- Email: [email protected]
Pain Limits the Effectiveness of Collaborative Care for Depression
Pain Limits the Effectiveness of Collaborative Care for Depression Stephen M. Thielke, M.D., M.A., Ming-Yu Fan, Ph.D., Mark Sullivan, M.D., Ph.D., Ju ¨ rgen Unu ¨ tzer, M.D., M.P.H.
Objective: To ascertain the effects of baseline pain on depression outcomes in a collaborative care treatment trial of depression for older adults. Methods: A secondary data analysis of 1,801 depressed older adults in the Improving Mood: Providing Access to Collaborative Treatment trial, comparing groups with no/low and high baseline pain using two pain interference variables. The primary outcome was a 50% reduction in depression score at 12 months. Analyses were performed separately for usual care and intervention groups, then examined for interactions. Results: In the treatment-as-usual group, there was no significant association of baseline pain status with depression outcomes. In the intervention group, higher pain interference was significantly associated with worse depression response: 48.9% of those with no/low pain interference achieved a depression response, compared with 37.4% of those with high pain (2 ⫽ 12.27, df⫽1, p⫽0.001). Arthritis pain interference showed a similar association (2 ⫽4.04, df⫽1, p⫽0.044). Controlling for sociodemographic and baseline characteristics did not diminish this association. A significant interaction effect on depression response was found between pain interference and the intervention, suggesting that higher pain differentially impairs depression response in collaborative care compared to usual care. Conclusion: A collaborative care intervention was significantly more successful in older adults with less pain. Pain may be an important barrier to improvement of depression and attending to pain might produce better depression outcomes. (Am J Geriatr Psychiatry 2007; 15:699–707)
C
hronic pain affects 60%– 80% of older adults, with the bulk of this related to age-related osteoarthritis.1,2 Pain interferes more with people’s everyday lives as they get older,3 and arthritis pain is associated with mental and physical impairments.4,5 Pain and depression frequently occur together6 – 8 and depression in the setting of pain is associated with higher functional limitations and economic bur-
dens than depression alone,9 and with more intensive use of general medical services.10 Pain with comorbid depressive symptoms is similarly associated with greater functional impairment than pain alone.11,12 Major depression is associated with higher health care costs, but the added costs may be accounted for by the subset of depressed patients with high pain interference.13
Received September 11, 2006; revised December 11, 2006; accepted December 12, 2006. From the Department of Psychiatry and Behavioral Sciences, University of Washington, Seattle, Washington. Send correspondence and reprint requests to Stephen M. Thielke, M.D., M.A., Department of Psychiatry and Behavioral Sciences, Box 356560, University of Washington, Seattle, WA 98115-6560. e-mail: [email protected]. © 2007 American Association for Geriatric Psychiatry
Am J Geriatr Psychiatry 15:8, August 2007
699
Effectiveness of Depression Collaborative Care The effects of pain on depression treatment response have been mixed in studies to date. Bair and colleagues found that adult patients in primary care with higher baseline pain complaints had worse depression responses from selective serotonin reuptake inhibitors at three months.14 Karp and colleagues examined the effect of bodily pain on time to depression response in 187 older adults in an open-label paroxetine trial, lasting up to 26 weeks.15 They found that although nonresponders reported more severe pain at baseline, after adjusting for severity of baseline depression, no effect was found for physical pain on time to response. In a younger sample treated with imipramine and interpersonal psychotherapy, Karp and colleagues found that higher levels of pain predicted a longer time to remission from depression and more suicidality, even when controlling for baseline depression.16 These were nonrandomized observational studies and the primary treatment was antidepressant medications. The Improving Mood: Providing Access to Collaborative Treatment (IMPACT) trial randomly assigned 1,801 depressed older adults with and without comorbid physical pain to a collaborative treatment intervention or to care as usual in the patient’s primary care clinic. The IMPACT intervention was more effective than usual care, with 45% of IMPACT participants showing a 50% reduction in depressive symptoms by 12 months, compared with only 19% of usual care patients.17 IMPACT care was also associated with less physical pain18,19 and better physical functioning20 than usual care, and these effects extended to 24 months.21 The intervention was effective regardless of age, sex, and baseline depression severity,16 ethnicity,22 baseline cognitive impairment,23 or comorbid medical illness.24 Although we were encouraged by the fact that collaborative care was more than twice as effective as usual care, we wanted to investigate potential causes for the high degree of persistent depression in the study. We followed the clinical suggestion of care managers who treated IMPACT intervention participants at the eight participating health care systems, who felt that physical pain was one of the major barriers to patients experiencing a full depression response. Subgroup analysis of patients in IMPACT with osteoarthritis has showed that greater
700
baseline pain was associated with less likely improvement in pain and disability at 12 months.19 Based on these observations, we hypothesized that higher baseline perceived pain interference would be associated with less likely depression response in the intervention at 12 months. Because of the other evidence implicating a qualitative difference between high and low pain groups,9,13 we also hypothesized that a similar association between greater pain and worse depression response would be seen in the usual care group.
METHODS The IMPACT trial involved a year-long collaborative care intervention to treat depression in primary care. It was conducted at 18 primary care clinics in 8 diverse health care organizations across five states. All 1,801 subjects were age 60 or older and met Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition criteria for major depression, dysthymia, or both. Half of the subjects were randomized to the IMPACT intervention. A nurse care manager collaborated with the patient and the patient’s primary care physician to provide multifaceted, stepped depression care. After gathering data and providing education and behavioral activation, the depression care manager helped patients determine and implement treatments. These included antidepressants, problem-solving therapy, and referrals to specialty mental health services. The depression care manager contacted the patients about every two weeks during acute-phase treatment and assessed depression outcomes. The care manager, a consulting psychiatrist, and an expert primary care physician met weekly to monitor progress and adjust treatment plans. Full details of the intervention can be found elsewhere.17 There was no specific intervention for pain or medical conditions related to pain. The control group received treatment as usual, after each of the control subjects’ primary care physicians were informed of the diagnosis of major depression or dysthymia. The intervention continued for 12 months, and independent assessments were administered to intervention and control participants at baseline, 3, 6, 12, 18, and 24 months. The patient data from all
Am J Geriatr Psychiatry 15:8, August 2007
Thielke et al. the clinics was combined in an unweighted manner. The data set was imputed for missing values and missing waves, with methods developed and tested using IMPACT data; full details of the imputation methods are discussed elsewhere.17,25 The method used, extended hot deck imputation, has been reexamined and found to compare favorably with other imputation techniques.26 In the usual care group, 134 (15.0%) had Hopkins Symptom Checklist (HSCL)-20 scores imputed at 12 months; 31 (3.5%) had died before that time. In the intervention, 98 (10.8%) of the intervention group had 12-month HSCL-20 scores imputed; 38 (4.2%) had died. Those who had died were excluded from the analysis. The HSCL-20 was used to measure depression, with the total score being the average of 20 items, each on a four-point scale.27 Pain interference was addressed in two measures.28 For the first measure, Pain Interference, 1,801 subjects were asked item 8 of the RAND 15-Item Short Form: “During the past month, how much did pain interfere with your normal work or other daily activities?” The scale on this item was: 1, not at all; 2, slightly; 3, moderately; 4, quite a bit; and 5, extremely. For the second measure, Arthritis Pain Interference, 1,002 (55%) subjects who described having osteoarthritis pain were asked, “In the past month, how much did your arthritis interfere with your daily activities rated,” where 0 was no interference and 10 was unable to carry out any activities. The primary outcome of the study was depression response, as defined by a 50% reduction in the HSCL-20 from baseline to 12 months. Those subjects who had an absolute HSCL-20 below 0.5 at both baseline and 12 months, and whose HSCL-20 score decreased (depression improved) during this time, were also considered to have had a response. There were three patients in this category in the intervention and one in the usual care group. The baseline pain measures were dichotomized as “no/low pain” and “high pain.” A Pain Interference (item 8 of the Short Form) score of 1 or 2 was defined as no/low, and 3, 4, or 5 as high. An Arthritis Pain Interference score of 0 – 4 was defined as no/low and 5–10 was defined as high. These cutoffs were selected because the keyed responses corresponded roughly to low and high pain states, and because this captured at least one quartile of the group.
