Development and validation of a method to screen for co-morbid depression by non-behavioral health practitioners treating musculoskeletal pain

2.1Sampling and procedures

The current paper is the byproduct of an observational study conducted with a convenience sample of English-speaking and Spanish-speaking adult participants recruited on admission at three “stand-alone” outpatient physical therapy centers in Southern California. The sponsoring research foundation’s Institutional Review Board approved the experimental protocol and monitored the research activities. All candidates for enrollment provided written informed consent prior to participation. Following completion of the consent procedures, 157 participants age 19 years to 69 (mean, SD = 44.8, 12.5) years with duration of presenting musculoskeletal pain disorder from acute to 47 (mean, SD = 2.4, 4.5) years completed baseline testing. Participants presented with soft-tissue spine impairment (41%), upper extremity impairment (35%), lower extremity impairment (24%). Between baseline testing and the 30-day follow-up testing, all participants continued in the active outpatient physical therapy program for musculoskeletal conditions consisting of rehabilitative exercises, manual therapies, and adjunctive modalities using a measurement-driven approach along with a home exercise program. Typically, each participant participated in eight therapy sessions over the 30-days.

2.2Data acquisition

Data were retained from all consented participants, including those whose data subsequently were excluded from the psychometric research study. At the conclusion of the study, telephone interviews were undertaken with all consented participants to determine work status. Because the participants who were excluded had continued to participate as usual in physical therapy and had given consent for participation, they were included in the telephone follow-up.

2.3Measures

Baseline testing was conducted with three instruments, the Older Adult Health and Mood Questionnaire [24] (OAHMQ), the Multidimensional Task Ability Profile (MTAP) [25], and a 10 cm visual analogue pain scale (VAS) [26]. The OAHMQ was recorded with paper and pen, while the MTAP and the VAS pain scale were recorded using a computer kiosk.

2.3.2Multidimensional Task Ability Profile (MTAP)

A 50-item computer-administered and scored patient-reported outcome measure consisting of drawings with text captions depicting a wide range of physical work tasks and activities of daily living (Fig. 1) with difficulty levels and a rating scale that have been calibrated using the item response theory rating scale method of Rasch analysis [32].

Fig. 1

Sample MTAP items: Panel “A” Carry a full laundry basket up one flight of stairs; Panel “B” Get into an automobile driver’s seat; Panel “C” Use an electric sander to smooth a table top; Panel “D” Write a shopping list with a pencil.

For each item, the MTAP uses a five-level ordinal rating scale (Able, Slightly Restricted, Restricted, Very Restricted, Unable). Items are presented one at a time, starting with items that have the least physical demand. Missed items are not allowed with the computer interface. The mean time to complete the MTAP is approximately 6 minutes to 8 minutes. Once the participant completes the test, the computer automatically scores the responses with this score-weighting rubric: Able = 4, Slightly Restricted = 3, Restricted = 2, Very Restricted = 1, Unable = 0. Weighted item scores are summed over the instrument to calculate a global score ranging from 0 to 200, with higher scores indicating more ability.

2.4Depression screening and triage process

In the original psychometric research study, to avoid contaminating the inter-test comparisons of the patient self-report measures, volunteer patients were screened by a physical therapist and/or a chiropractor for clinical depression, following the flowchart model in Fig. 2.

Fig. 2

Two-step non-behavioral health (NBH) depression screening and triage.

2.5Post-study telephone interview

One month after conclusion of the study, participants received a telephone interview by study personnel to assess work status through two questions: 1) “Were you working when you first started the study?” 2) “Are you working now?” Depending on when the participant concluded study participation, these data were collected 1–9 months (mean±SD: 155±80 days; range: 32–260 days) after visit 3.

4.1Demographic comparisons

A series of one-way analyses of variance comparing participants who were identified as experiencing depressive illness with non-depressed participants found no differences on demographic characteristics for age, the duration of disability, or body mass index (all p > 0.05). Chi-square analyses also found no differences on gender, native language, and whether or not the person was working at baseline or at follow-up (all p > 0.05).

4.2Depressed versus nondepressed participants

To examine the comorbidity of musculoskeletal pain and depression in this sample, a series of one-way analyses of variance were conducted, with groupwise results presented in Table 2.

