| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLES |
From the Departments of Psychiatry (D.S.C.) and Anesthesiology (N.J., E.B.B.), University of Medicine and Dentistry–New Jersey Medical School, Newark, New Jersey.
Address reprint requests to: Donald S. Ciccone, Department of Psychiatry, University of Medicine and Dentistry–New Jersey Medical School, 30 Bergen St., ADMC Bldg. 14, Newark, NJ 07107. Email: cicconds{at}umdnj.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
METHODS: A 2 x 2 factorial design was used in which patients were retrospectively assigned to one of four independent groups: low economic/low social reward, low economic/high social reward, high economic/low social reward, and high economic/high social reward. Of 265 consecutive patients enrolled at a tertiary pain service, 75 met eligibility criteria and had chronic nonmalignant back pain.
RESULTS: Preexisting differences in health status were not associated with differences in illness behavior or pain ratings. With social reward held constant, patients in the high economic reward group missed more days from work (p < .005), had more domestic disability (p < .05), and were more depressed (p < .05) than patients in the low economic reward group. With economic reward held constant, patients in the high social reward group missed more days from work (p < .05), had more domestic disability (p < .01), and were more depressed (p < .01) than patients in the low social reward group. Unlike patients in the high economic reward group, however, patients in the high social reward group had higher levels of pain (p < .05) and more nonspecific medical complaints (p < .01).
CONCLUSIONS: Economic and social rewards were both associated with increased disability and depression, but only social rewards were associated with increased symptom reporting. Exposure to economic and social rewards may account for unique variance in illness behavior that cannot be explained by differences in medical diagnosis, symptom duration, pain intensity, depression, or somatization.
Key Words: economic reward • social reward • back pain • disability • illness behavior • disincentives
Abbreviations: BDI = Beck Depression Inventory; HM = Home Managementscale; SIP = Sickness Impact Profile; SR = Sleep and Restscale.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Although based on far fewer studies, a similar conclusion seems warranted in the case of social reward. In a study that included both pain patients and their spouses, Flor et al. (4) found that pain ratings were best predicted by whether the patient’s spouse was perceived as solicitous. Patients with "solicitous" spouses reported higher levels of pain than patients who had less solicitous spouses. Similarly, patients with solicitous spouses reported lower levels of activity, whereas those with "punishing" spouses had higher levels of activity. A study by Romano et al. (5) offered confirmation that spouses of pain patients tend to act in a more solicitous manner than control spouses when confronted with either an overt or covert sign of pain. According to the authors, "these findings underscore the importance of spouse solicitous behavior as both a potential reinforcer and discriminative stimulus for chronic pain behaviors" (5). Additional support for the influence of social reward was provided by Lousberg et al. (6), who found that patients observed by a solicitous spouse spent less time walking on a treadmill than patients observed by a nonsolicitous spouse.
Taken collectively, the results of these studies show that abnormal illness behavior is associated with (although not necessarily caused by) exposure to economic and/or social reward. Despite the apparent correlation between reward and behavioral outcomes, it is possible that medical rather than psychological factors are responsible. One can argue, for example, that those who sustain the worst injuries are also the most likely to be "rewarded." The correlation between reward and behavior might thus be explained by preexisting differences in illness severity. As yet, however, there is little or no support for this purely medical explanation. In their meta-analysis of the compensation effect, Rohling et al. (3) found no differences in effect size as a function of preexisting health status. Specifically, 19 of the 36 outcome studies included in their review used a matching procedure or statistically controlled for pretreatment differences in health status. The results of these studies "revealed a weighted mean effect size of 0.57, virtually the same value as the effect size of 0.60 for all 36 samples" (3). They concluded that differences in physical health or injury severity could not explain the association between compensation and treatment outcome. In a related vein, we found that physician ratings of organic pathology could not account for individual differences in illness behavior among patients at a tertiary pain service (7). The correlation between pathology ratings and number of days absent from work (the same behavioral measure used in the present study) was only 0.19.