Am J Geriatr Psychiatry 15:8, August 2007
Data Analysis We analyzed the data using SPSS (SPSS Inc., version 13.0). Because the primary goal was to understand the effect of baseline pain symptoms on depression response, and not to calculate intervention effect, the control and intervention groups were first analyzed separately and then examined jointly for interactions. First, we examined the baseline characteristics in the no/low and high pain groups using 2 tests and Student’s t-tests for continuous and categorical variables, respectively. Second, we used 2 tests to compare the depression response for the no/low and high pain groups in the two pain measures. The comparison was carried out separately for intervention and usual care groups. Third, we created logistic regression models to examine the association between baseline pain symptoms and depression response after adjusting for age, sex, race, marital status, educational level, number of medical conditions, and baseline depression. Again, we ran separate models for intervention and usual care groups. Fourth, in order to examine the interaction effect between baseline pain and treatment group for depression response, we combined the data from both treatment groups and created a logistic regression model including baseline pain, indicator for treatment group, the interaction term between baseline pain and treatment group, and all the covariates used in the third step.
RESULTS The baseline characteristics of the control and intervention groups, categorized by baseline pain group, are presented in Table 1. There were no significant differences between control and intervention groups; they are grouped together for this table, although for the other analyses they are treated separately. Lower baseline pain on either of the pain measures was significantly associated with higher education, lower depression scores on the HSCL-20, and fewer medical comorbidities. Lower Pain Interference scores were significantly associated with being married or living with a partner but not with race, whereas lower Arthritis Pain Interference scores were associated with race but not partnered status. There was no
701
Effectiveness of Depression Collaborative Care
TABLE 1.
Baseline Characteristics of Sample Population Based on Pain Measures Pain Interference
Number (%) Interventiona Age, years (SD) Male (%) White (%) Married/living with partner (%) No. of chronic medical conditions (SD) Education: some college or more (%) Baseline HSCL-20 score (SD)
Arthritis Pain Interference
No/Low Pain
High Pain
p Value
No/Low Pain
High Pain
p Value
764 (42.4%) 50.8% 71.4 (7.59) 37.4% 78.7% 50.5% 2.98 (1.74) 64.9% 1.56 (0.59)
1037 (57.6%) 50.0% 71.0 (7.38) 33.5% 76.0% 43.3% 4.38 (1.86) 53.0% 1.77 (0.60)
0.727 0.327 0.081 0.179 0.002 ⬍0.001 ⬍0.001 ⬍0.001
399 (39.8%) 52.4% 72.2 (7.38) 35.1% 83.2% 46.1% 4.34 (1.65) 60.4% 1.62 (0.55)
603 (60.2%) 49.3% 71.9 (7.43) 29.5% 71.5% 43.0% 4.95 (1.66) 51.7% 1.79 (0.61)
0.333 0.514 0.064 ⬍0.001 0.324 ⬍0.001 0.007 ⬍0.001
Notes: For Pain Interference, no/low pain is 1–2 and high pain is 3–5 on the pain item of the RAND 15-Item Short Form. For Arthritis Pain Interference, no/low pain is defined as 0 – 4 and high pain as 5–10 on a 10-point scale. Comparisons are based on 2 or Student’s t-tests with equal variances not assumed. a No differences were observed between intervention and usual care groups for the other categories. SD: standard deviation.
significant difference between the pain groups in sex or age. Table 2 presents the unadjusted association between depression response and baseline pain measures, separated by intervention and control. Among controls, there was no difference observed in depression response based on baseline pain or arthritis category. In the intervention, lower levels of both Pain Interference and Arthritis Pain Interference were significantly associated with better depression response, (2(1)⫽12.23, p⫽0.001 for the difference in Pain Interference, and 2(1)⫽4.04, p⫽0.044 for the difference in Arthritis Pain Interference). The absolute difference between groups was roughly 10% and the relative difference was approximately 30%. Figure 1 represents the proportion of depression responders at 12 months for each incremental category of baseline Pain Interference. The presence or absence of arthritis, independent
TABLE 2.
of pain, was not associated with a difference in depression response. In the usual care group, 17.5% of those with arthritis (N⫽481) and 19.6% of those without arthritis (N ⫽ 383) had a depression response at 12 months (2(1) ⫽0.64, p⫽0.425). In the intervention, 40.3% of those with arthritis (N⫽ 484) and 45.1% of those without arthritis (N⫽384) had a depression response (2(1) ⫽1.99, p⫽ 0.158). The multiple logistic regression models of pain on depression improvement with the adjustment of baseline covariates are reported in Table 3. In the IMPACT intervention, higher Pain Interference and higher Arthritis Pain Interference were both significantly associated with lower odds of depression improvement (odds ratio [OR]⫽ 0.641, 2(1) ⫽8.22, p⫽ 0.003 for Pain Interference; OR⫽0.658, 2(1) ⫽4.54, p⫽0.033 for Arthritis Pain Interference). Age was a significant covariate in both the whole group (OR⫽ 0.970, 2(1) ⫽9.01, p⫽0.003) and the arthritis group
Percentage of Subjects With Depression Response in the Control and Intervention Groups by Baseline Pain
Category Usual care Pain interference Depression response (%) Arthritis pain interference Depression response (%) Intervention Pain interference Depression response (%) Arthritis pain interference Depression response (%)
No/Low Pain
High Pain
362 (41.8%) 18.8% 183 (38.4%) 18.6%
502 (58.2%) 18.1% 298 (61.6%) 16.8%
371 (43.1%) 48.9% 201 (41.7%) 45.8%
489 (56.9%) 37.4% 281 (58.3%) 36.6%
2
p Value
0.060
0.806
0.255
0.614
12.226
0.001
4.043
0.044
Notes: Comparisons based on 2 tests.
702
Am J Geriatr Psychiatry 15:8, August 2007
Thielke et al.
FIGURE 1.
Proportion of Respondents With Depression Response (50% or greater reduction in depression symptoms from baseline) Based on Treatment Group and Baseline Pain Interference Measure
depression response, we conducted logistic regression models including an interaction term between treatment group and pain category for both pain measures. For Pain Interference, the interaction term was significant at a 95% confidence level (OR⫽ 0.631, 2(1) ⫽4.01, p⫽0.045, 95% confidence interval: 0.402– 0.99). This suggests that the assumption of homogeneity between treatment groups by pain status should not be accepted, and the association between pain and depression outcome should be assessed separately for intervention and usual care groups. For Arthritis Pain Interference, the interaction term was not significant (OR⫽0.776, 2(1)⫽ 0.653, p⫽0.419).
