Obesity Surgery

, Volume 20, Issue 4, pp 426–431

The Utility of the Beck Depression Inventory in a Bariatric Surgery Population

Authors

    • Fay W. Boozman College of Public HealthUniversity of Arkansas for Medical Sciences
  • Kelli E. Friedman
    • Duke Center for Metabolic and Weight Loss SurgeryDuke University Medical Center
  • Katherine L. Applegate
    • Duke Center for Metabolic and Weight Loss SurgeryDuke University Medical Center
Research Article

DOI: 10.1007/s11695-008-9717-2

Cite this article as:
Krukowski, R.A., Friedman, K.E. & Applegate, K.L. OBES SURG (2010) 20: 426. doi:10.1007/s11695-008-9717-2

Abstract

Background

The Beck Depression Inventory (BDI) is commonly used in bariatric surgery psychological assessments. However, several items may be measuring physical consequences of obesity (e.g., sleep disturbance, chronic pain, or sexual dysfunction) rather than depressive symptoms.

Methods

Bariatric surgery candidates (n = 210) completed a series of assessments including the BDI, a chronic pain assessment, and a semistructured clinical interview. Total BDI scores, subscale scores, and endorsement patterns of somatic versus cognitive-affective items were examined based on (1) the presence or absence of a depressive diagnosis or (2) the presence or absence of chronic pain, and optimal cut points were determined.

Results

Both the total BDI and cognitive–affective subscale had good discriminating accuracy between participants with and without depression, with an optimal cut point of 12 for the BDI and 7 for the cognitive–affective subscale. Bariatric surgery candidates with chronic pain had significantly higher mean total scores on the BDI (M = 12.5 ± 7.5) than those without chronic pain (M = 9.02 ± 6.7; p < 0.01), and those with chronic pain were significantly more likely to endorse many of the physical items than those without chronic pain.

Conclusions

The BDI, with or without the somatic items, appears to be a reasonable screening measure for depressive symptoms among bariatric surgery candidates and the subpopulation of those with chronic pain, although future investigations may wish to examine whether other measures would have improved discrimination accuracy.

Keywords

Bariatric surgeryBeck Depression InventoryDepressionObesityChronic pain

The American Society for Metabolic and Bariatric Surgery estimated that 205,000 Americans with morbid obesity (body mass index (BMI) ≥ 40) will have weight loss surgery in 2007 [1], which is approximately double the estimate in 2004 (∼125,000 surgeries [2]). Many insurance companies now require a presurgical psychological evaluation for bariatric surgery candidates. As part of these evaluations, psychologists often administer standardized psychological measures, and the Beck Depression Inventory (BDI) is the most commonly used psychological measure for evaluating depressive symptoms among obese individuals who are applying for weight loss surgery [3, 4]. Recent research has found that approximately 25% of female candidates for obesity surgery had scores on the BDI that were indicative of a mood disorder, pointing to the importance of assessing depressive symptomatology in this population [5].

Although the BDI is a widely used and accepted measure of depressive symptoms, numerous researchers [68] have questioned whether the somatic items on the BDI are appropriate for assessing depressive symptoms in medical populations. As there are several somatic items on the BDI, it may be the case that endorsement of these items in certain populations, such as cancer patients [7], are more reflective of medical symptoms than mood or affective symptoms, which may influence the validity and clinical utility of the measure in certain medical populations. Similarly, several items on the BDI may be measuring physical consequences of obesity (e.g., sleep disturbance, chronic pain, or sexual dysfunction) rather than depressive symptomatology. For example, several items measure energy level and motivation; though these symptoms may be related to depression, they are also strongly associated with obstructive sleep apnea, which is common in a bariatric surgery population [9].

Munoz and colleagues [10] recently investigated the use of the BDI in a bariatric surgery population. They compared endorsements of the cognitive–affective items on the BDI to the somatic items in a bariatric surgery population. They found that bariatric patients were more likely to endorse somatic items as compared to cognitive–affective symptoms. Munoz and colleagues suggest that it may be important for clinicians to examine endorsement patterns of individual items on the BDI, rather than the total score, when assessing a bariatric surgery population. Although this initial study suggests that an elevated score on the BDI may reflect somatic complaints rather than depression, it did not evaluate the items (somatic versus cognitive–affective) on the BDI in relation to psychological diagnosis of a depressive disorder. Further, they did not examine possible comorbidities that might also explain symptom endorsement, such as chronic pain, which could also increase endorsement of some somatic items on the BDI [11]. Obesity has been consistently linked with several types of chronic pain, including headaches [12], and low back pain, fibromyalgia, gout, and various forms of arthritis [13]. Although there is some controversy about the relationship between chronic pain conditions, like fibromyalgia, and depressive symptoms, recent research indicates that fibromyalgia and depression are separate but often comorbid disorders [14].

