Depressive symptoms and bodily pain: the role of physical disability and social stress

Authors


Abstract

This study evaluates the bi-directional association between depressive symptoms and bodily pain, and examines the role of physical disability and perceived social stress in the depression–pain relationship. Data are employed from a two-wave panel study of Miami-Dade county residents (n = 1459) that includes a substantial over-sampling of individuals who identify as physically-disabled. Findings indicate that the bi-directional relationship between depression and pain is similar for those with and without a physical disability. Results also demonstrate that stress exposure, specifically recent life events and daily discrimination, partially mediated the relationship between prior levels of depression and changes in pain. Directions for future research and the need for a more comprehensive model of health incorporating physical, psychological, and social factors are discussed. Copyright © 2010 John Wiley & Sons, Ltd.

Introduction

The co-occurrence of depression and bodily pain is a major public health issue in the United States constituting substantial economic, social and personal costs (Gatchel, 2004). Although the association between depression and pain is consistently observed in the health literature (Baune, Caniato, Garcia-Alcaraz, & Berger, 2008; Benjamin, Morris, McBeth, MacFarlane, & Silman, 2000; Gamsa, 1990; Gayman, Turner, & Cui, 2008; Härmä, Kaltiala-Heino, Rimpelä, & Rantanen, 2002; Lindsay & Wyckoff, 1981; Magni, Caldleron, Rigatti-Luchini, & Merskey, 1990; Magni, Marchetti, Morechi, Merskey, & Luchini, 1993; Mathew, Weinman, & Mirabi, 1981), the nature of this relationship is still not well understood. Indeed, further investigation into the depression–pain relationship remains of both theoretical and practical significance (Bair, Robinson, Katon, & Kroenke, 2003; Romano & Turner, 1985).

The current investigation aims to further our understanding of the relationship between depression and pain by addressing three study objectives: First, evaluate the bi-directional relationship by modeling both pathways simultaneously using path analysis, as well as control for disability status in order to rule out the interpretation of spuriousness. Second, test whether the bi-directional relationship between depression and pain is moderated by psychical disability status. Third, evaluate the mediating effects of social stress in the depression–pain relationship.

Bi-directional relationship between depression and pain

Recent studies employing longitudinal designs have considered the temporal order of the association between depression and pain. While findings indicate that pain predicts depression over time (Bair et al., 2004; Breslau & Davis, 1992; Brown, 1990; Corruble & Guelfi, 2000; Gallagher & Verna, 1999; Von Korff, Deyo, Cherkin, & Barlow, 1993) and that pain appears to be a stronger predictor of depression than vice versa (Cairns, Adkins, & Scott, 1996; Fishbain, Cutler, Rosomoff, & Steele-Rosomoff, 1997), evidence also demonstrates that depression is associated with an increase in pain (Leino and Magni, 1993; Ruoff, 1996; Von Korff, Ormel, Katon, & Lin, 1991). Breslau, Davis, Schultz, and Peterson (1994) found that prior pain predicted an increase in risk for major depression and that depression predicted an increase in bodily pain. Consequently, researchers have concluded that the relationship between depression and pain is bi-directional (Bair et al., 2003; Leino & Magni, 1993; Von Korff & Simon, 1996). However, because a true test of a bi-directional association necessitates that both pathways are modeled simultaneously researchers cannot answer with confidence whether the depression–pain relationship is indeed bi-directional.

Moderation of physical disability

There are also grounds to anticipate that the relationship between depression and pain may be moderated (or conditioned) by disability status: whereby the relationship varies significantly for those with and without a physical disability. Prior research demonstrates that persons with physical disabilities are more prone to experiencing co-occurring pain and depression (Dickens, Jayson, Sutton, & Creed, 2000; Gatchel & Dersh, 2002; Polatin, Kinney, Gatchel, Lillo, & Mayer, 1993). Persons with a disability who experience considerable bodily pain may have greater difficulty in meeting the requirements of social roles, and consequently, increase risk for depressive mood. Similarly, persons with disabilities who experience higher levels of depressive symptoms may experience greater obstacles in the ability and motivation to engage in physical and social activities conducive to physical well-being, thus increasing risk for bodily pain. However, due to the paucity of research specifically testing for a bi-directional association between depression and pain, the potential significance of disability for understanding the shared linkage between these health indicators has also received little attention.