DISCUSSION
(OR⫽0.959, 2(1)⫽9.15, p⫽0.002). Higher baseline depression was significantly associated with better depression response in the Pain Interference model (OR⫽1.433, 2(1)⫽8.87, p ⫽0.003), but not in the Arthritis Pain Interference model (OR⫽1.048, 2(1) ⫽0.08, p⫽ 0.774). In the usual care group, after adjusting for the covariates, neither pain measure was significantly associated with depression response at 12 months (OR⫽1.037, 2(1)⫽0.02, p⫽0.860 for Pain Interference; OR⫽0.896, 2(1)⫽0.175, p⫽0.675 for Arthritis Pain Interference). In the Pain Interference model, only fewer baseline medical conditions (OR⫽0.876, 2(1) ⫽6.104, p⫽ 0.013) and higher baseline HSCL-20 score (OR⫽1.930, 2(1)⫽18.37, p ⬍0.001) were associated with higher odds of depression response. In the Arthritis Pain Interference model, the same two covariates were significant (OR⫽0.829, 2(1) ⫽ 5.378, p⫽0.020 for number of medical conditions; OR⫽2.202, 2(1)⫽12.746, p ⬍0.001 for baseline HSCL-20). The results in Table 3 suggest that treatment group might have a modifying effect on the association between baseline pain interference and 12month depression improvement. To examine the interaction of baseline pain with treatment group for
Am J Geriatr Psychiatry 15:8, August 2007
As had been suggested by the IMPACT care managers, intervention participants with high baseline pain interference measures had lower rates of depression response than those with no/low pain. In a subset of 1,002 subjects with arthritis, higher Arthritis Pain Interference had roughly the same effects on depression response in the intervention, but was less significant. The presence of arthritis was not in itself associated with differential depression outcomes. Adjusting for age, sex, race, and marital status did not diminish the association of baseline pain with depression response in the intervention group. Higher baseline depression was associated with better depression responses, but not in the subgroup of those with arthritis. Older age was significantly associated with a decreased depression response in all the models, although the effects of age were modest. Age was not significantly associated with pain status at baseline, and the difference of the mean age between the high and no/low pain groups was less than six months. The association of higher pain interference with worse depression response was not observed in the usual care group, where depression response rates were similarly low in both pain groups (18.8% for no/low versus 18.1% for high Pain Interference). The results of the regression analysis with the interaction term suggested that Pain Interference showed a significant differential effect in the collaborative care
703
Effectiveness of Depression Collaborative Care
TABLE 3.
Logistic Regression Models for the Intervention and Control Intervention
Pain Interference High vs low pain interference Older agea Higher baseline HSCL-20b No. of medical comorbidities Arthritis Pain Interference High vs low pain interference Older agea Higher baseline HSCL-20b No. of medical comorbidities Education: high school or less
Usual Care
Odds Ratio
Wald 2
df
p Value
Odds Ratio
Wald 2
df
p Value
0.641 0.970 1.433 0.932
8.22 9.01 8.87 3.07
1 1 1 1
0.004 0.003 0.003 0.080
1.037 0.980 1.930 0.876
0.02 2.57 18.37 6.10
1 1 1 1
0.860 0.109 ⬍0.001 0.013
0.658 0.959 1.048 0.932 0.665
4.54 9.15 0.08 1.40 4.34
1 1 1 1 1
0.033 0.002 0.774 0.236 0.037
0.896 0.969 2.202 0.829 0.986
0.175 3.22 12.74 5.38 0.01
1 1 1 1 1
0.675 0.073 ⬍0.001 0.020 0.956
Notes: Wald 2 test was used to compare depression response (50% reduction at 12 months), with pain category and baseline measures as covariates. Separate models are presented for the whole group and the subset of 1,002 arthritis subjects. Of the covariates age, sex, race, marital status, medical comorbidities, and baseline HSCL-20 score, only those found to be significant in either the treatment or the usual care group are listed. a Per increase in year from the minimum subject age of 65 years. b Per one point increase in HSCL-20 (on a 0 – 4 point scale). HSCL: Hopkins Symptom Checklist.
intervention compared with usual care. Arthritis Pain Interference did not show this interaction. The initial hypothesis that higher pain would be associated with less likely improvement in depression was not confirmed for the whole usual care group or the arthritis subgroup. Thus from the usual care data, there was no strong evidence that higher pain was generally associated with worse depression outcomes at 12 months. Although several of the other clinical variables assessed at baseline, including severity of depression, functional status, and number of chronic medical conditions, did not appear to diminish depression response in the intervention, baseline pain had a significant negative effect on depression response, which suggests a unique relationship between pain and depression in collaborative care treatment. The specific consequences of pain on depression outcomes could be from a number of factors, including more severe neurovegetative symptoms not captured fully by the HSCL-20, decreased ability to fully engage in pleasant events or behavioral activation such as physical or social activities, distraction from the depression intervention by pain, or a more fundamental neural correlation between being in pain and feeling depressed (Bair7 provides a discussion of the “depression-pain syndrome”). Regardless of the exact cause, our findings support the clinical observation of the study care managers who felt that phys-
704
ical pain was often a “sticking point” to depression response in this intervention. These findings differ from results published by Karp and colleagues,14 who concluded that after correcting for baseline depression, pain did not lower depression response rates or increase time to response in older adults. That trial involved only medications and was shorter in duration than the IMPACT intervention. In the usual-care arm of IMPACT, 57% of the subjects were receiving an antidepressant from their primary care physician at 12 months, and in this group baseline pain did not seem to be associated with a difference in depression response, which agrees with Karp’s findings. Yet in the intervention arm, 73% of subjects were receiving an antidepressant at 12 months and higher baseline pain was associated with significantly worse depression response. The IMPACT intervention involved more than medications, with all the intervention subjects receiving routine contact from care managers and 30% receiving problem-solving therapy. Pain might thus have different consequences for depression response in longer-term collaborative, multifaceted interventions such as IMPACT than in short-term antidepressant-only trials. Higher pain interference may inhibit the patient activation that is thought to be a crucial component of collaborative care, may distract patients from the content of the intervention, or may be associated with character-
Am J Geriatr Psychiatry 15:8, August 2007
Thielke et al. ologic traits that make depression response less likely.29 The recent finding that high perceived pain interference might be more closely associated with high medical costs than depression13 implies that depression with pain may have important qualitative differences from depression without pain. Our analysis corroborates this observation, and suggests that depression without pain may have fewer functional consequences and be more amenable to treatment than depression with pain. There may be qualitative differences in the “type” of depression in no/low and high-pain groups that is not captured in traditional measure of depression such as the HSCL-20, which focus more on psychological distress than on physical symptoms. Understanding the complex relationship between pain, depression, medical care, and treatment response merits further research. Our analysis raises the question of whether pain treatment might improve depression outcomes. There is to our knowledge no published evidence that independently treating pain improves outcomes for major depression in any age group. Some trials have shown improvement in mood with opioid treatment of chronic back pain30 and such findings suggest that if depressed older adults were to experience less pain, they might be more likely to respond to interventions for depression. Despite 58% of subjects at baseline describing moderate or greater pain interference, a previous analysis of IMPACT found that almost half of those with functional impairment from pain did not report using any analgesic medications.31 Pain was thus probably globally undertreated in this depressed cohort, and better treatment of it might increase the likelihood of depression response. Alternately, high perceived pain might be a marker of personality or attachment traits that would persist even if the pain were treated or nominally improved,32 and might continue to decrease the likelihood of improving with a depression intervention. Finally, we note that clinical efficacy studies of depression treatments (e.g., clinical drug trials) tend to produce better outcomes, with greater treatment effects, than effectiveness studies of depression interventions provided by usual providers in “real world” primary care settings. We believe that this
Am J Geriatr Psychiatry 15:8, August 2007
may be, in part, because efficacy studies often exclude patients with chronic medical illnesses that involve pain, medically ill patients with significant pain complaints, or patients who are taking regular medications.33 Our analyses suggest that some of this difference might derive from selection bias of subjects with less pain. From the current data, excluding those subjects who had moderate to high pain interference would result in roughly a 30% greater relative response in the intervention compared to the whole group, with little change in response in the control group. Such an exclusion would create the appearance of a greater intervention effect, but it would also ignore about 60% of the entire sample of depressed older primary care patients. Limitations This study has several limitations. Pain was assessed by self-report, single-item measures. There were no subjective measures of Activity of Daily Living impairment, or observation-based assessments of functional capacity. From the current data, there was no way to ascertain or correct for the use of analgesic medications. Treatment response, a 50% reduction in depression (HSCL-20) at 12 months, was used as the primary outcome measure instead of remission (e.g., a value below a defined threshold, such as 0.5 on the HSCL-20), but because the HSCL-20 and pain measures were cross-sectionally correlated, using an absolute score threshold would have increased the possibility of confounding.
