Validation of the Brief Symptom Inventory–18 Among Low-Income African American Adolescents Exposed to Community Violence

Abstract

African American youth, especially those who reside in low resourced communities, are exposed to higher levels of exposure to community violence relative to their counterparts from other race/ethnic groups. However, appropriate measures for assessing psychological stress related to such exposures are underresearched in the extant literature for this population. The aim of the current study was to examine the reliability and validity of the Brief Symptom Inventory (BSI-18) scale among African American youth exposed to community violence through classical test theory and item response theory methods. Internal reliability and construct validity were examined. Results indicated good internal reliability (α = .93). Construct validity of the BSI-18 was established through confirmatory factor analysis with the three-factor somatic, depression, and anxiety model. Construct validity was also determined with all items indicating adequate fit. Our study indicates good reliability and validity of the BSI-18 to assess psychological distress among African American youth exposed to community violence.

Keywords

BSI-18 validation African American urban youth community violence item response theory

Introduction

Exposure to community violence has been defined as witnessing or victimization through violence that takes place outside of the home in the community (Krug, Mercy, Dahlberg, & Zwi, 2002). Studies document that between 45% and 96% of urban, African American youth have witnessed violence in their communities, ranging from assault to murder (Gaylord-Harden, Cunningham, & Zelencik, 2011; Self-Brown et al., 2006), and 16% to 37% of youth have reported being victims of aforementioned violence (Farrell & Bruce, 1997; Spano & Bolland, 2013). In the United States, being African American youth and low-income places the highest burden of such exposures on those who fall within these identities. Homicide rates reflect one dimension of exposure to community violence. National data documented that such rates among African American male youth are 3 times higher compared with Hispanics/Latino males; 4 and 22 times higher than Caucasian/White and Asian male counterparts, respectively (Centers for Disease Control and Prevention, 2010). Moreover, African American youth who reside in low-income communities reported exposure to community violence rates that are twice those reported by same race counterparts from higher resourced communities (Harrell, Langton, Berzofsky, Couzens, & Smiley-McDonald, 2014).

The overall purpose of this study is to examine the reliability and validity of the Brief Symptom Inventory–18 (BSI-18) Scale (Derogatis, 2000), while exploring the prevalence of psychological symptoms within these youth and examining potential differences in item endorsement after accounting for the average difference in psychological distress. This study is conducted among a population of low-income African Americans. To date, there are few studies that have explored the reliability and validity of the BSI-18 with this population, despite the disproportionately high burden they bear with regard to exposures to community violence and its psychological sequelae. Without question, there is increasing recognition that measures validated with White populations cannot be applied to ethnic and racial minorities with proper validity and reliability testing, given that cultural differences may influence the presentation of distress symptoms (Asner-Self, Schreiber, & Marotta, 2006).

Exposures to Community Violence and Distress Symptoms Among African American Youth

Research has established that psychological symptoms are often correlated with exposures to community violence (for review, see Margolin & Gordis, 2000). Psychological sequelae often include symptoms of posttraumatic stress disorder (PTSD), depression, anxiety, and somatic symptoms. However, findings on the type and severity of psychological symptoms correlated with exposures to community violence among African American youth are often mixed suggesting that further research is warranted in this area, especially with regard to psychological testing and measurement. For instance, among the 185 participants who were African American inner city high school youths, those that were exposed to high versus low levels of community violence, reported more fears, anxiety, internalizing behavior, and negative life experiences (Cooley-Quille, Boyd, Frantz, & Walsh, 2001). In addition, between both groups, there were no reported differences with regard to depression or externalizing behaviors. In another study of 563 African American high school adolescents exposed to community violence, major findings documented that females experienced lower levels of exposures to community violence relative to their male counterparts, but reported higher levels of all psychological distress symptoms (i.e., PTSD, anxiety, withdrawal and aggressive symptoms; Voisin & Neilands, 2010). In addition, based on a sample of 125 African American urban mothers and their children, Schiff and Mckay (2003) found that girls displayed significantly higher levels of externalizing behavioral difficulties compared with boys in relation to their exposure to community violence. These findings challenge and complicate some gender normative research which documents that females tend to report less externalizing and more internalizing symptoms in comparison with their male counterparts in response to exposures to community violence (McGee et al., 2001; Zahn-Waxler, Shirtcliff, & Marceau, 2008). In conclusion, more research which focuses on the measurement of psychological distress among African American youth is warranted.