Thus, the primary objectives of this study were: (1) to examine the discrimination accuracy of the BDI total score and the BDI subscales (i.e., cognitive–affective and somatic); (2) to determine the optimal cut points on the BDI and the cognitive–affective subscale; and (3) to study the differential endorsement patterns (i.e., somatic versus cognitive–affective items). The study objectives will be examined: (a) based on the presence or absence of a depressive diagnosis and (b) based on the presence or absence of chronic pain.

Methods

Sample and Procedure

Individuals (n = 210) seeking bariatric surgery at a university-based surgical center offering both Roux-en-Y gastric bypass and laparoscopic banding procedures between 2003 and 2005 were included in this study. Only individuals with a BMI greater than or equal to 35 kg/m2 were scheduled for an evaluation in the surgery program. Weight loss surgery participants completed a battery of questionnaires that included the BDI [15], the West Haven–Yale Multidimensional Pain Inventory (MPI) [16], demographic information (e.g., age, race, gender), psychological treatment history, and psychotropic medication usage patterns. All participants also participated in a semistructured diagnostic clinical interview, during which a psychological diagnosis was determined by a Ph.D.-level clinical psychologist. The use of these clinical data for research purposes was approved by the Duke University Medical Center Institutional Review Board for Clinical Investigations.

Measurement

Beck Depression Inventory [15] is a 21-item self-report questionnaire assessing depressive symptomatology in the past week including today. Each item is scored on a four-point scale, ranging from 0 (absence of the symptom) to 3 (severe symptom; range 0 to 63). Scores of 0–9 are indicative of minimal depression, scores of 10–19 indicate mild depression; scores of 19–29 indicate moderate depression, and scores of 30 or greater indicate severe depression [17]. The reliability coefficient (Cronbach’s alpha) for the present study was 0.88, indicating a high internal consistency among the items. Consistent with previous research [6, 10], the BDI was divided into subscales: cognitive–affective (items 1–14) and somatic (items 15–21) subscales.

West HavenYale Multidimensional Pain Inventory [16] is a 52-item multifactorial questionnaire designed to assess a broad range of psychosocial variables related to chronic pain. Participants completed this measure only if they were experiencing chronic pain.

Data Analysis

Pearson’s correlations and t tests were calculated for associations between the BDI and demographic information. For the purposes of this study, participants were classified as having a depressive diagnosis or not during the clinical interview, based on whether they were diagnosed with any type of major depressive disorder (including major depressive episode—not otherwise specified and dysthymia). In addition, as participants only completed the MPI if they were experiencing pain, they were categorized into experiencing chronic pain or not experiencing chronic pain; neither the MPI subscales nor the individual items were examined for the purposes of this study. T tests were used to compare group means on the BDI and the BDI subscales by group (presence versus absence of a depressive diagnosis; presence versus absence of chronic pain).

Appropriateness of various BDI cut points was evaluated using receiver operating characteristic (ROC) curves. ROC analyses were used to show the change in sensitivity/1 − specificity pairs according to changes in the cutoffs for the BDI as well as cognitive–affective subscale, as this subscale omits the somatic items. In this study, sensitivity is the probability that a participant diagnosed as depressed based on the clinical interview has a score below the cut point, while specificity is the probability that a participant not diagnosed as being depressed has a score below the cut point. The area under the curve (AUC) statistic is a global measure of the overall diagnostic accuracy of the measure to determine clinical status and ranges between 1 (perfect test) and 0.5 (no better than chance). The Youden index (maximum [sensitivity + specificity − 1]) was used to assess the cutoff producing the maximized overall diagnostic effectiveness [18].

The item scores were dichotomized into “0” and “1 and above” in order to examine the presence or absence of symptoms on the BDI; chi-square tests were used to assess the endorsement of each item by those who had and those who had not been given a depressive diagnosis and those who reported and those who did not report chronic pain. An alpha level of 0.05 was established as a level of significance for all statistical tests. The data were analyzed using SPSS 16.0.