CONCLUSIONS Unlike a prior study showing that pain does not interfere with response to depression treatment in older adults, our study indicated that higher baseline pain, from arthritis or other causes, was associated with worse depression response during a 12-month collaborative care intervention. This association was not observed in controls who tended to have low response rates regardless of baseline pain status. The association of more pain with worse depression response in intervention subjects persisted even when adjusting for sociodemographic variables, medical comorbidities, and baseline depression. A significant interaction term indicated that the pain had a differ-
705
Effectiveness of Depression Collaborative Care ential effect in the intervention compared with usual care. Together, these findings suggest that pain may be a unique barrier to depression response during collaborative care interventions for depression. In conjunction with recent findings about the central importance of pain to medical costs,13 this work highlights the broad importance of pain to aging and mental health. To our knowledge, no research has determined if reducing pain would also bolster depression treatment, although this may be an action-
able means of improving outcomes in a substantial subset of older adults with depression.
This work was supported by the National Research Service Award T-32 Fellowship (SMT), the John A Hartford Foundation (JU), the California HealthCare Foundation (JU), and the Paul Beeson Physician Faculty Scholars Award from the American Federation for Aging Research (JU).
REFERENCES 1. Centers for Disease Control: Projected Prevalence of Self-Reported Arthritis or Chronic Joint Symptoms Among Persons 65 years—United States, 2005-2030. MMRW; 2003 2. Yelin E, Callahan LF: The economic cost and social and psychological impact of musculoskeletal conditions. National Arthritis Data Work Groups. Arthritis Rheum 1995; 38:1351–1362 3. Thomas E, Peat G, Harris L, et al: The prevalence of pain and pain interference in a general population of older adults: cross-sectional findings from the North Staffordshire Osteoarthritis Project (NorStOP). Pain 2004; 110:361–368 4. Stang PE, Brandenburg NA, Lane MC, et al: Mental and physical comorbid conditions and days in role among persons with arthritis. Psychosom Med 2006; 68:152–158 5. Verbrugge LM, Juarez L: Profile of arthritis disability: II. Arthritis Rheum 2006; 55:102–113 6. Arnow BA, Hunkeler EM, Blasey CM, et al: Comorbid depression, chronic pain, and disability in primary care. Psychosom Med 2006; 68:262–268 7. Bair MJ, Robinson RL, Katon W, Kroenke K: Depression and pain comorbidity: a literature review. Arch Intern Med 2003; 163: 2433–2445 8. Ohayon MM, Schatzberg AF: Using chronic pain to predict depressive morbidity in the general population. Arch Gen Psychiatry 2003; 60:39–47 9. Emptage NP, Sturm R, Robinson RL: Depression and comorbid pain as predictors of disability, employment, insurance status, and health care costs. Psychiatr Serv 2005; 56:468–474 10. Bao Y, Sturm R, Croghan TW: A national study of the effect of chronic pain on the use of health care by depressed persons. Psychiatr Serv 2003; 54:693–697 11. Mossey JM, Gallagher RM: The longitudinal occurrence and impact of comorbid chronic pain and chronic depression over two years in continuing care retirement community residents. Pain Med 2004; 5:335–348 12. Mossey JM, Gallagher RM, Tirumalasetti F: The effects of pain and depression on physical functioning in elderly residents of a continuing care retirement community. Pain Med 2000; 1:340–350 13. Gameroff MJ, Olfson M: Major depressive disorder, somatic pain, and health care costs in an urban primary care practice. J Clin Psychiatry 2006; 67:1232–1239 14. Bair MJ, Robinson RL, Eckert GJ, et al: Impact of pain on depression treatment response in primary care. Psychosom Med 2004; 66:17–22 15. Karp JF, Weiner D, Seligman K, et al: Body pain and treatment
706
response in late-life depression. Am J Geriatr Psychiatry 2005; 13:188–194 16. Karp JF, Scott J, Houck P, et al: Pain predicts longer time to remission during treatment of recurrent depression. J Clin Psychiatry 2005; 66:591–597 17. Unu ¨ tzer J, Katon W, Callahan CM, et al: IMPACT Investigators. Improving Mood-Promoting Access to Collaborative Treatment. Collaborative care management of late-life depression in the primary care setting: a randomized controlled trial. JAMA 2002; 288:2836–2845 18. Lin EH, Katon W, Von Korff M, et al: IMPACT Investigators. Effect of improving depression care on pain and functional outcomes among older adults with arthritis: a randomized controlled trial. JAMA 2003; 290:2428–2429 19. Lin EH, Tang L, Katon W, et al: Arthritis pain and disability: response to collaborative depression care. Gen Hosp Psychiatry 2006; 28:482–486 20. Callahan CM, Kroenke K, Counsell SR, et al: IMPACT Investigators. Treatment of depression improves physical functioning in older adults. J Am Geriatr Soc 2005; 53:367–373 21. Hunkeler EM, Katon W, Tang L, et al: Long term outcomes from the IMPACT randomised trial for depressed elderly patients in primary care. BMJ 2006; 332:259–263 22. Arean PA, Ayalon L, Hunkeler E, et al: IMPACT Investigators. Improving depression care for older, minority patients in primary care. Med Care 2005; 43:381–390 23. Steffens DC, Snowden M, Fan MY, et al: IMPACT Investigators. Cognitive impairment and depression outcomes in the IMPACT study. Am J Geriatr Psychiatry 2006; 14:401–409 24. Harpole LH, Williams JW Jr, Olsen MK, et al: Improving depression outcomes in older adults with comorbid medical illness. Gen Hosp Psychiatry 2005; 27:4–12 25. Katon WJ, Schoenbaum M, Fan MY, et al: Cost-effectiveness of improving primary care treatment of late-life depression. Arch Gen Psychiatry 2005; 62:1313–1320 26. Tang L, Song J, Belin TR, Unutzer J: A comparison of imputation methods in a longitudinal randomized clinical trial. Stat Med 2005; 24:2111–2128 27. Derogatis LR, Lipman RS, Rickels K, et al: The Hopkins Symptom Checklist (HSCL): a self-report symptom inventory. Behav Sci 1974; 19:1–15 28. Von Korff M, Jensen MP, Karoly P: Assessing global pain severity by self-report in clinical and health services research. Spine 2000; 25:3140–3151 29. Asghari A, Nicholas MK: Personality and pain-related beliefs/
Am J Geriatr Psychiatry 15:8, August 2007
Thielke et al. coping strategies: a prospective study. Clin J Pain 2006; 22: 10–18 30. Jamison RN, Raymond SA, Slawsby EA, et al: Opioid therapy for chronic noncancer back pain. A randomized prospective study. Spine 1998; 23:2591–2600 31. Unu ¨ tzer J, Ferrell B, Lin EH, Marmon T: Pharmacotherapy of pain in depressed older adults. J Am Geriatr Soc 2004; 52:1916–1922
Am J Geriatr Psychiatry 15:8, August 2007
32. Ciechanowski P, Sullivan M, Jensen M, et al: The relationship of attachment style to depression, catastrophizing and health care utilization in patients with chronic pain. Pain 2003; 104:627–637 33. Keitner GI, Posternak MA, Ryan CE: How many subjects with major depressive disorder meet eligibility requirements of an antidepressant efficacy trial? J Clin Psychiatry 2003; 64:1091– 1093
707
Objective: To ascertain the effects of baseline pain on depression outcomes in a collaborative care treatment trial of depression for older adults. Methods: A secondary data analysis of 1,801 depressed older adults in the Improving Mood: Providing Access to Collaborative Treatment trial, comparing groups with no/low and high baseline pain using two pain interference variables. The primary outcome was a 50% reduction in depression score at 12 months. Analyses were performed separately for usual care and intervention groups, then examined for interactions. Results: In the treatment-as-usual group, there was no significant association of baseline pain status with depression outcomes. In the intervention group, higher pain interference was significantly associated with worse depression response: 48.9% of those with no/low pain interference achieved a depression response, compared with 37.4% of those with high pain (2 ⫽ 12.27, df⫽1, p⫽0.001). Arthritis pain interference showed a similar association (2 ⫽4.04, df⫽1, p⫽0.044). Controlling for sociodemographic and baseline characteristics did not diminish this association. A significant interaction effect on depression response was found between pain interference and the intervention, suggesting that higher pain differentially impairs depression response in collaborative care compared to usual care. Conclusion: A collaborative care intervention was significantly more successful in older adults with less pain. Pain may be an important barrier to improvement of depression and attending to pain might produce better depression outcomes. (Am J Geriatr Psychiatry 2007; 15:699–707)
