29)	22.78 (2.13)	32.56 (2.05)	13.20 (2.78)
Interference	37.39 (30.73)	37.53 (26.42)	47.53 (33.17)	18.53 (23.40)
Average pain	4.22 (1.98)	3.91 (1.82)	4.69 (2.12)	3.90 (1.81)
Baseline disability (z- score)	0.00 (1.00)	−0.06 (.87)	0.20 (1.08)	−0.27 (.97)
PCOQ% success
Pain	69.30 (26.83)	68.28 (23.59)	73.40 (26.24)	63.62 (32.00)
Fatigue	60.13 (33.59)	55.47 (33.73)	66.31 (29.94)	57.38 (38.19)
Emotional distress	54.68 (61.46)	52.66 (35.86)	57.76 (83.57)	51.68 (41.91)
Interference	66.37 (49.89)	67.53 (41.34)	69.89 (39.62)	55.43 (80.03)
PCOQ %expected
Pain	62.45 (52.04)	63.20 (51.14)	61.23 (59.07)	63.43 (37.88)
Fatigue	59.01 (47.04)	59.10 (37.71)	58.49 (57.49)	59.81 (40.30)
Emotional distress	47.14 (91.42)	50.49 (47.17)	44.66 (127.35)	45.63 (51.43)
Interference	64.30 (66.95)	69.16 (41.17)	60.12 (90.23)	63.12 (47.03)
PCOQ importance
Pain	84.66 (29.43)	93.68 (9.49)	97.91 (5.13)	44.99 (41.28)
Fatigue	63.12 (39.41)	51.94 (33.20)	97.38 (5.74)	19.25 (28.68)
Emotional distress	54.58 (43.41)	33.33 (33.25)	98.38 (3.94)	10.35 (19.77)
Interference	75.18 (37.20)	88.81 (14.15)	98.57 (3.94)	9.05 (14.28)

Cluster Reproduction

A similar 3-cluster solution was observed (Fig. 3). Cluster 1 (n = 165) was labeled PFI and was characterized by high scores for the importance of improvement in the domains of pain and interference. Cluster 2 (n = 178) was labeled MDI and was characterized by high scores for the importance of improvement in all domains. Cluster 3 (n = 97) was labeled PI and was characterized by generally low scores for the importance of improvement in all domains, with the highest importance attached to pain.

Figure 3

Cluster ratings of importance for different domains of pain from the OSPRO longitudinal study. Y axis = importance attributed to improvement in each domain of pain with 0 (not at all important) to 100 (most important). MDI = multiple domains important cluster; PFI = pain and function important cluster; PI = pain important cluster.

Clinical Outcomes for Outcome Importance Subgroups

The clusters differed in terms of average pain rating at baseline (P < .01), with MDI reporting significantly higher levels than PFI (P < .01) and PI (P < .01). Clusters differed significantly in disability (P < .01), with MDI reporting higher disability at baseline than PFI and PI.

A main effect for pain was observed (F(3, 250) = 56.65; P < .01, partial eta squared = 0.41) with significant improvements observed between baseline and 4 weeks (P < .01) and 4 weeks and 6 months (P < .01). Cluster dependent differences in pain response were not observed (F(6, 502) = 2.00; P = .07, partial eta squared = 0.02). An interaction was observed for disability (F(6, 498) = 2.89; P = .01, partial eta squared = 0.03). Deconstruction of the interaction observed significant differences from baseline to 4 weeks between PI and both PFI (P < .01) and MDI (P < .01).

A greater percentage of PFI (P = .02) expressed being somewhat satisfied to very satisfied at 6 months to the questions “If you had to spend the rest of your life with the symptoms you have right now, how would you feel about it?” The clusters did not differ in response to this question at 12 months (P = .56). A greater percentage of PFI (P = .01) indicated probably yes to definitely yes to the question, “Would you have the same PT treatment again if you had the same condition?” at 6 months. The clusters did not differ in response to this question at 12 months (P = .22). The clusters did not differ at 6 months (P = .23) or 12 months (.19) in their responses to “How would you rate the overall results of your PT treatment?”