The association between reward and illness behavior is well documented, but, to our knowledge, no study has addressed the possibility that economic and social rewards may act selectively. Social reward, for example, may influence domestic activity more than economic reward, whereas the latter may be more highly associated with work disability. These separate "effects" may be additive or even interactive in nature. If different rewards are found to be associated with different behavioral outcomes, then it may be useful to perform a more comprehensive assessment of each patient’s reward status. For example, patients exposed to high levels of both economic and social reward might have different outcomes than patients exposed to high levels of economic reward only. This issue was addressed in the present study using a rudimentary grading system to quantify both economic and social reward. These separate reward variables allowed us to conduct a 2 x 2 factorial study in which patients were retrospectively assigned to one of four independent groups: low economic/low social reward, low economic/high social reward, high economic/low social reward, and high economic/high social reward. The present study also examined the possibility that economic and social rewards do not exert a unique influence on patient behavior but rather share their influence with other, nonorganic variables, such as depression and somatization. This issue was addressed using regression analysis to control for the influence of other variables before the contribution of reward status was assessed. Finally, it is not known whether reward status exerts an all-or-none influence on patient behavior or whether increasing exposure to reward is associated with increasing levels of disability. The continuous measure of reward used in the present study allowed us to address this issue by plotting the patient’s reward status directly alongside his or her level of disability.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
Data Analysis
Once patients were assigned to groups, factorial analysis of variance was used to examine the main as well as interactive effects of economic and social reward on measures of illness behavior, symptom reporting, and affective distress (see Assessment). Aside from these group comparisons, we wished to determine whether economic and social rewards had unique effects on illness behavior or shared their influence with other variables, such as pain intensity, depression, and somatization (exaggerated symptom reporting). This issue was addressed by forcing each variable into a hierarchical regression model before entering the reward variable. This allowed us to assess the amount of unique variance (change in R2) explained by either economic or social reward. In the absence of a specific hypothesis about when or under what circumstances one type of reward might take precedence over another, we did not examine the possibility that economic reward might account for all or some of the variance explained by social reward (or vice versa). Finally, the continuous nature of each reward scale allowed us to compute pairwise correlations between level of reward and corresponding measures of work and domestic disability. Similar correlations were computed for measures of symptom reporting and affective distress.
Study Sample
From May 1995 through September 1997, a total of 265 outpatients were enrolled at the pain service of the New Jersey Medical School. Selected items of medical and demographic information obtained at the time of enrollment, along with the results of psychological testing, were entered into a clinical database. These clinical data were available for retrospective analysis and served as the basis for all comparisons reported in the present study. The protocol was approved by the institutional review board of the New Jersey Medical School. All clinic attendees were entered in the database, but only those who met the following criteria were included in the study sample: more than 19 and less than 75 years old, fluent in oral and written English, no sensory or motor impairment that interfered with reading or writing, and willingness to complete a psychological battery. To ensure that estimates of work disability were equally applicable to all participants, only patients who were employed (full or part time) or who were disabled (unable to work) at the time of enrollment were included in the sample (full-time homemakers, unemployed persons, students, and retirees were excluded). Finally, all patients with cancer, multiple (unrelated) pain disorders, or disabling illness unrelated to pain (eg, multiple sclerosis) were excluded to avoid possible confounding. The distribution of pain disorders among the remaining 144 patients was as follows: back pain, 52%; myofascial pain, 28%; neuropathy, 27%; reflex sympathetic dystrophy, 10%; orofacial pain, 8%; arthritis, 7%; and headache, 6%. The sum of these percentages exceeds 100 because a number of patients met the criteria for more than one disorder (eg, back pain with radiculopathy, headache secondary to myofascial dysfunction, etc.). To minimize the possibility of confounding due to uncontrolled differences in diagnosis, we excluded from the final sample all patients who did not have a primary diagnosis of chronic nonmalignant back pain. This reduced the total sample population to 75 patients who had one or more of the following diagnoses: lumbar radiculopathy, failed back syndrome, facet joint syndrome, sacroiliac joint pathology, spinal stenosis, back pain of myofascial origin, and other painful back disorders. The final sample consisted of 39 men and 36 women with persistent back pain for at least 6 months. Average duration of symptoms was 5.5 years. Mean age of the entire sample was 44.4 years. Demographic characteristics of the sample were as follows: married, 83%; white, 72%; African American, 20%; at least a high school education, 84%; and disabled at the time of enrollment, 80%.