Results

Sample characteristics are presented in Table 1. Age was not significantly related to BDI total score or the cognitive–affective or somatic subscales (ps > 0.05). There were no significant differences in mean total BDI score or the cognitive–affective or somatic subscales between African American and Caucasian participants (p > 0.05). There were also no gender-related differences in BDI total score or the cognitive–affective or somatic subscales (p > 0.05).
Table 1

Means (SD) and percentages for sample characteristics

Gender (female) (%)

84.3

Age (years)

43.3 (9.4)

Body mass index (kg/m2)

48.9 (6.9)

Marital status (%)

 Never married

12.9

 Married or partnered

70.9

 Divorced, separated, or widowed

16.2

Race (%)

 Caucasian

82.4

 African American

16.2

 Other

1.4

Education (%)

 <High school degree

2.4

 High school degree or partial college

58.4

 College degree or greater

39.2

Beck Depression Inventory (total)

11.7 (7.4)

 Cognitive–affective subscale

6.5 (5.2)

 Somatic subscale

5.2 (2.8)

Beck Depression Inventory score category (%)

 Minimal (0–9)

43.3

 Mild (10–18)

41

 Moderate (19–29)

14.3

 Severe (30–63)

1.4

Clinical diagnosis of depression (%)

42.4

Current psychological–psychiatric treatment (%)

15.2

Current psychotropic medication usea (%)

43.3

Presence of pain symptoms (%)

78.1

aParticipants were primarily taking antidepressant medications (mostly selective serotonin reuptake inhibitors) and sleep medications

Diagnostic Discrimination of the BDI Total Score and Item Endorsement

Individuals who were diagnosed with depression in the clinical interview (M = 16.2 ± 7.4) had significantly higher BDI scores than those who were not diagnosed with depression (M = 8.4 ± 5.4; p < 0.001). The ROC analysis showed good discriminating accuracy of the BDI in identifying absence–presence of depression (AUC = 0.80, 95% CI 0.74, 0.87). The optimal BDI cut point for depression with the maximal Youden index was ≥12, with sensitivity of 0.73 and specificity of 0.77.

Individuals with a depressive diagnosis also had significantly higher scores on both the cognitive–affective and somatic subscales (M = 9.8 ± 5.5, 6.4 ± 2.7, respectively) compared to those who were not diagnosed with depression (M = 4.0 ± 3.4, 4.4 ± 2.6, respectively; ps < 0.001). The ROC analysis of the cognitive–affective subscale also had good discriminating accuracy (AUC = 0.82, 95% CI 0.75, 0.88). The optimal cognitive–affective subscale cut point for depression was ≥7, with sensitivity of 0.73 and specificity of 0.81.

A higher proportion of participants who were diagnosed with depression endorsed all of the cognitive–affective subscale items on the BDI than those individuals who were not diagnosed with depression (Table 2). There were no significant differences between the endorsement of participants diagnosed with depression and those not diagnosed with depression for somatic items 17–19 (tiredness, changes in appetite, and weight changes). Those without a depressive diagnosis were, nonetheless, likely (more than 50% of participants) to endorse the following BDI items: loss of pleasure, self-dislike, appearance, motivation to work, changes in sleep, tiredness, changes in appetite, worry about health, and loss of interest in sex.
Table 2

Proportion of individuals with and without a depressive diagnosis who endorsed BDI items

Item number

Item

Depressive diagnosis (N = 89)

No depressive diagnosis (N = 121)

p value

1

Sadness

56.2

14.9

<0.001

2

Pessimism

43.8

14.9

<0.001

3

Failure

42.7

14.9

<0.001

4

Loss of pleasure

82.0

62.8

<0.01

5

Guilty feelings

38.2

14.9

<0.001

6

Punishment

15.7

2.5

<0.01

7

Self-dislike

84.3

52.1

<0.001

8

Self-critical

70.8

42.1

<0.001

9

Suicidal thoughts

16.8

0.8

<0.001

10

Crying

37.1

12.4

<0.001

11

Irritation

62.9

36.4

<0.001

12

Loss of interest

51.7

24.8

<0.001

13

Indecisiveness

47.2

9.1

<0.001

14

Appearance

85.4

61.2

<0.001

15

Motivation to work

84.3

55.4

<0.001

16

Changes in sleep

67.4

52.9

<0.05

17

Tiredness

94.4

86.8

n.s.

18

Changes in appetite

45.5

54.5

n.s.

19

Weight change

1.1

2.5

n.s.