Mediation of social stress

The hypothesis that pain may increase risk for depression and depression may increase risk for pain, vis-à-vis associated difficulties in overcoming life's hardships, is consistent with research in the stress process tradition. Stress process theory posits that social experiences are translated into health outcomes through exposure to social stress and one's ability to effectively cope with stress (Pearlin, 1989; Pearlin, Lieberman, Menaghan, & Mullan, 1981). Because depression is associated with increased stress exposure (Hammen, 1991, 2006) and psychological stress is related to physical pain (Diepenmaat, van der Wal, de Vet, & Hirasing, 2006; Hwang et al., 2008; Ullrich, Turner, Ciol, & Berger, 2005), social stress may mediate the relationship between prior depression and changes in pain. Conversely, to the extent that bodily pain increases risk for failure to meet the requirements of social roles, pain may increase risk for role-related chronic stressors, and in turn, increase depressive mood.

The potential mediating effects of social stress in the depression–pain relationship is also supported by research examining shared physiological attributes between depression and pain (Bair et al., 2003) and shared biological pathways that respond similarly to clinical treatment (Blier & Abbott, 2001; Gallagher & Verna, 1999). Evidence from research on allostatic load (Geronimus, Hicken, Keene, & Bound, 2006; McEwen & Seeman, 1999; Seeman, Singer, Rowe, Horwitz, & McEwen, 1997) and psychoneuroimmunology (Benham, Nash, & Baldwin, 2009; Glaser, 2005; Fan et al., 2009) have specified a range of biomarkers that are affected by social stress and that mediate the linkage between stress exposure and health. Schell, Theorell, Hasson, Arnetz, and Saraste (2008) recently demonstrated an association between stress biomarkers and bodily pain. In the case of depression, stress has been shown to influence neuroendocrine responses that are correlates or expressions of depressive symptoms (Nierop, Bratsikas, Zimmerman, & Ehlert, 2006; Robles, Glaser, & Kiecolt-Glaser 2005). Because symptoms of depression and pain may follow the same pathways in the central nervous system (Bair et al., 2003), it seems likely that stress plays an important role in shaping the influence of depression on pain and vice versa.

Method

Study procedures and sample

The data employed in this study were from a two-wave (W1 and W2) panel study of Miami Dade County residents that included a substantial oversampling of individuals with a physical disability. This study was approved by the Institutional Review Board at the University of Miami and has been described in detail elsewhere (Gayman et al., 2008; Turner, Lloyd, & Taylor, 2006). Described in the sampling framework illustrated below (Figure 1), a total of 10,000 randomly selected households were screened with respect to gender, age, ethnicity, disability status and language preference. Using this sampling frame, the study sample was drawn such that there were even numbers of women and men, even numbers of people screened as having a physical disability and those not, and equivalent numbers of the four major ethnic groups comprising more than 90 percent of all Miami-Dade County residents (non-Hispanic Whites, Cubans, non-Cuban Hispanics and African Americans). The presence of disability status was established in response to the question: ‘Do any adults in the household have a physical health condition or physical handicap that has resulted in a change in their daily routine or that limits the kind of or amount of activity they can carry out?’

Figure 1.

Sampling framework—Physical challenge and health study

From 2000–2001, 1986 interviews were completed, with a success rate of 82 percent. Included in the study were 1086 adults who screened as having no physical disability and 900 individuals who screened as having a disability. Of the 900 reported by a family member as having activity limitations within the screening process, only 559 confirmed this status within the actual interview. Presumably, this discrepancy arises from differing views about the level of activity limitation that defines disability.

Second-wave interviews of 1495 W1 participants were completed approximately three years later (2003–2004). Excluding 100 W1 participants who had died and 59 who were too ill to be interviewed, the second-wave success rate was 82.5 percent. Those lost to attrition at W2 reported lower levels of social stress, socioeconomic status and pain, and slightly higher levels of both physical limitations and depressive symptoms at W1. The magnitude of these differences varied from 0.10  SD to slightly more than 0.30  SD. In our view, the nature and magnitude of these differences do not seriously challenge the representativeness of the subpopulation employed in these analyses or in any way suggest that selection biases may have accounted for the results presented.