Assessment
On arriving at the clinic, patients were required to produce records of previous medical treatment, undergo a medical examination, complete a battery of self-report tests (described below), and participate in a structured psychological interview. Measures of work disability (number of disability days) and frequency of medical visitation were obtained in a collaborative manner at the time of interview by comparing responses obtained on questionnaires with information available in the medical record. Patients were permitted to revise their estimates when necessary to conform with data reported by the insurance company or referring physician. Measures of economic and social reward were also subject to review and possible revision by the patient during the initial psychological interview (see below).
Work disability.
This measure was borrowed from a questionnaire developed by VonKorff et al. (8) and later modified by us to ensure that patients were basing their estimates on objective sources of disability information (7). The specific wording of the question was as follows: "Over the last 6 months (180 days), how many days have you been unable to perform at least 50% of your usual activities because of pain and/or how many days have you been unable to attend work because of pain? Give a number from 0 to 180." As noted above, homemakers, unemployed persons, students, and retirees were excluded from the sample to standardize the estimation task for all participants. We also took precautions to ensure that disability information was accurate by comparing it with information available in the medical chart. In the event of a discrepancy, patients were given an opportunity to consider all sources of relevant information and revise their estimates if necessary to more accurately reflect the actual number of days they were disabled secondary to back pain.
Medical visitation.
Before the assessment of medical visitation, the patient’s medical records were reviewed in the patient’s presence for accuracy and completeness. The records, which included provider names and dates of service, were available to patients during a frequency rating task. The task required patients to estimate the number of medical visits they made during the preceding year. The scale included six response categories: 0, "not at all in past year"; 1, "once in past year"; 2, "six times in past year"; 3, "once a month"; 4, "twice a month"; and 5, "every week." Separate ratings were elicited for each of the following providers: family physician or general practitioner, specialist or surgeon, dentist, and chiropractor or physical therapist. Ratings were then summed over all providers to yield a single estimate of visitation frequency. Patients were instructed to include all medical visits (not just those related to back pain). Patient estimates of medical visitation frequency are known to be highly correlated with objective medical records. Green et al. (9), for example, found that different types of medical visitation were accurately recalled 80% to 91% of the time.
Economic Reward scale.
The presence of contingent economic reward was first assessed by asking patients to complete a brief, self-report questionnaire. Patients were asked to indicate whether they were currently receiving or applying for 1) workman’s compensation benefits, 2) Social Security disability benefits, 3) other disability income (eg, private disability insurance), 4) a court settlement, and/or 5) financial assistance from family members or friends. Our goal was to derive a measure of economic reward that would reflect the potential influence of all these income sources. Ideally, such an index would include current sources of monetary gain (such as monthly Social Security income) as well as future sources of gain (such as a lucrative court settlement). According to social learning theory, patients learn vicariously or through first-hand experience that illness behavior can have desirable (eg, economic) consequences (10). For some, these expectations serve as powerful determinants of illness behavior. If social learning theory is correct, then economic reward should not be defined on the basis of total disability income. An anticipated but as yet unrealized legal settlement or other future economic gain may well represent a cognitive source of motivation. For this reason, we included anticipated gain in the definition by counting the number of different income sources the patient received or expected to receive if his or her disability continued. This index provides an approximate estimate of economic gain that reflects both current and future sources of reward (even though the latter could not be assigned a monetary value). The income questionnaire administered to patients at the time of their enrollment was coded with a number ranging from 0 (no sources of income) to 5 (five sources of income). One disadvantage of this strategy was that patients with a single income source were given lower reward scores than patients with multiple income sources even when the former were more lucrative than the latter. If patients with a single lucrative income source were prevalent, we would underestimate the impact of economic reward (or fail to detect its presence altogether) because all patients of this type would be assigned to the low economic reward group. To ensure accurate clinical assessment, we reviewed information available in the medical record and permitted patients to revise their responses on the income questionnaire as necessary to reflect the actual number of income sources they were receiving or anticipating. The largest number of economic benefits reported by any patient in our sample was three (although the scale theoretically permitted five). Mean levels of economic reward are presented in Table 1 for each of the four groups.