Discussion

Patient-centered care requires a collaborative approach to both treatment options and goals. Subsequently, understanding which outcomes are important to patients seeking care for pain is an important consideration for shared decision-making. We identified 3 subgroups of individuals presenting for physical therapy with musculoskeletal pain based on the importance attributed to different domains of pain. Prior studies using similar methodology have identified pain-focused and multi-dimension–focused subgroups based on the importance attributed to improvements in outcome domains in patients attending physical therapy and seeking care from a chronic pain clinic. Our findings are similar and add to these by identifying a third subgroup citing both improvement in pain and function as important. Importantly, this 3-group cluster structure was relatively consistent across 2 separate samples with clusters of similar subgroup composition and frequency distribution in each. Future research is needed to confirm these findings and to refine what the expected frequencies are for each of the cluster subgroups. Overall, the PCOQ appears to be a reliable and valid measure for determining the importance of different outcomes domains for patients seeking care for musculoskeletal pain.

Our findings characterize subgroups based on the importance attached to improvements in key outcomes domains. We observed subgroup-related differences in baseline pain, disability, demographic factors, and psychological factors. Specifically, the cluster considering improvement in all domains of pain important (MDI) reported higher baseline levels of pain, disability, depression, fear, catastrophizing, and anxiety as well as lower levels of self-efficacy. These findings added credibility to the derived subgroups as those with the higher levels of pain-related fear and emotional distress valued improvements in related domains. Furthermore, the identification of this subgroup demonstrated psychological factors are associated with both pain and disability. However, the patients themselves indicated high importance in resolving the psychological component along with pain and function.

The derived subgroups did not differ statistically in the 4-week and 6-month pain intensity responses. This was an interesting finding because 1 of the subgroups (PI) was characterized by the importance of pain improvements, but this subgroup did not have larger pain intensity change scores after receiving physical therapy. This finding could question the clinical utility of these subgroups if the sole interest was in predicting pain intensity responses. Our findings indicate that these subgroups may not be useful in distinguishing pain intensity outcomes, but there may be other ways in which their clinical utility can be expressed. For example, knowledge of a particular subgroup could facilitate shared decision-making and give the clinician an idea of which outcome domains should be emphasized during a given treatment episode.

The derived subgroups did differ in other areas of treatment response. The subgroup considering improvements in pain as most important (PI) displayed a worsening in function over the first 4 weeks following the initiation of physical therapy. This effect was negated at 6 months and 1 year. The importance attached by our PI subgroup to improvements in pain do not align with current treatment recommendations advocating for a focus on lessening the impact of pain on function and suffering rather than pain intensity. We did not control for nor document individual physical therapy treatment approaches and are therefore unable to determine the role that any specific treatment approach or modality may play. We observed that patients with a sole focus on lessening of pain intensity had worse short-term functional outcomes than those valuing improvement in multiple dimensions. It is beyond the purposes of this study to determine if matching treatment emphasis with outcome domain importance results in better clinical outcomes for function and patient satisfaction. Future studies are required to replicate these findings and also use designs that allow for better determination of the impact on clinical outcomes of matching treatment approach with outcome domain importance subgroup characteristics.

Patient satisfaction is a core measure in many patient-centered approaches, as it a reflection of treatment delivery or treatment outcome. Furthermore, satisfaction is associated with pain perception, and risk of being disabled 6 to 12 months after injury. Patient satisfaction has been recommended as a core outcome measure for chronic pain, although distinction between treatment delivery and outcome is not part of that recommendation. Making such a distinction is important because levels of patient satisfaction will vary substantially whether the focus is on treatment delivery or treatment outcomes., In this analysis, patients valuing improvement in both pain and function expressed greater satisfaction with both treatment outcomes as well as treatment delivery at 6 months. These findings add to this body of knowledge by suggesting patients valuing improvement in pain and function may have overall greater 6-month satisfaction with physical therapy treatment delivery and outcomes, suggesting that the alignment with pain and function as treatment goals works particularly well for this patient subgroup. Physical therapists traditionally emphasize outcomes related to functional restoration and may not be as comfortable in addressing other domains such as fatigue and emotional distress. In this analysis, the multidimensional subgroup did not have as high 6-month satisfaction ratings, and we speculate that focus on broader treatment approaches (eg, psychologically informed physical therapy) may be necessary to see higher 6-month satisfaction ratings in treatment delivery and treatment outcomes for this patient subgroup.