20

Worry about health

86.5

74.4

<0.05

21

Loss of interest in sex

69.7

51.2

<0.05

n.s. nonsignificant

Participants with Chronic Pain

While participants with chronic pain were no more likely to be diagnosed with depression (p > 0.05), participants who reported chronic pain had significantly higher mean total scores on the BDI (M = 12.5 ± 7.5) than those without chronic pain (M = 9.0 ± 6.7; p < 0.01). In addition, participants with chronic pain had significantly higher scores on both the cognitive–affective and somatic subscales (M = 6.9 ± 5.4, 5.6 ± 2.7, respectively) than those who did not have chronic pain (M = 5.0 ± 4.5, 4.0 ± 2.9, respectively; ps < 0.05).

When examining only those participants with chronic pain, the ROC analysis indicated good discriminating accuracy of the BDI in identifying absence–presence of depression in this subpopulation (AUC = 0.79, 95% CI 0.71, 0.86). The optimal BDI cut point for depression with the maximal Youden index was ≥13, with sensitivity of 0.70 and specificity of 0.79. The ROC analysis of the cognitive–affective subscale also had good discriminating accuracy (AUC = 0.80, 95% CI 0.73, 0.87). The optimal cognitive–affective subscale cut point for depression was ≥7, with sensitivity of 0.73 and specificity of 0.79. Participants with chronic pain also were significantly more likely to endorse several BDI items, as compared to participants who did not report chronic pain: loss of pleasure, self-dislike, appearance, motivation to work, changes in sleep, tiredness, and worry about health (Table 3).
Table 3

Proportion of individuals with and without chronic pain who endorsed BDI items

Item number

Item

Chronic pain (N = 164)

No chronic pain (N = 46)

p value

1

Sadness

34.1

26.1

n.s.

2

Pessimism

29.9

17.4

n.s.

3

Failure

28.7

19.6

n.s.

4

Loss of pleasure

76.2

52.2

<0.01

5

Guilty feelings

26.2

19.6

n.s.

6

Punishment

9.1

4.3

n.s.

7

Self-dislike

69.5

52.2

<0.05

8

Self-critical

54.9

52.2

n.s.

9

Suicidal thoughts

8.5

4.3

n.s.

10

Crying

24.4

17.4

n.s.

11

Irritation

50.0

39.1

n.s.

12

Loss of interest

39.6

23.0

n.s.

13

Indecisiveness

26.8

19.6

n.s.

14

Appearance

75.6

56.5

<0.05

15

Motivation to work

71.3

54.3

<0.05

16

Changes in sleep

64.6

39.1

<0.01

17

Tiredness

92.7

80.4

<0.05

18

Changes in appetite

17.2

10.9

n.s.

19

Weight change

1.8

2.2

n.s.

20

Worry about health

85.4

58.7

<0.001

21

Loss of interest in sex

61.0

52.2

n.s.

n.s. nonsignificant

Discussion

Both the total BDI and the cognitive–affective subscale had good discriminating accuracy between participants with and without a depressive diagnosis (based on clinical interview). The optimal total BDI cut point of ≥12 correctly identified 73% with a depressive diagnosis and 77% who did not have a depressive diagnosis, while the optimal cognitive–affective subscale cut point for depression was ≥7, which accurately classified 73% those with depression and 81% of those without a depressive diagnosis. In previous research in general chronic pain samples, researchers found similar optimal cut points on the total BDI for the greatest sensitivity and specificity in detecting depression [19, 20]. In contrast, other researchers found an optimal cut point of 21 on the total BDI and a cut point of 12 on the cognitive–affective subscale in a chronic pain sample [21]. However, consistent with Geisser and colleagues [21], using the cognitive–affective subscale, rather than the overall BDI, did not appear to significantly improve classification accuracy in this sample of bariatric surgery candidates.

Participants with depressive diagnoses were significantly more likely to endorse all of the cognitive–affective items than participants without a depressive diagnosis. Consistent with previous research, the participants in the present study were also likely to endorse many of the somatic items, even among those without a depressive diagnosis, including appearance, motivation to work, changes in sleep, tiredness, changes in appetite, worry about health, and loss of interest in sex [10]. However, unlike Munoz and colleagues [10], the participants in the present study did not have higher scores on the somatic subscale, as compared to the cognitive–affective subscale. In addition, those with a depressive diagnosis had significantly higher scores on the cognitive–affective subscale and the somatic subscale, compared to participants without a depressive diagnosis (reflecting overall higher total scores).