A total of 1459 study participants provided valid data across all study variables employed in the current analysis. The working sample included 786 females (54%), 345 non-Hispanic Whites (24%), 351 Cuban-Americans (24%), 320 non-Cuban Hispanics (22%) and 443 African Americans (30%). The sample distribution by physical disability status was 410 disabled (28%) and 1049 non-disabled (72%). The average age of respondents at Wave 1 was 57 years (SD = 17 years; range = 18–93 years).

Measures

Depressive symptoms

Depressive symptomatology was assessed at W1 and W2 using the highly reliable and widely used Center for Epidemiologic Studies Depression Scale (CES-D) (Radloff, 1977). From the original 20-item CES-D index, six items related to somatic problems were omitted from analyses to avoid measurement confounding between depression and physical health (Ormel, Rijsdijk, Sullivan, van Sonderen, & Kempen, 2002). These omitted items include problems with eating, keeping your mind on what you are doing, effort, restlessness, talking and getting going. In completing the CES-D items, participants were asked about their experiences over the past month using response categories of ‘not at all’ (1) to ‘almost all the time’ (4) (range = 0–56). Items were coded such that higher scores indicated higher levels of depressive symptoms and summed for analysis (Wave 1 α = 0.81; Wave 2 α = 0.84).

Bodily pain

Indicators of physical pain were assessed by asking respondents at W1 and W2 whether they had experienced any bodily pain over the preceding 4 weeks. Those who answered affirmatively were questioned on the frequency and average intensity of pain experienced. Response categories for the intensity item ranged from ‘very mild’ (1) to ‘very severe’ (5). Those for frequency ranged from ‘once or twice’ (1) to ‘everyday or almost everyday’ (5). Respondents who reported no pain over the past four weeks were coded as ‘none’ (0) for intensity of pain and ‘never’ (0) for frequency of pain. Both items were coded so that higher values indicate greater intensity and frequency of pain. While the measurement of pain varies widely across studies, the severity of pain has important implications for depression (Bair et al., 2003). As used in prior research (Gayman et al., 2008), we derive the pain measure by multiplying scores on the two dimensions as an indicator of pain severity (range = 0–25).

Social stress

Measured at W1 and W2, stress exposure was measured using three separate indicators of perceived social stress: chronic stress, daily discrimination and recent life events. The chronic stress measure is an adaptation of Wheaton's (1994) scale, modified to better capture the kinds of enduring stressors older individuals are likely to experience. This index includes 39 items (0–1 scale; range = 0–39) relating to general experiences, employment/unemployment, intimate partnerships/no partners, children, recreation and health concerns. Daily discrimination (nine items; 1–5-point scale; range = 9–45) and life events within the past 12-moths (32 items; 0–1 scale; range = 0–32) were indexed by previously employed measures (Turner & Avison, 2003; Williams, Yu, Jackson, & Anderson, 1997). Indicators of daily discrimination include statements such as ‘You are treated with less courtesy than other people’, ‘People act as if they think you are dishonest’ and ‘You are called names or insulted’. Recent life events involve undesirable experiences that occur to the respondent or those close to them and include, for example: being in a serious accident, death of spouse/partner, death of child and major financial crisis. A detailed list of all stress items included in this study was previously published by Turner and Avison (2003).

Physical disability status

At Wave 1 interviewers asked the following question of respondents who were classified as having a disability in the screening process: ‘Earlier someone in your household [or the respondent] told us that you had a condition or physical health problem that limits the kind or amount of activity that you can carry out (such as work, housework, school, recreation, shopping, or participation in social or community activities). I just want to confirm with you now whether that is correct’. Respondents who responded affirmatively were coded as 1 for disability; those who did not were coded as 0.

Controls

The control variables were assessed at Wave 1 and included gender, age, race-ethnicity and socioeconomic status (SES). Gender is coded 1 for females and 0 for males. Age was measured in years as a continuous variable. Race-ethnicity is measured using a set of dummy indicators for non-Hispanic Whites (reference group), African Americans, Cubans and non-Cuban Hispanics. Estimates of SES were based on a composite score that equally weights occupational level (Hollingshead, 1957), educational attainment, and household income of each participant. To avoid problems of missing data, particularly on the variable of household income, scores on each of these dimensions were standardized, summed and divided by the number of dimensions on which data were available.