C
hronic pain affects 60%– 80% of older adults, with the bulk of this related to age-related osteoarthritis.1,2 Pain interferes more with people’s everyday lives as they get older,3 and arthritis pain is associated with mental and physical impairments.4,5 Pain and depression frequently occur together6 – 8 and depression in the setting of pain is associated with higher functional limitations and economic bur-
dens than depression alone,9 and with more intensive use of general medical services.10 Pain with comorbid depressive symptoms is similarly associated with greater functional impairment than pain alone.11,12 Major depression is associated with higher health care costs, but the added costs may be accounted for by the subset of depressed patients with high pain interference.13
Received September 11, 2006; revised December 11, 2006; accepted December 12, 2006. From the Department of Psychiatry and Behavioral Sciences, University of Washington, Seattle, Washington. Send correspondence and reprint requests to Stephen M. Thielke, M.D., M.A., Department of Psychiatry and Behavioral Sciences, Box 356560, University of Washington, Seattle, WA 98115-6560. e-mail: [email protected]. © 2007 American Association for Geriatric Psychiatry
Am J Geriatr Psychiatry 15:8, August 2007
699
Effectiveness of Depression Collaborative Care The effects of pain on depression treatment response have been mixed in studies to date. Bair and colleagues found that adult patients in primary care with higher baseline pain complaints had worse depression responses from selective serotonin reuptake inhibitors at three months.14 Karp and colleagues examined the effect of bodily pain on time to depression response in 187 older adults in an open-label paroxetine trial, lasting up to 26 weeks.15 They found that although nonresponders reported more severe pain at baseline, after adjusting for severity of baseline depression, no effect was found for physical pain on time to response. In a younger sample treated with imipramine and interpersonal psychotherapy, Karp and colleagues found that higher levels of pain predicted a longer time to remission from depression and more suicidality, even when controlling for baseline depression.16 These were nonrandomized observational studies and the primary treatment was antidepressant medications. The Improving Mood: Providing Access to Collaborative Treatment (IMPACT) trial randomly assigned 1,801 depressed older adults with and without comorbid physical pain to a collaborative treatment intervention or to care as usual in the patient’s primary care clinic. The IMPACT intervention was more effective than usual care, with 45% of IMPACT participants showing a 50% reduction in depressive symptoms by 12 months, compared with only 19% of usual care patients.17 IMPACT care was also associated with less physical pain18,19 and better physical functioning20 than usual care, and these effects extended to 24 months.21 The intervention was effective regardless of age, sex, and baseline depression severity,16 ethnicity,22 baseline cognitive impairment,23 or comorbid medical illness.24 Although we were encouraged by the fact that collaborative care was more than twice as effective as usual care, we wanted to investigate potential causes for the high degree of persistent depression in the study. We followed the clinical suggestion of care managers who treated IMPACT intervention participants at the eight participating health care systems, who felt that physical pain was one of the major barriers to patients experiencing a full depression response. Subgroup analysis of patients in IMPACT with osteoarthritis has showed that greater
700
baseline pain was associated with less likely improvement in pain and disability at 12 months.19 Based on these observations, we hypothesized that higher baseline perceived pain interference would be associated with less likely depression response in the intervention at 12 months. Because of the other evidence implicating a qualitative difference between high and low pain groups,9,13 we also hypothesized that a similar association between greater pain and worse depression response would be seen in the usual care group.
METHODS The IMPACT trial involved a year-long collaborative care intervention to treat depression in primary care. It was conducted at 18 primary care clinics in 8 diverse health care organizations across five states. All 1,801 subjects were age 60 or older and met Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition criteria for major depression, dysthymia, or both. Half of the subjects were randomized to the IMPACT intervention. A nurse care manager collaborated with the patient and the patient’s primary care physician to provide multifaceted, stepped depression care. After gathering data and providing education and behavioral activation, the depression care manager helped patients determine and implement treatments. These included antidepressants, problem-solving therapy, and referrals to specialty mental health services. The depression care manager contacted the patients about every two weeks during acute-phase treatment and assessed depression outcomes. The care manager, a consulting psychiatrist, and an expert primary care physician met weekly to monitor progress and adjust treatment plans. Full details of the intervention can be found elsewhere.17 There was no specific intervention for pain or medical conditions related to pain. The control group received treatment as usual, after each of the control subjects’ primary care physicians were informed of the diagnosis of major depression or dysthymia. The intervention continued for 12 months, and independent assessments were administered to intervention and control participants at baseline, 3, 6, 12, 18, and 24 months. The patient data from all
Am J Geriatr Psychiatry 15:8, August 2007
Thielke et al. the clinics was combined in an unweighted manner. The data set was imputed for missing values and missing waves, with methods developed and tested using IMPACT data; full details of the imputation methods are discussed elsewhere.17,25 The method used, extended hot deck imputation, has been reexamined and found to compare favorably with other imputation techniques.26 In the usual care group, 134 (15.0%) had Hopkins Symptom Checklist (HSCL)-20 scores imputed at 12 months; 31 (3.5%) had died before that time. In the intervention, 98 (10.8%) of the intervention group had 12-month HSCL-20 scores imputed; 38 (4.2%) had died. Those who had died were excluded from the analysis. The HSCL-20 was used to measure depression, with the total score being the average of 20 items, each on a four-point scale.27 Pain interference was addressed in two measures.28 For the first measure, Pain Interference, 1,801 subjects were asked item 8 of the RAND 15-Item Short Form: “During the past month, how much did pain interfere with your normal work or other daily activities?” The scale on this item was: 1, not at all; 2, slightly; 3, moderately; 4, quite a bit; and 5, extremely. For the second measure, Arthritis Pain Interference, 1,002 (55%) subjects who described having osteoarthritis pain were asked, “In the past month, how much did your arthritis interfere with your daily activities rated,” where 0 was no interference and 10 was unable to carry out any activities. The primary outcome of the study was depression response, as defined by a 50% reduction in the HSCL-20 from baseline to 12 months. Those subjects who had an absolute HSCL-20 below 0.5 at both baseline and 12 months, and whose HSCL-20 score decreased (depression improved) during this time, were also considered to have had a response. There were three patients in this category in the intervention and one in the usual care group. The baseline pain measures were dichotomized as “no/low pain” and “high pain.” A Pain Interference (item 8 of the Short Form) score of 1 or 2 was defined as no/low, and 3, 4, or 5 as high. An Arthritis Pain Interference score of 0 – 4 was defined as no/low and 5–10 was defined as high. These cutoffs were selected because the keyed responses corresponded roughly to low and high pain states, and because this captured at least one quartile of the group.