Social Reward scale.
The presence of contingent social reward was assessed by administering a brief, face-valid checklist of positive social outcomes. The outcomes used for the questionnaire were borrowed, in part, from the Solicitous Response scale of the Multidimensional Pain Inventory (11) and were consistent with categories of spouse behavior used by Romano et al. (5) in their analysis of social reward. Patients were instructed to read each of six statements and to indicate whether it was true or false. The statements were 1) "Since my pain problem started, I’ve had to avoid unpleasant social obligations" (avoidance of obligations); 2) "Since my pain problem started, my spouse or other family members have to do more of the household chores than they used to" (avoidance of domestic chores); 3) "Since my pain problem started, my spouse or other family members pay more attention (or talk to me more) than they used to" (increased positive attention); 4) "Since my pain problem started, my spouse or other family members don’t place as many demands on me as they used to" (diminished role expectations); 5) "Since my pain problem started, my spouse or other family members offer to help me more often than they used to" (solicitous behavior by others); and 6) "Since my pain problem started, I can say ‘No’ to requests by my spouse or other family members because I have to do what is best for me" (socially sanctioned self-interest). Each item was weighted equally by summing the total number of "yes" responses for each patient so that scores ranged from 0 to 6. Each item on the scale was discussed during the initial psychological interview, and, when necessary, questions were clarified to encourage more accurate responses. Among the 75 patients included in the final subject pool, the average interitem correlation (coefficient 
) for the scale was 0.65, suggesting that the six items were conceptually related but did not overlap in content. Table 1 shows the mean levels of social reward (ranging from 0 to 6) for each of the four groups.
Illness behavior.
Two additional measures of illness behavior were administered to all study participants. Both are subscales of the SIP developed by Bergner et al. (12). The SIP is a standardized measure that has well-established psychometric properties (including reliable and valid subscales) and is often used to assess disability in patients with chronic nonmalignant pain (cf. Ref. 13). Each item corresponds to a specific behavioral problem that is either endorsed or not endorsed (eg, "I am not doing any of the house cleaning that I would usually do"). Total scores for each subscale provide a measure of disability ranging from 0% to 100%. The 10-item HM scale and the 7-item SR scale were used to measure domestic disability and inactivity, respectively. Following is a representative item from the HM scale: "I am doing less of the regular daily work around the house than I would usually do." Although referred to as the "sleep and rest" scale, six of the seven items on the SR scale pertain to daytime activity or "downtime." The following are typical SR items: "I spend much of the day lying down in order to rest," and "I sleep or nap more during the day." Validity data for each of these subscales were provided by Bergner et al. (14), who correlated physician ratings of disability with corresponding SIP percent scores. In the case of first-year residents (who spent more time with patients than other raters), these correlations were 0.86 and 0.63 for the SR and HM subscales, respectively. Pollard et al. (15) reported acceptable test-retest reliabilities for the SR and HM subscales (0.69 and 0.62, respectively). Because changes in patient function were possible between test administrations, these reliability coefficients may represent conservative estimates.
Pain survey.
A brief questionnaire developed by VonKorff et al. (8) was used to assess the patient’s current, usual, and highest (worst) level of pain over the 2 weeks before enrollment. Responses were elicited on a numerical scale from 0 ("no pain") to 10 ("pain as bad as could be"). Patients were also asked to provide the date (month and year) of their back injury (ie, date of pain onset) along with an estimate of how many days they were disabled or unable to attend work during the preceding 6 months (as described above). Medical records were used, whenever possible, to confirm patient estimates.