Statistical analyses

We began with descriptive analyses comparing study variables by disability status for Waves 1 and 2 (Table I). We then examined bivariate associations for the study variables using a correlation matrix (Table II). Next, using AMOS software (Arbuckle, 2005), we conducted path analyses to test for bi-directional associations between depression and pain over a 3-year period, independent of sociodemographic controls and disability status (Figure 2, Objective 1). Specifically, we simultaneously evaluated the association between pain at Wave 1 and changes in depression from Wave 1 to Wave 2 and depression at Wave 1 and changes in pain from Wave 1 to Wave 2. Path analyses are particularly useful when assessing bi-directional relationships as multiple outcomes can be modeled simultaneously (see Bollen, 1989; Kline, 2005).

Figure 2.

Bi-directional relationship between depressive symptoms and bodily pain. Standardized regression coefficients are shown. N = 1986. Age, gender, race-ethnicity, socioeconomic status and physical disability are controlled for in the model but not shown. Model is saturated. W: wave; e1: error 1; e2: error 2. *p ≤ 0.05; **p ≤ 0.01

Table I.  Description of study variables by disability status
VariablePhysical disabilityNo physical disabilityTotal
(N = 410)(N = 1049)(N = 1459)
Wave 1Wave 2Wave 1Wave 2Wave 1Wave 2
Mean (SD)Mean (SD)Mean (SD)Mean (SD)Mean (SD)Mean (SD)
  • *

    * p < 0.05;

  • **

    ** p < 0.01.

  •  Frequency × intensity of bodily pain over previous four weeks (range = 0–25).

  •  Center for Epidemiologic Studies—Depression scale (range = 0–56).

  • §

    § Range = 0–39.

  •  Range = 0–32.

  • ††

    †† Range = 9–45.

  • Notes: N = 1459. Shown are unstandardized mean scores (SD) for those who participated at both study waves and had valid data on all study variables. Significance tests compare those who self-confirm a physical disability to those without a physical disability by study wave.

Bodily pain8.63**5.24**2.711.904.372.84
(8.28)(6.12)(5.49)(4.03)(6.92)(4.94)
Depressive symptoms23.26**23.24**20.9320.8621.5921.53
(7.07)(7.25)(6.39)(6.34)(6.67)(6.69)
Chronic stress§4.90*5.07**4.216.274.405.93
(5.24)(5.15)(4.77)(5.39)(4.92)(5.35)
Recent life events1.19*.99.99.861.05.89
(1.54)(1.49)(1.49)(1.26)(1.51)(1.33)
Daily discrimination††14.41**14.30**13.3013.3813.6113.64
(5.54)(5.36)(4.81)(4.74)(5.05)(4.94)
Table II.  Correlation matrix for study variables
  (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)
  1. Notes: N = 1459. Only significant correlations are shown (p ≤ 0.05) for those who participated at both study waves and had valid data on all study variables.

  2. W: wave; P: frequency × intensity of bodily pain; DS: depressive symptoms (Center for Epidemiologic Studies—Depression scale); CS: chronic stress; RLE: recent life events; DD: daily discrimination; PD: self-reported physical disability.

(1)P (W1)          
(2)P (W2)0.42         
(3)DS (W1)0.120.12        
(4)DS (W2)0.090.160.35       
(5)CS (W1)0.160.060.250.12      
(6)CS (W2)−0.05 0.08 0.29     
(7)RLE (W1)0.140.100.18 0.470.15    
(8)RLE (W2)0.050.130.080.050.180.370.18   
(9)DD (W1)0.12 0.15 0.340.170.290.14  
(10)DD (W2) 0.06 0.070.120.300.130.170.35 
(11)PD (W1)0.380.300.160.160.06−0.100.06 0.100.08

To evaluate the moderating effects of disability status on the depression–pain relationship, the model from Objective 1 was estimated separately for those with and without a physical disability (results reported in text only, Objective 2). Using multiple group comparison (i.e. chi-square test), we compared the separate models to evaluate whether the bi-directional relationship between depression and pain significantly differed for those with and without a physical disability.

Path analyses were also employed in order to evaluate the mediating effects of Wave 2 stress exposure in relation to Wave 1 depression predicting changes in pain from Wave 1 to Wave 2, and vice versa (Figures 3 and 4, Objective 3). We used Sobel tests (Sobel, 1982) to formally assess the significance of the mediating effects. These tests determine the level of mediation based upon four factors: (1) the extent to which the independent variable influences the mediator; (2) the extent to which the independent variable affects the dependent variable in the absence of the mediator; (3) the extent to which a mediator has a unique effect on the dependent variable and (4) the influence of the independent variable on the dependent variable controlling for the mediating variable in the model (Mackinnon & Dwyer, 1993).