Am J Geriatr Psychiatry 15:8, August 2007
Data Analysis We analyzed the data using SPSS (SPSS Inc., version 13.0). Because the primary goal was to understand the effect of baseline pain symptoms on depression response, and not to calculate intervention effect, the control and intervention groups were first analyzed separately and then examined jointly for interactions. First, we examined the baseline characteristics in the no/low and high pain groups using 2 tests and Student’s t-tests for continuous and categorical variables, respectively. Second, we used 2 tests to compare the depression response for the no/low and high pain groups in the two pain measures. The comparison was carried out separately for intervention and usual care groups. Third, we created logistic regression models to examine the association between baseline pain symptoms and depression response after adjusting for age, sex, race, marital status, educational level, number of medical conditions, and baseline depression. Again, we ran separate models for intervention and usual care groups. Fourth, in order to examine the interaction effect between baseline pain and treatment group for depression response, we combined the data from both treatment groups and created a logistic regression model including baseline pain, indicator for treatment group, the interaction term between baseline pain and treatment group, and all the covariates used in the third step.
RESULTS The baseline characteristics of the control and intervention groups, categorized by baseline pain group, are presented in Table 1. There were no significant differences between control and intervention groups; they are grouped together for this table, although for the other analyses they are treated separately. Lower baseline pain on either of the pain measures was significantly associated with higher education, lower depression scores on the HSCL-20, and fewer medical comorbidities. Lower Pain Interference scores were significantly associated with being married or living with a partner but not with race, whereas lower Arthritis Pain Interference scores were associated with race but not partnered status. There was no
701
Effectiveness of Depression Collaborative Care
TABLE 1.
Baseline Characteristics of Sample Population Based on Pain Measures Pain Interference
Number (%) Interventiona Age, years (SD) Male (%) White (%) Married/living with partner (%) No. of chronic medical conditions (SD) Education: some college or more (%) Baseline HSCL-20 score (SD)
Arthritis Pain Interference
No/Low Pain
High Pain
p Value
No/Low Pain
High Pain
p Value
764 (42.4%) 50.8% 71.4 (7.59) 37.4% 78.7% 50.5% 2.98 (1.74) 64.9% 1.56 (0.59)
1037 (57.6%) 50.0% 71.0 (7.38) 33.5% 76.0% 43.3% 4.38 (1.86) 53.0% 1.77 (0.60)
0.727 0.327 0.081 0.179 0.002 ⬍0.001 ⬍0.001 ⬍0.001
399 (39.8%) 52.4% 72.2 (7.38) 35.1% 83.2% 46.1% 4.34 (1.65) 60.4% 1.62 (0.55)
603 (60.2%) 49.3% 71.9 (7.43) 29.5% 71.5% 43.0% 4.95 (1.66) 51.7% 1.79 (0.61)
0.333 0.514 0.064 ⬍0.001 0.324 ⬍0.001 0.007 ⬍0.001
Notes: For Pain Interference, no/low pain is 1–2 and high pain is 3–5 on the pain item of the RAND 15-Item Short Form. For Arthritis Pain Interference, no/low pain is defined as 0 – 4 and high pain as 5–10 on a 10-point scale. Comparisons are based on 2 or Student’s t-tests with equal variances not assumed. a No differences were observed between intervention and usual care groups for the other categories. SD: standard deviation.
significant difference between the pain groups in sex or age. Table 2 presents the unadjusted association between depression response and baseline pain measures, separated by intervention and control. Among controls, there was no difference observed in depression response based on baseline pain or arthritis category. In the intervention, lower levels of both Pain Interference and Arthritis Pain Interference were significantly associated with better depression response, (2(1)⫽12.23, p⫽0.001 for the difference in Pain Interference, and 2(1)⫽4.04, p⫽0.044 for the difference in Arthritis Pain Interference). The absolute difference between groups was roughly 10% and the relative difference was approximately 30%. Figure 1 represents the proportion of depression responders at 12 months for each incremental category of baseline Pain Interference. The presence or absence of arthritis, independent
TABLE 2.
of pain, was not associated with a difference in depression response. In the usual care group, 17.5% of those with arthritis (N⫽481) and 19.6% of those without arthritis (N ⫽ 383) had a depression response at 12 months (2(1) ⫽0.64, p⫽0.425). In the intervention, 40.3% of those with arthritis (N⫽ 484) and 45.1% of those without arthritis (N⫽384) had a depression response (2(1) ⫽1.99, p⫽ 0.158). The multiple logistic regression models of pain on depression improvement with the adjustment of baseline covariates are reported in Table 3. In the IMPACT intervention, higher Pain Interference and higher Arthritis Pain Interference were both significantly associated with lower odds of depression improvement (odds ratio [OR]⫽ 0.641, 2(1) ⫽8.22, p⫽ 0.003 for Pain Interference; OR⫽0.658, 2(1) ⫽4.54, p⫽0.033 for Arthritis Pain Interference). Age was a significant covariate in both the whole group (OR⫽ 0.970, 2(1) ⫽9.01, p⫽0.003) and the arthritis group
Percentage of Subjects With Depression Response in the Control and Intervention Groups by Baseline Pain
Category Usual care Pain interference Depression response (%) Arthritis pain interference Depression response (%) Intervention Pain interference Depression response (%) Arthritis pain interference Depression response (%)
No/Low Pain
High Pain
362 (41.8%) 18.8% 183 (38.4%) 18.6%
502 (58.2%) 18.1% 298 (61.6%) 16.8%
371 (43.1%) 48.9% 201 (41.7%) 45.8%
489 (56.9%) 37.4% 281 (58.3%) 36.6%
2
p Value
0.060
0.806
0.255
0.614
12.226
0.001
4.043
0.044
Notes: Comparisons based on 2 tests.
702
Am J Geriatr Psychiatry 15:8, August 2007
Thielke et al.
FIGURE 1.
Proportion of Respondents With Depression Response (50% or greater reduction in depression symptoms from baseline) Based on Treatment Group and Baseline Pain Interference Measure
depression response, we conducted logistic regression models including an interaction term between treatment group and pain category for both pain measures. For Pain Interference, the interaction term was significant at a 95% confidence level (OR⫽ 0.631, 2(1) ⫽4.01, p⫽0.045, 95% confidence interval: 0.402– 0.99). This suggests that the assumption of homogeneity between treatment groups by pain status should not be accepted, and the association between pain and depression outcome should be assessed separately for intervention and usual care groups. For Arthritis Pain Interference, the interaction term was not significant (OR⫽0.776, 2(1)⫽ 0.653, p⫽0.419).