The BSI-18

The BSI-18 (Derogatis, 2000) is an abbreviated version of the original 90-item Symptom Checklist (SCL-90; Derogatis, 1977), which was developed to assess and screen for psychological distress. An earlier abbreviation of SCL-90, the Brief Symptom Inventory, a 53-item checklist for nine dimensions (Derogatis, 1993), has also been widely used and demonstrated good psychometric properties, showing satisfactory indexes of internal consistency and test–retest reliability, and sensitivity to change (Derogatis, 1993, 1994). However, its weakness is found in the structural dimension, which has been the subject of numerous criticisms (Andreu et al., 2008). Therefore, Derogatis (2000) developed the BSI-18, which is a recent version of this series instruments and consists of only three dimensions: somatization, depression, and anxiety. Many studies apply this measure to verify the reliability and validity in different samples and cultures (Andreu et al., 2008). Although some studies argued the four-factor model of BSI-18 (i.e., anxiety is divided with panic) showed the best fit (Asner-Self et al., 2006; Wang et al., 2010), the three-factor model of BSI-18 is also well established in many studies comparing with SCL-90 and BSI-53 (Abraham, Gruber-Baldini, Harrington, & Shulman, 2017; Recklitis et al., 2006; Torres, Miller, & Moore, 2013; Wang, Kelly, Liu, Zhang, & Hao, 2013). In particular, three-factor models of BSI-18 have been identified with various samples and cultures including Caucasian/White samples (Weisner et al., 2010), Asian samples (Wang et al., 2013), and Hispanic samples (Torres et al., 2013).

The BSI-18 (Derogatis, 2000) constitutes a new and promising tool for distress assessment in clinical and community populations. Although the BSI-18 has been widely used in a number of studies focused on youth and young adults exposed to violence and trauma (Al-Krenawi, Graham, & Kanat-Maymon, 2009; Contractor et al., 2014 Lancaster, Mcrea, & Nelson, 2016), there are few studies that have examined the reliability and validity of the measure within African American youth affected by trauma. In addition, few studies to date have utilized item response theory (IRT) methods to assess item level reliability/validity and differential item functioning of the BSI-18 and have been limited to adult populations (Mieijer, de Vries, & Bruggen, 2011; Wang et al., 2010; Wang et al., 2013; Weisner et al., 2010).

The Current Study

This study addresses several gaps in the extant literature. We cannot assume that models of psychological distress for African American youth fit established models conducted with predominantly Caucasian/White study samples. In fact, Weisner and colleagues (2010) found among a national representative sample of U.S. adults that the established three-factor structure of the BSI-18 (anxiety, depression, somatic symptoms; Derogatis, 2000) adequately fit, but three items of the anxiety domain were endorsed differently comparing African American women and Caucasian women after accounting for average scale score differences. It is possible that measurement error may partly result from utilizing different measures to assess psychological symptoms; some of which have not been validated with African American samples. Measurement error might result in under- or overestimating the prevalence of psychological distress across genders which could result in missteps with regard to treatment recommendations or practices. Consequently, the aims of the current study are to (a) examine the reliability and validity of the BSI-18 psychological distress scale among African American youth who have been exposed to community trauma using both classical test theory and IRT methods; and (b) examine the prevalence of psychological distress among African American youth exposed to community violence.

Method

Sample and Procedure

Data for this study come from the Resilience Project, a study examining the risk and protective factors related to sexual risk behaviors of African American adolescents living in urban neighborhoods of concentrated poverty in Chicago. To obtain an adequate sample size from the same geographical clusters, youth were recruited from three high schools, one youth church group, two community youth programs, and four public venues. An overall response rate of 87% was achieved. These participants were recruited from low-income communities consisting predominantly of African American residents, where the average annual median income ranged from US$24,049 to US$35,946, below the Chicago city average of US$43,628. The percentage of single-mother households in these areas ranged from 28.9% to 32.3%, with the city average being 13.9% (City-Data, 2015).

To recruit adolescent participants, flyers with information regarding the study were posted at schools, community programs, and churches, where the school principals as well as leaders of church groups and youth programs had given permission to recruit participants for the study. Each participant was required to have both active parental consent (if younger than 18 years) and youth assent to participate in the study. Trained research assistants introduced the study to all potential participants, recruited from aforementioned locations with a detailed letter describing the study along with parental consent forms. Youth who returned consent forms signed by a parent or guardian and provided assent were enrolled in the study. Youth recruited in public venues were only asked to participate if both a parent was present to provide consent as well as youth providing assent. Participants recruited from schools, community programs, and churches were administered a questionnaire at those respective locations. Individuals who were recruited in public venues (e.g., parks and fast food venues) were given questionnaires in quiet spaces at or near those venues. The questionnaire took approximately 45 min to complete, after which, the youth participant was given a US$10 cash compensation. The University Institutional Review Board approved the study.