Figure 3.

Depressive symptoms predicting changes in bodily pain: mediating effects of social stress. Standardized regression coefficients are shown. N = 1986. Age, gender, race-ethnicity, socioeconomic status, and physical disability are controlled for in the model but not shown. Model is saturated. Mediation effects using Sobel test: chronic stress z = 0.359, p = .720; recent life events z = 2.793, p = 0.005; daily discrimination z = 1.939, p = 0.052. W: wave. *p ≤ 0.05; **p ≤ 0.01

Figure 4.

Bodily pain predicting changes in depressive symptoms: mediating effects of social stress. Standardized regression coefficients are shown. N = 1986. Age, gender, race-ethnicity, socioeconomic status and physical disability are controlled for in the model but not shown. Model is saturated. W: wave. *p ≤ 0.05; **p ≤ 0.01

Finally, because the data showed little indication of bias due to selective attrition, rather than deleting cases with missing data at Wave 2, the present study used Full Information Maximum Likelihood (FIML) to test the predicted relationships among constructs. FIML (Little & Rubin, 1987; Rubin, 1976; Schafer, 1997) computes maximum likelihood estimates and standard errors for the path models from data with missing values. This approach provides efficient estimation of statistical parameters from incomplete data, allowing the retention of the complete sample for analysis. Parameter estimates from FIML provide less biased information than ad hoc procedures such as listwise deletion, pairwise deletion or imputation of means (Schafer, 1997). Consequently, all 1986 Wave 1 participants are included in the path analyses.

Results

Descriptives

Table I provides descriptive statistics for study variables at each wave by disability status. Results indicate that persons who self-identify as having a physical disability were disadvantaged in terms of pain, depression and stress exposure at both Wave 1 and 2. For example, persons with a physical disability reported significantly greater bodily pain at Wave 1 (M = 8.63) and Wave 2 (M = 2.71) than their non-physically limited counterparts (M = 5.24 and M = 1.90, respectively).

Table II provides correlations among study variables; indicating significant bivariate associations between study variables at both Wave 1 and 2. Of interest, Wave 1 pain was associated with Wave 2 depressive symptoms, and vice versa. The results also showed significant stability in both pain (r = 0.42, p < 0.001) and depressive symptoms (r = 0.35, p < 0.001) over a 3-year period. In addition, measures of social stress were associated with both pain and depressive symptoms. Together, the correlations provided initial support for the bi-directional and mediation hypotheses.

Bi-directional relationship between depression and pain

Figure 2 provides results from path analysis examining the hypothesized bi-directional association between depressive symptoms and bodily pain. We included age, gender, race/ethnicity, SES and disability status as control variables but do not show them in order to more clearly convey the primary findings. Several findings are noteworthy. Firstly, controlling for demographics and disability status, the stability coefficients for depression and pain remained statistically significant for depressive symptoms (β = 0.27, p < 0.01) and pain (β = 0.33, p < 0.01). Secondly, independent of disability status, the level of depressive symptoms at Wave 1 predicts changes in bodily pain from Wave 1 to Wave 2 (β = 0.05, p < 0.05), and vice versa (β = 0.04, p < 0.05).

Moderation of physical disability

In order to evaluate whether the bi-directional association between depression and pain is moderated by disability status, multiple group comparisons were conducted (not shown). Specifically, the sample was divided by physical status to create two subsamples: one included those with a physical disability (N = 559) and the other included those without a physical disability (N = 1427). First the model in Figure 2 was evaluated for both groups simultaneously with corresponding paths in the model for the two groups freely estimated. Second, corresponding pairs of paths were constrained to be equal and chi-square changes between models were evaluated to test the difference. The findings indicate that the bi-directional relationship between depression and pain was statistically equivalent—in both direction and strength—for those with and without a physical disability. In other words, the relationship between depression and pain is both bi-directional and equally robust for those with and without a physical disability.