DISCUSSION
(OR⫽0.959, 2(1)⫽9.15, p⫽0.002). Higher baseline depression was significantly associated with better depression response in the Pain Interference model (OR⫽1.433, 2(1)⫽8.87, p ⫽0.003), but not in the Arthritis Pain Interference model (OR⫽1.048, 2(1) ⫽0.08, p⫽ 0.774). In the usual care group, after adjusting for the covariates, neither pain measure was significantly associated with depression response at 12 months (OR⫽1.037, 2(1)⫽0.02, p⫽0.860 for Pain Interference; OR⫽0.896, 2(1)⫽0.175, p⫽0.675 for Arthritis Pain Interference). In the Pain Interference model, only fewer baseline medical conditions (OR⫽0.876, 2(1) ⫽6.104, p⫽ 0.013) and higher baseline HSCL-20 score (OR⫽1.930, 2(1)⫽18.37, p ⬍0.001) were associated with higher odds of depression response. In the Arthritis Pain Interference model, the same two covariates were significant (OR⫽0.829, 2(1) ⫽ 5.378, p⫽0.020 for number of medical conditions; OR⫽2.202, 2(1)⫽12.746, p ⬍0.001 for baseline HSCL-20). The results in Table 3 suggest that treatment group might have a modifying effect on the association between baseline pain interference and 12month depression improvement. To examine the interaction of baseline pain with treatment group for
Am J Geriatr Psychiatry 15:8, August 2007
As had been suggested by the IMPACT care managers, intervention participants with high baseline pain interference measures had lower rates of depression response than those with no/low pain. In a subset of 1,002 subjects with arthritis, higher Arthritis Pain Interference had roughly the same effects on depression response in the intervention, but was less significant. The presence of arthritis was not in itself associated with differential depression outcomes. Adjusting for age, sex, race, and marital status did not diminish the association of baseline pain with depression response in the intervention group. Higher baseline depression was associated with better depression responses, but not in the subgroup of those with arthritis. Older age was significantly associated with a decreased depression response in all the models, although the effects of age were modest. Age was not significantly associated with pain status at baseline, and the difference of the mean age between the high and no/low pain groups was less than six months. The association of higher pain interference with worse depression response was not observed in the usual care group, where depression response rates were similarly low in both pain groups (18.8% for no/low versus 18.1% for high Pain Interference). The results of the regression analysis with the interaction term suggested that Pain Interference showed a significant differential effect in the collaborative care
703
Effectiveness of Depression Collaborative Care
TABLE 3.
Logistic Regression Models for the Intervention and Control Intervention
Pain Interference High vs low pain interference Older agea Higher baseline HSCL-20b No. of medical comorbidities Arthritis Pain Interference High vs low pain interference Older agea Higher baseline HSCL-20b No. of medical comorbidities Education: high school or less
Usual Care
Odds Ratio
Wald 2
df
p Value
Odds Ratio
Wald 2
df
p Value
0.641 0.970 1.433 0.932
8.22 9.01 8.87 3.07
1 1 1 1
0.004 0.003 0.003 0.080
1.037 0.980 1.930 0.876
0.02 2.57 18.37 6.10
1 1 1 1
0.860 0.109 ⬍0.001 0.013
0.658 0.959 1.048 0.932 0.665
4.54 9.15 0.08 1.40 4.34
1 1 1 1 1
0.033 0.002 0.774 0.236 0.037
0.896 0.969 2.202 0.829 0.986
0.175 3.22 12.74 5.38 0.01
1 1 1 1 1
0.675 0.073 ⬍0.001 0.020 0.956
Notes: Wald 2 test was used to compare depression response (50% reduction at 12 months), with pain category and baseline measures as covariates. Separate models are presented for the whole group and the subset of 1,002 arthritis subjects. Of the covariates age, sex, race, marital status, medical comorbidities, and baseline HSCL-20 score, only those found to be significant in either the treatment or the usual care group are listed. a Per increase in year from the minimum subject age of 65 years. b Per one point increase in HSCL-20 (on a 0 – 4 point scale). HSCL: Hopkins Symptom Checklist.
intervention compared with usual care. Arthritis Pain Interference did not show this interaction. The initial hypothesis that higher pain would be associated with less likely improvement in depression was not confirmed for the whole usual care group or the arthritis subgroup. Thus from the usual care data, there was no strong evidence that higher pain was generally associated with worse depression outcomes at 12 months. Although several of the other clinical variables assessed at baseline, including severity of depression, functional status, and number of chronic medical conditions, did not appear to diminish depression response in the intervention, baseline pain had a significant negative effect on depression response, which suggests a unique relationship between pain and depression in collaborative care treatment. The specific consequences of pain on depression outcomes could be from a number of factors, including more severe neurovegetative symptoms not captured fully by the HSCL-20, decreased ability to fully engage in pleasant events or behavioral activation such as physical or social activities, distraction from the depression intervention by pain, or a more fundamental neural correlation between being in pain and feeling depressed (Bair7 provides a discussion of the “depression-pain syndrome”). Regardless of the exact cause, our findings support the clinical observation of the study care managers who felt that phys-
704
ical pain was often a “sticking point” to depression response in this intervention. These findings differ from results published by Karp and colleagues,14 who concluded that after correcting for baseline depression, pain did not lower depression response rates or increase time to response in older adults. That trial involved only medications and was shorter in duration than the IMPACT intervention. In the usual-care arm of IMPACT, 57% of the subjects were receiving an antidepressant from their primary care physician at 12 months, and in this group baseline pain did not seem to be associated with a difference in depression response, which agrees with Karp’s findings. Yet in the intervention arm, 73% of subjects were receiving an antidepressant at 12 months and higher baseline pain was associated with significantly worse depression response. The IMPACT intervention involved more than medications, with all the intervention subjects receiving routine contact from care managers and 30% receiving problem-solving therapy. Pain might thus have different consequences for depression response in longer-term collaborative, multifaceted interventions such as IMPACT than in short-term antidepressant-only trials. Higher pain interference may inhibit the patient activation that is thought to be a crucial component of collaborative care, may distract patients from the content of the intervention, or may be associated with character-
Am J Geriatr Psychiatry 15:8, August 2007
Thielke et al. ologic traits that make depression response less likely.29 The recent finding that high perceived pain interference might be more closely associated with high medical costs than depression13 implies that depression with pain may have important qualitative differences from depression without pain. Our analysis corroborates this observation, and suggests that depression without pain may have fewer functional consequences and be more amenable to treatment than depression with pain. There may be qualitative differences in the “type” of depression in no/low and high-pain groups that is not captured in traditional measure of depression such as the HSCL-20, which focus more on psychological distress than on physical symptoms. Understanding the complex relationship between pain, depression, medical care, and treatment response merits further research. Our analysis raises the question of whether pain treatment might improve depression outcomes. There is to our knowledge no published evidence that independently treating pain improves outcomes for major depression in any age group. Some trials have shown improvement in mood with opioid treatment of chronic back pain30 and such findings suggest that if depressed older adults were to experience less pain, they might be more likely to respond to interventions for depression. Despite 58% of subjects at baseline describing moderate or greater pain interference, a previous analysis of IMPACT found that almost half of those with functional impairment from pain did not report using any analgesic medications.31 Pain was thus probably globally undertreated in this depressed cohort, and better treatment of it might increase the likelihood of depression response. Alternately, high perceived pain might be a marker of personality or attachment traits that would persist even if the pain were treated or nominally improved,32 and might continue to decrease the likelihood of improving with a depression intervention. Finally, we note that clinical efficacy studies of depression treatments (e.g., clinical drug trials) tend to produce better outcomes, with greater treatment effects, than effectiveness studies of depression interventions provided by usual providers in “real world” primary care settings. We believe that this
Am J Geriatr Psychiatry 15:8, August 2007
may be, in part, because efficacy studies often exclude patients with chronic medical illnesses that involve pain, medically ill patients with significant pain complaints, or patients who are taking regular medications.33 Our analyses suggest that some of this difference might derive from selection bias of subjects with less pain. From the current data, excluding those subjects who had moderate to high pain interference would result in roughly a 30% greater relative response in the intervention compared to the whole group, with little change in response in the control group. Such an exclusion would create the appearance of a greater intervention effect, but it would also ignore about 60% of the entire sample of depressed older primary care patients. Limitations This study has several limitations. Pain was assessed by self-report, single-item measures. There were no subjective measures of Activity of Daily Living impairment, or observation-based assessments of functional capacity. From the current data, there was no way to ascertain or correct for the use of analgesic medications. Treatment response, a 50% reduction in depression (HSCL-20) at 12 months, was used as the primary outcome measure instead of remission (e.g., a value below a defined threshold, such as 0.5 on the HSCL-20), but because the HSCL-20 and pain measures were cross-sectionally correlated, using an absolute score threshold would have increased the possibility of confounding.