Clinical Implications

Our findings have clinical implications for physical therapists treating patients seeking care for musculoskeletal pain. Patient-centered care may add value to the treatment process for patients presenting with pain; however, physical therapists do not routinely include the patient in the decision-making process., Subsequently, a structured approach is likely necessary for shared decision-making to be successfully implemented into routine clinical practice. The PCOQ provides a standardized manner to determining outcomes identified as important to the patient. For clinical application, a patient presenting to physical therapy would complete the PCOQ and allow the therapist to determine which outcomes are important and what desired, expected, and successful outcome levels are for the patient. Treatment goals aligned with these outcomes could then be discussed, agreed on, and targeted through the collaborative interaction of practitioner’s expertise and patient’s preference for identified outcomes.,, Additional research is needed to determine the best manner to identify PCOQ importance subgroup membership in routine clinical practice. Future studies should also consider whether such approaches using the PCOQ are more effective in producing shared decision-making and subsequently result in improved outcomes. Furthermore, patients support changes in multiple domains of pain as important when seeking physical therapy for pain. Physical therapists do not consistently assess domains such as emotional distress in their patients with pain, and may lack confidence in addressing these domains. Moreover, accurate assessment of emotional distress requires a systematic approach and cannot be done accurately based on instinct. Psychologically informed physical therapist practice has recently been advocated as a more effective management strategy for patients in pain. Our findings suggest a shared decision-making approach to managing patients presenting to physical therapy with pain may necessitate measuring factors such as emotional distress as well as addressing this domain with treatment as outcomes in this domain are important to a subgroup of patients.

This study has several limitations when interpreting the results. Our results are representative of outpatient orthopedic and sports medicine population and may not be representative or arbitrarily applied to other physical therapy populations. Additionally, our data were collected by means of convenience sampling and may not be indicative of all individuals seeking rehabilitation. In particular, we did not characterize the total relevant patient population from which this convenience sample was derived and are unable to determine what role response bias played for those deciding to participate in this study. We only assessed importance at baseline. Patient-specific factors such as their success criteria for treatment change over time,, and our approach does not allow us to determine if the importance in change patients attach to different outcome domains changes over an episode of care as well. This is an area for future studies utilizing the PCOQ to develop cut-off scores to determine patient classification-based outcomes of importance and how cut-off scores could be used to drive treatment and changing expectations.

Additionally, despite the highest ratings for importance attached to improvements in pain, our PI subgroup expressed relatively low importance for this domain compared with the other subgroups. The PI subgroup also had the lowest importance ratings for the other domains of the PCOQ (ie, fatigue, emotional distress, and interference) and while they had comparable pain intensity ratings at baseline, they also had correspondingly lower ratings on the region-specific disability scores (Tab. 2). Collectively these data support the notion that while the PI subgroup places the highest importance on pain improvement, there are low levels of concurrent distress, interference, and disability. We speculate that the PI subgroup may be seeking care for pain relief, and this is why it is important to them. However, because there are overall low levels of distress, interference, and disability, the overall importance ratings are lower than the other subgroups. An alternative explanation to consider is that the domains of the PCOQ were based on expert consensus for key outcome domains in patients with pain. While indicating the highest importance for improvements in pain, patients in our PI subgroup may place great value and be better identified by the importance attached to change in other domains not reflected by the 4 domains of the PCOQ. Future research is needed to better characterize the PI subgroup and determine how this subgroup influences provider clinical decision-making on treatment approaches.

Conclusion

Grouping patients by the importance of improvement in outcome domains may lead to alternative models of delivering patient-centered care. The MDI, PFI, and PI subgroups identified in this cohort differed in key psychological measures, pain and disability at baseline, as well as in treatment response. These findings suggest the patient’s perspectives on importance of outcome domain may influence clinical decision-making for patients seeking care of musculoskeletal pain conditions.