Depression.
The 21-item BDI was administered to all patients at the time of enrollment (16). The scale is widely used as a screening instrument and has well-established psychometric properties (17). Each item on the scale consists of four statements arranged in increasing order of severity. Patients are instructed to select the one statement that "best describes the way you have been feeling over the past two weeks, including today." A number of investigators have argued that total scores on the BDI may give a misleading impression when administered to medically ill patients. Cavanaugh et al. (18), for example, showed that medical patients endorse somatic items on the BDI even if they are not depressed. To address this problem, Wesley et al. (19) recommend the use of separate Cognitive and Somatic subscales. The former eliminates vegetative and other somatic items that may be confounded with illness (items 15–21) and thus provides a more reliable estimate of depression. In accordance with this suggestion, we used the Cognitive-Affective scale (items 1–14) as well as the total BDI to assess depression in the present study.
Anxiety.
A slightly modified version of the Cognitive-Somatic Anxiety Questionnaire was administered to patients on their enrollment (20). The scale consists of seven cognitive items (eg, "I worry too much over something that doesn’t really matter") and seven somatic items (eg, "My heart beats fast"). Instructions called for patients to "Read each of the following sentences that describe feelings or symptoms people sometimes have. For each one, rate how often the symptom happened over the past two weeks, including today." Frequency ratings were elicited using a five-point scale: 0, "not at all"; 1, "rarely"; 2, "occasionally"; 3, "often"; and 4, "very often." The scale has been used previously to assess anxiety in patients with chronic nonmalignant pain (21).
Somatization.
This is a brief symptom checklist developed by Derogatis et al. (22) to assess the presence of nonorganic or exaggerated symptom reporting. Respondents are given 12 nonspecific medical symptoms and instructed to rate them on a scale of 0 to 4 to indicate "how much discomfort that problem has caused you during the past two weeks, including today." Higher scores reflect increasing discomfort arising from multiple organ systems. Items on the scale include headaches, faintness or dizziness, pains in heart or chest, pains in lower back, hot or cold spells, etc. The Somatization scale (22) has well-established psychometric properties, including test-retest reliability (0.82) and internal consistency (0.87). The dependent measures used in the present study were 1) the sum of discomfort ratings for all 12 symptoms and 2) the rating assigned to the back pain item. We previously showed that length of disability and activity level in patients with chronic nonmalignant pain are significantly associated with symptom reporting on the Somatization scale even after controlling for the influence of organic pathology (7).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
2 tests for 2 x 2 contingency tables were used to examine the distribution of demographic variables across two levels of economic reward and two levels of social reward. Neither men nor women seemed to be distributed disproportionately, with 2 = 2.27 (NS) for men and <1 for women. Similarly, there was no association between race and level of reward, with 2 < 1 for both whites and African Americans. Married patients were evenly distributed between groups (2 < 1), as were patients with at least a high school education (2 < 1). With respect to continuous variables, patients classified as receiving low economic reward had an average symptom duration of 7.9 years, compared with 3.1 years for those classified as receiving high economic reward. A 2 x 2 factorial analysis of variance confirmed that symptom duration was indeed shorter for those with higher levels of economic reward (F(1,73) = 8.0, p < .01). The low economic reward group necessarily included patients who exhausted their disability benefits and thus had pain for a longer period of time than those who did not exhaust their benefits. To examine the possibility that symptom duration may have confounded our analysis of economic and social reward, we computed bivariate correlations between duration and each dependent variable in the study. The highest correlation obtained was between duration and work disability (r = -0.21, p < .08). None of the other correlations approached statistical significance (range, 0.00 to -0.16). As a precaution, however, we included symptom duration as a covariate in each factorial analysis of variance reported below. No differences in duration were found between different levels of social reward (F < 1), and there was no social-by-economic group interaction (F < 1). A 2 x 2 factorial analysis of variance did not reveal group differences in age as a function of reward.