These analyses found the expected difference on the OAHMQ score (F_1,155 = 178.54, p < 0.001), along with significant differences for scores on the VAS score (F_1,154 = 4.11, p = 0.044), and the MTAP score (F_1,154 = 12.54, p = 0.001). Depressed participants had higher VAS scores and lower MTAP scores.

4.3Work status

Of the 136 participants for whom both pre-treatment and post-treatment work status data were collected, 60 were working prior to treatment and 67 were working at follow-up. A chi-square analysis revealed no categorical trends in terms of gender or native language at either time (both p > 0.05).

To examine the effect of depressive symptoms on work status outcomes, grouping the comparison of outcomes in terms of the clinical value was undertaken using the ordinal ranking rubric described in Table 3.

This method of ranking clinical value is based on the transition from not working to working having the highest value of clinical significance, the transition from working to not working having the lowest value, and the remaining two transition groups demonstrating no change. The ranked value of these transition groups was used as a factor to examine comparisons among the groups in terms of scores on the OAHMQ, MTAP, and VAS. The data describing the scores for the three ranked groups are presented in Table 4.

A series of one-way analyses of variance comparing outcome transitions across the three baseline measures found that the VAS pain scale score did not significantly predict work outcome (F_2,132 = 2.612, p = 0.077), but that there were significant differences among the work outcome groups with regard to both the OAHMQ score (F_2,133 = 3.128, p = 0.047) and the MTAP score (F_2,132 = 5.007, p = 0.008). A post-hoc analysis based on Fisher’s least significant difference test is depicted in Table 5.

Participants who transitioned over the course of treatment from either negative to positive or from neutral to positive had significantly lower OAHMQ total scores, consistent with depressive symptoms being inversely related to successful return to work transition. Participants who transitioned from negative to positive over the course of treatment demonstrated significantly higher MTAP scores, but relationship between MTAP score and transition from neutral to either positive or negative was not found.

5.1Hypotheses confirmation

A method for nonbehavioral health practitioners to screen for comorbid depression among patients presenting with musculoskeletal pain is described and tested in terms of its concurrent validity and predictive validity. The hypothesized concurrent relationships between depression identified by the screening method and both self-reported pain and functional ability at intake are found, as are the predictive relationships between depression identified at the initiation of musculoskeletal treatment and eventual return to work transition.

5.2Screening for depression

Depression is beyond the scope of practice for NBH practitioners although the prevalence of depression is much higher for people with chronic musculoskeletal pain. Depression is associated with lower levels of treatment efficacy including delay in return to work and duration of disability [33], as well as both reduced work participation and reduced work functioning [34], all of which are important outcomes for treatment of musculoskeletal pain. The use of a screening method to identify musculoskeletal pain patients who are experiencing comorbid depression can be used in value-based payment programs to compensate for the poorer outcomes likely to be found when treating people with depression.

This study presents a two-step method described in Figure 2 that was developed for NBH practitioners to address depression in outpatient musculoskeletal pain rehabilitation. This method is based on the Older Adult Health and Mood Questionnaire (OAHMQ), a 22-item depression screening instrument with two subscales that tap complimentary components of depressive illness. The OAMHQ was developed for use by NBH professionals to screen for depressive symptoms in a physical medicine and rehabilitation setting and is available in both English and Spanish. The use of a score of 9/22 as the trigger for the nonbehavioral health practitioner to seek behavioral health consultation is borne out by the current research. Across the three community-based musculoskeletal pain rehabilitation clinics in which the two-step process was implemented, 22.3% of the patients had comorbid depression. This is similar to the 24.5% for moderate or worse depression among outpatients with low back pain identified by the Depression Anxiety Stress Scales-21 [35] in similar outpatient settings in Australia [22].