Measurement Instruments

Community trauma was assessed using the Lifetime Prevalence of Trauma Exposure Probe (Stein, Walker, Hazen, & Forde, 1997). Six items inquired about witnessing violent acts (close relative or friend died violently, close relative or friend seriously injured, close relative or friend robbed or attacked, seen someone being beaten, seen a dead body in the community, and witnessed gun-related incident), and one item inquired about being a victim of violence. Items were rated on a 7-point scale (0 times to more than 6 times). For the current study, we dichotomized the items (i.e., “0” did not experience and “1” experienced 1 or more times). The Cronbach’s alpha was .87 for the current study, which was within the .71 to .87 range noted in prior studies (So, Gaylord-Harden, Voisin, & Scott, 2018; Voisin, Neilands, & Hunnicutt, 2011).

The BSI-18 (Derogatis, 2000) was used to assess psychological distress for the current study. For each symptom, respondents were asked to indicate how much the symptoms bothered them in the past 7 days. Examples of questions were, “in the past seven days including today have you felt lonely?” and “in the past seven days including today have you been suddenly scared for no reason.” Responses were rated on a 5-point Likert-type scale (0 = not at all, 1 = a little bit, 2 = moderate, 3 = quite a bit, and 4 = extremely). The BSI-18 constitutes three domains (i.e., anxiety, depression, and somatic symptoms), with six questions for each domain. The global severity index of distress indicates the sum of the three domains ranging from 0 to 72 with higher scores indicating more severe levels of psychological distress (Derogatis, 2000). Based on previous studies, we based a cutoff total score of 11 as an indication of significant psychological distress (De Beurs, 2011; Merckelbach, Langeland, de Vries, & Draijer, 2014).

Data Analysis Steps

Multiple statistical software packages were utilized for the current study, including STAT (StataCorp, 2013), Mplus v7.3 (Muthén & Muthén, 2014), and Acer Conquest 3 (Adams, Wu, & Wilson, 2012). The analyses included the following steps: (a) exploratory analyses to provide descriptive statistics, (b) confirmatory factor analysis (CFA; based on the common three-factor BSI model established in the literature, Derogatis, 2000) to examine the dimensionality and underlying factor structure of the data, and (c) Rasch Rating Scale IRT modeling to examine the performance of each item and difficulty range of items along the latent construct of the BSI-18.

Exploratory Data Analysis

Exploratory data analyses examined basic demographics of the sample and the distribution of symptom responses. Average total scale scores for the global severity index and each of the domains were also generated for the sample. Cronbach’s alpha estimate (α) was generated to assess internal consistency reliability. To assess the difference in total scale score across gender, we conducted an independent samples t test.

CFA

To conduct IRT analyses, which has the assumption of unidimensionality (Embretson & Reise, 2013), we first tested the factor structure of the underlying trait of the BSI-18 using CFA. If the best fitting model is not unidimensional, a multidimensional model should be used to account for separate factors of the underlying psychological distress trait (Embretson & Reise, 2013). As such, we tested a unidimensional model as well as higher order model and bifactor model. Due to the severe floor effect in the response, we treated the items as ordinal indicators and used the weighed least squares mean and variance adjusted estimation method (WLSMV; Brown, 2006). We used the full information maximum likelihood estimator to deal with missing data (Muthén & Muthén, 2014). Absolute fit of the CFA was examined using global fit indices, including the comparative fit index (CFI), the Tucker–Lewis index (TLI), the root mean square error of approximation (RMSEA), and the Weighted Root Mean Square Residual (WRMR). RMSEA values lower than .06, WRMR values close to 1, and TLI/CFI values above .95 are indicative of good model fit (Hu & Bentler, 1998).

IRT Model

To assess construct validity and item fit of the BSI-18, a Multidimensional two parameter Rasch Rating Scale IRT model was used (De Ayala, 2013). The underlying structure of psychological distress of the BSI-18 was determined through item location (i.e., difficulty parameters) from the IRT analyses. Difficulty parameter estimates are expressed in logits. Logits are the natural logarithm of the odds of a participant endorsing a symptom (or item), which is equivalent to the ratio of the probability of not endorsing the item to the probability of endorsing the item (De Ayala, 2013; Duncan, Bode, Lai, Perera, & Glycine Antagonist in Neuroprotection Americas Investigators, 2003). For example, an item with a difficulty parameter of 0.40 logits is equivalent to a participant having 1.49 higher odds of endorsing that item compared with other items in the scale. The hierarchy of difficulty parameter estimates are also indicative of the underlying structure of the latent construct (or construct validity) of psychological distress, so that items with higher logits are items that are endorsed among participants with more severe levels of psychological distress (De Ayala, 2013). For example, how a respondent will answer the item feeling faintness is based on the severity of the person’s psychological distress and where the item feeling faintness is on the continuum (or range of symptoms from least difficult to most difficult to endorse) of the underlying latent psychological distress construct.