Preexisting differences in health status were assessed by comparing two different diagnostic groups. The first had back pain with no evidence of radiculopathy, and the second had back pain with evidence of radicular pain in one or more extremities. Of 75 patients in the total sample, 21 fell in the first group and 54 in the second. No differences in reward status were observed because patients with radicular symptoms were evenly distributed among high and low economic reward (2 < 1, NS) and high and low social reward groups (2 = 1.5, NS). The mean number of disability days (of the preceding 180) was 144 for those without radiculopathy and 134 for those with it (t(73) < 1, NS). The level of domestic disability for those with and without radiculopathy was identical (37%). Despite evidence of increased illness severity, patients with radicular involvement rated their usual pain as 8.2 on a scale of 0 to 10, compared with 8.5 for those without this complication (t(71) < 1, NS). Moreover, patients assigned to low reward groups were just as likely to be disabled at the time of enrollment as those assigned to high reward groups (2 < 1). They were also just as likely to be taking opioid medication for pain (2 < 1). Judging by these indicators of health status obtained at enrollment, there was no evidence to suggest that group comparisons were confounded by preexisting differences in illness severity.
Illness Behavior
The effects of reward on illness behavior are summarized in Table 2. Group means for each dependent measure are presented in 2 x 2 contingency tables along with corresponding analyses of variance. After controlling for differences in symptom duration and after ruling out preexisting differences (due to medical diagnosis), we found that patients exposed to multiple sources of contingent economic reward had higher levels of work disability than patients exposed to either one or no sources of economic reward (p < .005). Patients exposed to high levels of social reward were also more disabled (reported more disability days) than patients exposed to low levels of social reward (p < .05). In the absence of an interaction between economic reward and social reward (F < 1), exposure to both had additive effects on length of disability. An identical pattern of results was found in the case of domestic disability (Table 2). Impairment in household chores was measured by performance on the HM subscale of the SIP. As above, the effects of reward appeared to be additive (because there was no evidence of an interaction). Daytime inactivity or "downtime," as measured by the SR subscale of the SIP, was found to increase in the high social reward group (p < .05) but not in the high economic reward group (Table 2). Interestingly, there were no main effects of reward on frequency of medical visitation, but there was evidence of an interaction (Table 2). Patients who were not exposed to either economic or social reward (low economic/low social group) visited healthcare providers significantly less often than patients who were exposed to at least one contingent reward (p < .05). Exposure to both economic and social rewards did not further increase visitation frequency.
|
|
Affective Distress
The effects of reward on affective distress are summarized in Table 4. Mean scores on each of the dependent measures are presented in 2 x 2 contingency tables along with results of analyses of variance. Total scores on the BDI along with scores on the Cognitive-Affective subscale are presented first. Using the Cognitive-Affective subscale to control for the confounding influence of physical illness, economic and social reward were each associated with increased levels of depression. Subjects who were exposed to fewer rewards (those in the low economic/low social reward group) seemed to report lower levels of depression than those exposed to higher levels of reward, but this interaction was not confirmed by the analysis of variance.
|
Unique vs. Shared Variance
It is not known whether economic and social rewards have unique effects on illness behavior or whether their influence is shared with other nonorganic variables (such as depression and somatization). The unique variance associated with economic reward was assessed in the present study using hierarchical regression to control for symptom duration, pain intensity, depression, and somatization. The sum of economic benefits on the Economic Reward scale was used as a continuous measure of reward for this analysis (with a range of 0–3). As shown in Table 5, economic reward was found to have a significant influence on both work and domestic disability that could not be explained by any other predictor variable. An identical regression analysis is presented in Table 6 for social reward. In this case, the sum of social benefits on the Social Reward scale was used as a continuous measure of reward. After controlling for the same independent variables, the influence of social reward on work disability was eliminated. When the regression model was simplified by excluding all predictors (except symptom duration), the continuous measure of social reward achieved the same level of significance (p < .05) as the dichotomous measure achieved in the factorial analysis of variance. As shown in Table 6, social reward accounted for unique variance in domestic disability even after controlling for other sources of influence (p < .01).