While participants with chronic pain were no more likely to be diagnosed with depression, they had significantly higher total scores on the BDI, as well as significantly higher scores on the cognitive–affective and somatic subscales, compared to those participants who did not endorse chronic pain. A slightly higher cut point (≥13) was optimal in this subpopulation in accurately categorizing individuals based on whether they had a depressive diagnosis or not; however, the optimal cut point on the cognitive–affective subscale was the same as in the overall sample (≥7). Nonetheless, these findings indicate that the subpopulation of bariatric surgery candidates with chronic pain may be particularly likely to obtain higher scores on the BDI, particularly on specific items (i.e., loss of pleasure, self-dislike, appearance, motivation to work, changes in sleep, tiredness, and worry about health), as was found in the present study. Based on these findings, clinicians may wish to evaluate bariatric surgery candidates for chronic pain and to be aware that the BDI total score, the subscale scores, and the item endorsements may reflect a chronic pain condition rather than a depressive diagnosis.

A possible alternative measure in this population could be the BDI-II [22], as several of the somatic items (i.e., appearance, weight change, work motivation, worry about health) are omitted in this version with the addition of items that are generally more cognitive–affective in focus (“feelings of worthlessness,” “concentration,” and “agitation”). It may be less likely that bariatric surgery candidates without depressive symptoms would endorse these additional items on the BDI-II. As the BDI-II has altered or omitted many of the items that appear to have been overendorsed in this bariatric surgery population, it may be particularly beneficial to use the BDI-II in this population. In addition, while the appetite and the sleep disturbance questions were modified on the BDI-II to include both insomnia and hypersomnia as well as increases and decreases in appetite, bariatric surgery candidates may still endorse these items without the presence of other depressive symptomatology.

The BDI for Primary Care (BDI-PC), which has been used to avoid spuriously high estimates of major depressive disorder in inpatient and outpatient medical populations [23, 24], would be another possible alternative instrument for bariatric surgery populations. Beck and his colleagues proposed that a set of seven nonsomatic items (i.e., sadness, pessimism, failure, loss of pleasure, self-dislike, self-critical, and suicidal thoughts) from the BDI-II could be used to screen medical patients in order to identify those patients who should be further evaluated for depression [23]. While the cognitive–affective subscale did not significantly improve the classification accuracy in the current study, the subset of items in the BDI-PC may be a possibility for assessing depressive symptoms among obese populations, including bariatric surgery candidates and individuals with chronic pain.

It is important to consider several limitations when interpreting the results of the current study. The methodology could have been improved by determining interrater reliability for the diagnosis of depression, based on the clinical interview. However, a review reported that interrater reliability for depressive diagnoses based on semistructured clinical interviews tends to be quite high (i.e., kappa = 0.64–0.93) [25]. While this study improved on previous research [10] by examining the comorbidity of chronic pain, a common condition in a bariatric surgery population, this study did not evaluate the presence of other chronic diseases prevalent among severely overweight individuals (e.g., sleep apnea), which may influence the endorsement of somatic items on the BDI. In addition, utilization of the MPI in this study was limited to determining the presence or absence of self-reported chronic pain for the participants; future investigations may wish to have a more comprehensive assessment of chronic pain in this population. Further, recent research has pointed to the importance of considering the possible effect of impression management in this population [26, 27] (based on participants’ concerns about eligibility for surgery), which may have attenuated the BDI scores in a portion of this sample. Finally, while there were no significant differences in BDI total scores or subscale scores on any demographic characteristics in this study, the sample was relatively homogeneous and the generalizability to other samples is unknown.

While it is unlikely that a self-report questionnaire would be used to the exclusion of a more comprehensive clinical interview in evaluating depressive symptoms in a bariatric surgery population, it is important to have a reasonable screening instrument, which does not reflect medical symptoms rather than depressive symptomatology. Based on these results, the BDI, with or without the somatic items, appears to be a reasonable screening measure for depressive symptoms among bariatric surgery candidates as well as specifically bariatric surgery candidates with chronic pain. However, clinicians should be aware that bariatric surgery candidates, particularly those with chronic pain, may be likely to have higher scores on the BDI and may endorse somatic items on the BDI, even without the presence of a depressive diagnosis. Further investigations may wish to examine whether alternative measures of depressive symptoms (the BDI-II, the BDI-PC, or various other instruments) may have better sensitivity and specificity in a bariatric surgery population with a high prevalence of chronic pain and determine which measures of depressive symptoms are most appropriate in this population.

Acknowledgements

A portion of this research was presented at the 2007 Obesity Society conference in New Orleans, LA.

This research was conducted when the first author was at the Duke Center for Metabolic and Weight Loss Surgery.

Copyright information

© Springer Science + Business Media, LLC 2008