CONCLUSIONS Unlike a prior study showing that pain does not interfere with response to depression treatment in older adults, our study indicated that higher baseline pain, from arthritis or other causes, was associated with worse depression response during a 12-month collaborative care intervention. This association was not observed in controls who tended to have low response rates regardless of baseline pain status. The association of more pain with worse depression response in intervention subjects persisted even when adjusting for sociodemographic variables, medical comorbidities, and baseline depression. A significant interaction term indicated that the pain had a differ-
705
Effectiveness of Depression Collaborative Care ential effect in the intervention compared with usual care. Together, these findings suggest that pain may be a unique barrier to depression response during collaborative care interventions for depression. In conjunction with recent findings about the central importance of pain to medical costs,13 this work highlights the broad importance of pain to aging and mental health. To our knowledge, no research has determined if reducing pain would also bolster depression treatment, although this may be an action-
able means of improving outcomes in a substantial subset of older adults with depression.
This work was supported by the National Research Service Award T-32 Fellowship (SMT), the John A Hartford Foundation (JU), the California HealthCare Foundation (JU), and the Paul Beeson Physician Faculty Scholars Award from the American Federation for Aging Research (JU).
REFERENCES 1. Centers for Disease Control: Projected Prevalence of Self-Reported Arthritis or Chronic Joint Symptoms Among Persons 65 years—United States, 2005-2030. MMRW; 2003 2. Yelin E, Callahan LF: The economic cost and social and psychological impact of musculoskeletal conditions. National Arthritis Data Work Groups. Arthritis Rheum 1995; 38:1351–1362 3. Thomas E, Peat G, Harris L, et al: The prevalence of pain and pain interference in a general population of older adults: cross-sectional findings from the North Staffordshire Osteoarthritis Project (NorStOP). Pain 2004; 110:361–368 4. Stang PE, Brandenburg NA, Lane MC, et al: Mental and physical comorbid conditions and days in role among persons with arthritis. Psychosom Med 2006; 68:152–158 5. Verbrugge LM, Juarez L: Profile of arthritis disability: II. Arthritis Rheum 2006; 55:102–113 6. Arnow BA, Hunkeler EM, Blasey CM, et al: Comorbid depression, chronic pain, and disability in primary care. Psychosom Med 2006; 68:262–268 7. Bair MJ, Robinson RL, Katon W, Kroenke K: Depression and pain comorbidity: a literature review. Arch Intern Med 2003; 163: 2433–2445 8. Ohayon MM, Schatzberg AF: Using chronic pain to predict depressive morbidity in the general population. Arch Gen Psychiatry 2003; 60:39–47 9. Emptage NP, Sturm R, Robinson RL: Depression and comorbid pain as predictors of disability, employment, insurance status, and health care costs. Psychiatr Serv 2005; 56:468–474 10. Bao Y, Sturm R, Croghan TW: A national study of the effect of chronic pain on the use of health care by depressed persons. Psychiatr Serv 2003; 54:693–697 11. Mossey JM, Gallagher RM: The longitudinal occurrence and impact of comorbid chronic pain and chronic depression over two years in continuing care retirement community residents. Pain Med 2004; 5:335–348 12. Mossey JM, Gallagher RM, Tirumalasetti F: The effects of pain and depression on physical functioning in elderly residents of a continuing care retirement community. Pain Med 2000; 1:340–350 13. Gameroff MJ, Olfson M: Major depressive disorder, somatic pain, and health care costs in an urban primary care practice. J Clin Psychiatry 2006; 67:1232–1239 14. Bair MJ, Robinson RL, Eckert GJ, et al: Impact of pain on depression treatment response in primary care. Psychosom Med 2004; 66:17–22 15. Karp JF, Weiner D, Seligman K, et al: Body pain and treatment
706
response in late-life depression. Am J Geriatr Psychiatry 2005; 13:188–194 16. Karp JF, Scott J, Houck P, et al: Pain predicts longer time to remission during treatment of recurrent depression. J Clin Psychiatry 2005; 66:591–597 17. Unu ¨ tzer J, Katon W, Callahan CM, et al: IMPACT Investigators. Improving Mood-Promoting Access to Collaborative Treatment. Collaborative care management of late-life depression in the primary care setting: a randomized controlled trial. JAMA 2002; 288:2836–2845 18. Lin EH, Katon W, Von Korff M, et al: IMPACT Investigators. Effect of improving depression care on pain and functional outcomes among older adults with arthritis: a randomized controlled trial. JAMA 2003; 290:2428–2429 19. Lin EH, Tang L, Katon W, et al: Arthritis pain and disability: response to collaborative depression care. Gen Hosp Psychiatry 2006; 28:482–486 20. Callahan CM, Kroenke K, Counsell SR, et al: IMPACT Investigators. Treatment of depression improves physical functioning in older adults. J Am Geriatr Soc 2005; 53:367–373 21. Hunkeler EM, Katon W, Tang L, et al: Long term outcomes from the IMPACT randomised trial for depressed elderly patients in primary care. BMJ 2006; 332:259–263 22. Arean PA, Ayalon L, Hunkeler E, et al: IMPACT Investigators. Improving depression care for older, minority patients in primary care. Med Care 2005; 43:381–390 23. Steffens DC, Snowden M, Fan MY, et al: IMPACT Investigators. Cognitive impairment and depression outcomes in the IMPACT study. Am J Geriatr Psychiatry 2006; 14:401–409 24. Harpole LH, Williams JW Jr, Olsen MK, et al: Improving depression outcomes in older adults with comorbid medical illness. Gen Hosp Psychiatry 2005; 27:4–12 25. Katon WJ, Schoenbaum M, Fan MY, et al: Cost-effectiveness of improving primary care treatment of late-life depression. Arch Gen Psychiatry 2005; 62:1313–1320 26. Tang L, Song J, Belin TR, Unutzer J: A comparison of imputation methods in a longitudinal randomized clinical trial. Stat Med 2005; 24:2111–2128 27. Derogatis LR, Lipman RS, Rickels K, et al: The Hopkins Symptom Checklist (HSCL): a self-report symptom inventory. Behav Sci 1974; 19:1–15 28. Von Korff M, Jensen MP, Karoly P: Assessing global pain severity by self-report in clinical and health services research. Spine 2000; 25:3140–3151 29. Asghari A, Nicholas MK: Personality and pain-related beliefs/
Am J Geriatr Psychiatry 15:8, August 2007
Thielke et al. coping strategies: a prospective study. Clin J Pain 2006; 22: 10–18 30. Jamison RN, Raymond SA, Slawsby EA, et al: Opioid therapy for chronic noncancer back pain. A randomized prospective study. Spine 1998; 23:2591–2600 31. Unu ¨ tzer J, Ferrell B, Lin EH, Marmon T: Pharmacotherapy of pain in depressed older adults. J Am Geriatr Soc 2004; 52:1916–1922
Am J Geriatr Psychiatry 15:8, August 2007
32. Ciechanowski P, Sullivan M, Jensen M, et al: The relationship of attachment style to depression, catastrophizing and health care utilization in patients with chronic pain. Pain 2003; 104:627–637 33. Keitner GI, Posternak MA, Ryan CE: How many subjects with major depressive disorder meet eligibility requirements of an antidepressant efficacy trial? J Clin Psychiatry 2003; 64:1091– 1093
707