Mediation of social stress

We next tested the mediating hypotheses. We proposed that the association between prior depression and changes in pain, and vice versa, would be attenuated by Wave 2 social stress (Figures 3 and 4). Again, we included age, gender, race/ethnicity, SES and disability status as control variables (not shown).

Figure shows that Wave 1 depression predicted each measure of Wave 2 stress and that Wave 2 stress (recent life events, daily discrimination) was associated with Wave 2 pain. Compared with Figure 2, the direct path from Wave 1 depression to Wave 2 pain was reduced from 0.05 to 0.03. With respect to mediating effects, Sobel tests revealed that both recent life events (z = 2.792, p = 0.005) and daily discrimination (z = 1.939; p = 0.052) significantly mediated the relationship between prior depression and changes in bodily pain. Chronic stress, however, was not found to be a significant mediator between prior depression and changes in bodily pain.

Conversely, Figure evaluated the mediating effects of social stress for the association between prior pain and changes in depressive symptoms. With demographic controls and all Wave 2 stress measures modeled simultaneously, Wave 1 pain was not associated with Wave 2 stressors and stress was not associated with Wave 2 depressive symptoms. In addition to revealing no changes in the relationship between prior pain and changes in depression, the results failed to meet the requirements of mediation effects. We therefore conclude that the association between prior pain and changes in depressive symptoms was not mediated by the stress indictors in this study.

Discussion

The present study addressed three research questions. Is the relationship between depressive symptoms and bodily pain bi-directional, independent of sociodemographic characteristics and physical disability status? Is the bi-directional association between depression and pain moderated by disability status? And, does perceived social stress mediate the depression–pain relationship?

Regarding the first research question, the current investigation builds on previous studies (Breslau et al., 1994) by simultaneously modeling the pathway from depression to pain and from pain to depression. The results support the conclusion that the relationship between depression and pain is bi-directional. In addition, although prior research has shown that the depression–pain relationship is partially explained by disability status (Dickens et al., 2000), the current investigation demonstrates that the bi-directional relationship between depression and pain is independent of physical disability status.

It should be noted that, although the reported effect sizes for the cross-lagged associations for depression and pain were modest, small bi-directional effects are common in this kind of model given the high stability coefficients between adjacent waves of data (Cui, Donnellaan, & Conger, 2007). Moreover, small effects are not necessarily indicative of trivial associations, especially when such relationships reflect an ongoing process that is related to the health and well-being of individuals [see Rosenthal and Rubin (1979) for a discussion of the importance of small effects]. Simply put, the observed bi-directional relationship may have a substantial and clinically significant impact on individual functioning over time. In addition, given that the data employed here were based on a community sample, the bi-directional and exacerbating effects of depression on pain (and vice versa) may be even more robust among clinical populations.

The bi-directionality between depression and pain identified here is in contrast with a recent report employing the same sample but that assessed the relationship between depressive symptoms and physical limitations. Specifically, Gayman and colleagues (2008) demonstrated that physical limitations predict changes in depressive symptoms, but found no evidence of the reverse association over the 3-year period. They concluded that the translation of depression to physical limitations is likely to occur over an extended period of time.

In light of the current findings, we posit that the influence of depression on bodily pain, and vice versa, may reflect a more pervasive linkage between mental and physical health—a linkage that may arise partly from shared biological pathways, as previously noted (Blier & Abbott, 2001; Gallagher & Verna, 1999). Thus, the time it takes for depression to translate into perceptions of pain may be relatively short, at least compared with the time needed for depression to manifest into physical limitations. In addition, depression may translate into physical limitations through experiences of bodily pain. Though data limitations prevent us from testing these hypotheses, we recommend future research to consider the linkage between depressive symptoms, bodily pain and physical limitations over longer periods of time and the role of pain in the translation of depression to physical limitations.

The second research question assessed whether the bi-directional association between depression and pain was moderated by disability status. The findings revealed that the bi-directional relationship between depression and pain was not conditioned by disability status. Although persons with a disability may be at increased risk for co-occurring pain and depression (Dickens et al., 2000; Gatchel & Dersh, 2002; Polatin et al., 1993), the translation of depressive symptoms to bodily pain, and vice versa, appears to be equally robust for those with and without a physical disability. We posited that persons with a disability experiencing bodily pain may have greater difficulty in meeting the requirements of social roles, and consequently, face increased risk for depressive mood. We also suggested that persons with a disability experiencing higher levels of depressive symptoms may confront greater obstacles in their ability and motivation to engage in activities that are conducive to physical well-being, and thus, experience an increased risk for bodily pain. However, given our results, it may be that persons with a physical disability are only at increased risk for the translation of depressive symptoms into pain, and vice versa, in the presence of difficulties with fulfilling the requirements of social roles and engaging in physical/social activities. Further research is needed to evaluate the potential moderating and/or mediating effects of such difficulties in the relationship between depressive symptoms and bodily pain by disability status.