|
|
|
|
Except for the relationship between social reward and work disability in Figure 2, a, each increase in reward was associated with a corresponding increase in disability. This appears to argue against a threshold model in which the effects of reward are viewed dichotomously and suggests that future research should use continuous (rather than dichotomous) measures, which may better elucidate the association between level of reward and illness behavior.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
After controlling for differences in medical diagnosis, symptom duration, pain intensity, depression, and somatization, we found that level of economic reward accounted for unique variance in both work and domestic disability, whereas level of social reward accounted for unique variance in domestic disability only. This means that measures of reward were able to account for variance in illness behavior that could not otherwise be explained. Moreover, the effects of reward appeared to be continuous rather than dichotomous in nature, with evidence of increasing disability as a function of increasing reward.
The fact that patients in the high economic reward group did not differ from those in the low economic reward group on either measure of pain intensity is consistent with findings of other studies that have relied on nonspecific pain rating scales. For example, Tait et al. (23) and Leavitt (1) found no differences in pain intensity between compensated and noncompensated patients when the pain rating task did not specify the anatomical locus of pain. Both studies found the usual differences in work-related disability, with noncompensated patients recovering more quickly than compensated patients. In contrast, Rainville et al. (2) used a more specific method of pain assessment and obtained a very different result. Rather than elicit overall pain ratings, they instructed their patients, all of whom had long-standing back injury, to rate the pain in their back. The result was that compensated patients rated their back pain as significantly more intense than noncompensated patients (p < .005). Similarly, when patients in the present high economic reward group were asked to rate the discomfort associated with their back injury, they reported significantly higher levels of discomfort than those in the low economic reward group (Table 3). In the present study as well as the one by Rainville et al. (2), patients in the compensation group based their claim of disability on back injury. In both cases, the severity of back pain was directly related to the patient’s eligibility for economic reward. This raises the possibility that pain ratings may be influenced by economic reward when the location of pain is specifically tied to the claim of disability.
Given the limitations of retrospective research and the use of reward scales that have not been rigorously validated, we believe the results of the present study should be interpreted cautiously. It remains to be seen whether these findings can be replicated on a different set of participants and whether different measures of reward would lead to similar or dissimilar results. We have described a simple numerical strategy for computing the patient’s reward status but recognize that this strategy has distinct limitations and acknowledge that a more objective measure based on actual compensation payments may have been preferable. As noted previously, the reason we adopted the index in question (number of income sources) was that it accounted for the influence of anticipated gain. The role of expected income is significant because patients who sustain a chronic back injury have good reason to believe they will receive a monetary reward. This expectation provides a cognitive source of motivation that can be just as important as a tangible asset (cf. Ref. 10). An obvious problem with our definition, however, was that we arbitrarily assigned an equivalent reward value to different sources of income. The assumption underlying this assignment was probably incorrect, because patients who received a lot of money from a single source would have obtained a lower reward score than patients who received less money from multiple sources. We have already noted that such patients would be classified as low economic reward by the present coding scheme despite their higher income status. As a result, the current index might have led us to underestimate (rather than overestimate) the effects of contingent monetary gain. The fact that we observed consistent and robust differences between the low and high economic reward groups after controlling for diagnosis and symptom duration suggests that such a bias was not a major factor in the present study. The unique variance explained by economic reward (Table 5) may also be construed as preliminary support for our numerical coding strategy. The validity of a predictor variable (such as economic reward) depends on its ability to account for "behavior that is external to the measuring instrument" (24).