The current study found that none of the 35 participants who scored 9/22 or above on the OAHMQ were not in need of care for depressive illness. This indicates that using the score of 9/22 as the cut-point is too high; a lower score would provide better balance between sensitivity and specificity, to identify more participants with depression as well as a few “false positives”. In the original research on development of the OAHMQ [24], 23% of the participants had scores of three or less, while 37% had scores of 4 through 10, which the test developers proposed to indicate a depressive disorder that was not major, and 40% had scores above 10. Of this latter group, 76% of the participants were found to have a major depressive disorder. In the current study, 60% of the participants had scores of 4 or greater. In future studies with this instrument, a score of 7/22 or 8/22 should be used to screen for significant depressive symptoms in Step 1 of the process.

Using the 9/22 score in the process described in Fig. 2, there were no differences in terms of gender, native language, or work status at the start of the program or after treatment concluded. The expected group-wise differences were found for both the VAS pain scale score and the score for the Multidimensional Task Ability Profile (MTAP). The cross-sectional nature of this study does not support causality, but the biopsychosocial model presumes interaction among the variables of pain, depression, and self-perceived physical function [36]. Providing to nonbehavioral health practitioners the tools to screen for depression allows the sole practitioner to become a potential entry point for the patient into an interdisciplinary biopsychosocial program.

5.3Efficacy of the OAMHQ in screening

The OAMHQ subscales to measure dysphoric mood and cognitive-behavioral-physiologic symptoms of depression were useful during Step 2 of this process. The primary advantage of the OAMHQ is the availability of the two subscales, with dysphoria modestly predominant for participants in this study who were identified as experiencing depressive illness. The opportunity to focus nonbehavioral health interventions on the even-numbered items that reflect symptoms that are reasonably considered amenable to intervention within the scope of NBH practice seems to be important. For example, OAMHQ symptoms involving motivation (item 2 and item 20), insomnia (item 4), fatigue (item 6), appetite and digestion (item 14), and inactivity (item 18) are not only “psychiatric”, but can reasonably be addressed by physical therapists, occupational therapists, and chiropractic professionals.

Although dysphoric mood is often the predominant symptom, depression also impacts cognition, social behavior, and physiology. Screening measures for depressive illness that are sensitive to these components should be used whenever possible. As the biopsychosocial model continues to evolve, methods that recognize the interplay among the three major components of the model must be developed.

5.4Work outcomes

In addition to describing a method for physical therapists to screen for depression in physical therapy practice, this study examined the impact of depression on work outcomes through the use of a rubric that ranked work transition pre-post treatment. This rubric is the simplest approach to ordinal ranking both program intake and exit; it provides unequivocal data that are easy to collect. Using this approach, the VAS pain scale was found to not be predictive of outcome, but the OAHMQ total score, as well as the MTAP score were associated with successful work transitions. Pain patients with depression can become more activated through physical therapy, with an increased likelihood of successful return to work outcomes.

5.5Practical applications

Depression is a major cause of disability and lost work productivity and can interfere with physical therapy. While the biopsychosocial approach to musculoskeletal pain rehabilitation is widely endorsed, barriers to implementation exist for individual practitioners who do not have direct access to behavioral health colleagues. This study has demonstrated the feasibility of a routine depression screening process for patients with musculoskeletal pain disorders in outpatient practices. Screening can help nonbehavioral health practitioners participate in interdisciplinary biopsychosocial management by referral to behavioral health services. In addition, as value-based insurance reimbursement proliferates, the proper weighting of difficulty in treating musculoskeletal pain patients with comorbid depression is necessary.

5.6Limitations

This study used a convenience sample of patients at three “stand-alone” outpatient physical therapy clinics in Southern California that were not affiliated with a hospital, rehabilitation center, or surgical practice. Thus, natural selection has probably biased these results. However, the range of chronicity of the musculoskeletal conditions treated in these three practices is sufficiently broad that the findings are likely to be pertinent in most other community-based physical therapy settings.

As noted above, the cut-point of 9/22 is valid and useful, but likely to be too high in that there were no “false positive” patients identified without depression. This argues for a lower cut-point coupled with a robust screening program that identifies patients who have elevated scores but are not in need of treatment for depression. A false positive rate of 5% to 10% is reasonable to consider.

5.7Future research

Research should be undertaken with the OAMHQ (and other instruments so utilized) to determine an optimal score that balances sensitivity and specificity. While treating dysphoric mood is important, other aspects of depression such as physical activation and social participation should be assessed by nonbehavioral health practitioners.