The third research question evaluated the potential mediating effects of social stress in the depression–pain relationship. The findings demonstrated that stress exposure plays a linking role in the depression–pain relationship. Specifically, exposure to social stressors such as recent life events and daily discrimination partially mediated the association between prior depressive symptoms and changes in pain. Consistent with research indicating an increased risk for stress exposure among persons with depression (Hammen, 1991, 2006) and a linkage between psychological stress and bodily pain (Hwang et al., 2008), our findings indicate that depressed persons are at heightened risk for stress exposure that, in turn, increases risk for bodily pain.

Stress exposure did not mediate the relation-ship between prior pain and changes in depressive symptoms, however. It is possible that the translation of prior pain to depression via stress may only be realized in the presence of activity limitations (Von Korff & Simon, 1996) and when social requirements are not met. In addition, given the association between bodily pain and stress biomarkers (Schell et al., 2008), the extent to which bodily pain translates into depressed mood may depend on physiological consequences of pain.

Limitations

Several limitations to the current investigation are worthy of note. Firstly, our measure of bodily pain is limited to the multiplicative of two indicators and refers to reports of general pain not specific to a given part of the body. Pain emanating from certain areas of the body may increase risk for depressed mood. However, studies employing different measures and specified locations of pain support the conclusion that physical pain, in general, is consistently linked to depression (Bair et al., 2003).

Secondly, analyses presented only considered the role of depressive symptomatology, though it should be noted that other forms of psychological distress such as anxiety have also been shown to predict levels of pain (Casten et al., 1995). Although data limitations prevented us from considering the significance of contemporaneous anxiety symptomatology, preliminary analyses controlling for lifetime major depression, anxiety disorders and substance use dependence produced similar results.

Thirdly, the relationship between physical and mental health may also stem from state dependence bias—those who currently experience higher levels of depressive symptoms, for example, report higher levels of social stress and bodily pain. Although we can not rule out this alterative interpretation, the longitudinal nature of the study helps temper this concern. At a minimum, this builds upon prior cross-sectional studies on the relationship between depression and bodily pain.

Conclusion

Such limitations notwithstanding, the present study builds on prior research in several important ways. Firstly, it offers a prospective assessment to evaluate the bi-directional relationship between depressive symptoms and bodily pain by modeling both pathways simultaneously. Secondly, the oversampling of persons with a physical disability living in the community provides a unique opportunity to evaluate whether the bi-directional relationship between depression and pain is an artifact of disability status and/or whether the relationship is moderated by disability status. In this vein, results from this study can be generalized to the community from which the sample was drawn, as opposed to being limited to a clinical population characteristic of many prior studies. Thirdly, few community-based studies account for such a wide array of stressors allowing for a more comprehensive understanding of the mediating effects of social stress in the depression–pain relationship.

In sum, the findings leave little doubt that the relationship between depression and pain is, indeed, bi-directional and that this bi-directional linkage is not an artifact of having a physical disability. In addition, the bi-directional relationship between depression and pain was similar for those with and without a physical disability, further highlighting the pervasive nature of the relationship. The findings also demonstrate that stress exposure among those with a history of depression has important implications for experiencing persistent and increasing levels of physical pain. Although the current investigation could not account for biological indicators of health, taken together, the findings support the call for a more comprehensive model of health incorporating physical, psychological and social factors.

Acknowledgements

The authors would like to thank Dr. Tim Carey for his critiques on an earlier draft of this manuscript. This work was supported by grants R01DA13292 and R01DA16429 from The National Institute on Drug Abuse to R. Jay Turner. This research was also supported by the National Research Service Award Postdoctoral Traineeship from the National Institute of Mental Health sponsored by Cecil G. Sheps Center for Health Services Research, University of North Carolina at Chapel Hill (Grant No: T32 MH019117).

Ancillary