In a related vein, we decided to examine the possibility that group comparisons in the present study were based on an arbitrary scoring rule rather than on real differences in reward status. To test this possibility, we developed an alternative coding strategy based on only two sources of economic reward: current disability income and a potentially lucrative legal settlement. Patients who were eligible for neither were given a score of 0, those who reported one source of reward but not the other were given a 1, and those who reported both were given a 2. The distribution of this reward variable was as follows: 0, N = 13; 1, N = 32; and 2, N = 30. A regression analysis identical to the one reported in Table 5 was performed, except that the new simplified variable was used as the index of economic reward. With work disability (disability days) as the dependent measure, the change in R2 for the new reward variable was 0.15, compared with 0.14 for the old measure. The F value associated with this change in explained variance was 12.77 for the new variable, compared with 12.23 for the old variable. This suggests that the original analysis reported in Table 5 did not depend on an arbitrary scoring rule. The results are also consistent with a prediction derived from social learning theory, which holds that anticipated gain (in the form of an expected legal settlement) may have as much influence on illness behavior as a reward contingency already in effect.
As we have already noted, the observed association between work disability and economic reward does not allow us to infer causality. There are a number of explanations for the increased disability often observed among compensated patients. First, compensated patients may sustain more serious injuries than noncompensated patients and become more disabled as a result (the so-called organic hypothesis). Second, compensated patients may consciously or unconsciously augment their symptom report to seek financial gain (the compensation hypothesis). Third, compensated patients may not be motivated by monetary gain but instead may pursue a claim of disability to receive social rewards or benefits (the social reward hypothesis). Finally, the association between compensation and disability may be a mere artifact of measurement (the artifactual hypothesis). According to this explanation, reward status is correlated with disability because disabled patients are necessarily compensated in our society and because compensated patients are necessarily disabled. This definitional redundancy gives the impression of an association, but, in fact, the association may be purely tautological. As we have already seen, the available evidence casts doubt on the organic hypothesis by showing that compensated patients are not, as a rule, more severely injured than noncompensated patients (3). This evidence, however, does not rule out the possibility that economic reward may simply be an artifact of measurement. If we assume the organic hypothesis is (at least in part) incorrect, we are left with the task of explaining the association between reward and disability in nontautological terms.
A direct test of the compensation hypothesis is virtually impossible because it would require that we alter the contingency between disability and economic compensation. However, an "experiment" of this sort was conducted in Australia in 1987, when whiplash victims were required, for the first time, to report their injuries to police and pay a small portion of their medical expenses before filing a claim (25). The number of whiplash claims declined by as much as 68% when this new rule went into effect. Although it fails to establish a causal link between financial gain and disability, this study contradicts the suggestion that such gain is simply an artifact of measurement. The possibility remains that patients who become disabled are not motivated by financial gain but rather by expectations of a less demanding or more rewarding social role. We have presented evidence in the present study that seems consistent with this social reward hypothesis. In Table 2, for example, we showed that time spent resting or reclining during the day was significantly higher among patients who received increased attention or other social benefits contingent on illness behavior. Similarly, Table 3 showed that these same patients were more likely to complain of diffuse medical symptoms. In Table 6, social reward accounted for unique variance in domestic disability (but not in work disability) after controlling for the effects of symptom duration, pain intensity, depression, and somatization. Additional support for the social reward interpretation comes from a study of menstrual illness behavior by Whitehead et al. (26), who found increased symptom reporting and disability among patients who were 1) socially rewarded for reporting menstrual symptoms during childhood and/or 2) exposed to models of exaggerated menstrual behavior. In a similar vein, childhood models of illness behavior have been shown to influence the extent of disability among headache patients (27). In summary, we cannot yet infer a causal link between reward status and illness behavior, but the evidence, taken collectively, seems to favor either a compensation or social reward hypothesis (or both) and to cast doubt on a purely organic or purely artifactual interpretation.
Received for publication April 8, 1998.
Revision received May 4, 1999.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  F. J. Keefe, M. A. Lumley, A. L. H. Buffington, J. W. Carson, J. L. Studts, C. L. Edwards, D. J. Macklem, A. K. Aspnes, L. Fox, and D. Steffey Changing Face of Pain: Evolution of Pain Research in Psychosomatic Medicine Psychosom Med, November 1, 2002; 64(6): 921 - 938. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