Item fit is determined by the mean square error (MNSQ) infit and outfit statistics (see Table 5). The infit and outfit statistics adopt slightly different techniques for assessing an item’s fit to the Rasch model. The infit statistic gives relatively more weight to the performances of persons closer to the item value (Bond & Fox, 2015). The argument is that persons whose ability is close to the item’s difficulty should give a more sensitive insight into that item’s performance. The outfit statistic is not weighted, and therefore is more sensitive to the influence of outlying scores. It is for this reason that users of the Rasch model routinely pay more attention to infit scores than to outfit scores (Bond & Fox, 2015). For this study, infit statistics were predominantly used to assess items.

The standard range of infit statistics varies according to disciplines and sciences (Linacre & Wright, 1994). In the behavioral sciences, reasonable MNSQ infit statistics for rating scale models are regarded as above 0.5 and below 1.5 (Linacre, 2002; Wright, Linacre, Gustafson, & Martin-Lof, 1994). Outside of the acceptable range can be detrimental to measurement development and accurate assessment. Below 0.5 suggests too little variation in response patterns (or an item with no error), and above 1.5 suggests too much variation (or high amounts of error; Linacre, 2002; Wright et al., 1994). As a multidimensional two parameter Rasch Rating Scale model was used, the correlation between the dimensions was also modeled for each analysis.

Results

Exploratory Data Analysis

The total combined data included N = 638 participants. The majority of participants were girls (54.21%). Participants were between the ages of 13 and 22 years (M = 15.8 years old, SD = 1.4). Almost two thirds (60.5%) of the overall sample qualified for free or reduced school lunch, indicating low-income households. With regard to sexual orientation, the majority of adolescents (81.20%) self-identified as heterosexual.

Findings indicated high rates of exposure to community violence, namely, 66.1% of the sample had a close relative or friend die violently and 71.2% of the sample witnessed a relative or friend robbed or attacked. In addition, 54.5% witnessed a relative or friend being beaten, 38.61% had been a victim of violence, 37.3% had seen a dead body not at a funeral, and 44.3% had witnessed a gun-related incident.

For the BSI-18, nine participants were missing all items (less than 2%) and were removed from all subsequent analyses. Internal consistency reliability for items combined was good (α = .93). Internal consistency reliability for each domain was also good with α = .82 for somatic, α = .83 for depression, and α = .82 for anxiety. The distribution of responses to the BSI-18 symptom questions show that just over half of respondents indicated not at all “0” experiencing the symptoms (Table 1). However, the total scale score average (or general severity index) was 11.88 (SD = 12.42), slightly above the established cutoff for psychological distress (Table 2). The total average scale score for each of the dimensions was as follows: for somatic symptoms, it was 4.17 (SD = 4.56); for depression symptoms, it was 4.2 (SD = 4.74); and for anxiety, it was 3.87 (SD = 4.58). Results revealed a significant difference in total average scale score between girls and boys, with a mean BSI score of girls (M = 13.38, SD = 12.85) 3.36 points higher than boys (M = 10.02, SD = 11.63, t(570) = −3.36, p < .05; 95% confidence interval [CI] = [–5.39, –1.32]).

Table 1.

Frequency (%) for Each BSI Item (N = 629).

Items^a	0 (Not at all)	1 (A Little Bit)	2 (Moderate)	3 (Quite a Bit)	4 (Extremely)
Somatization
1. Faintness	366 (58.65)	155 (24.84)	55 (8.81)	42 (6.73)	6 (0.96)
2. Heart pain	374 (59.84)	139 (22.24)	60 (9.60)	38 (6.08)	14 (2.24)
3. Nausea/upset stomach	336 (53.50)	136 (21.66)	80 (12.74)	50 (7.96)	26 (4.14)
4. Trouble breathing	431 (68.96)	79 (12.64)	55 (8.80)	40 (6.40)	20 (3.20)
5. Numbness	412 (66.13)	112 (17.98)	57 (9.15)	30 (4.82)	12 (1.93)
6. Feeling weak	354 (56.64)	131 (20.96)	71 (11.36)	51 (8.16)	18 (2.88)
Depression
7. Feeling no interest	219 (35.27)	175 (28.18)	121 (19.48)	83 (13.20)	24 (3.86)
8. Feeling lonely	344 (54.95)	123 (19.65)	67 (10.70)	52 (8.31)	40 (6.39)
9. Feeling blue	416 (66.35)	100 (15.95)	50 (7.97)	44 (7.02)	17 (2.71)
10. Feeling worthless	459 (73.32)	78 (12.46)	45 (7.19)	25 (3.99)	19 (3.04)
11. Feeling hopeless	439 (70.35)	86 (13.78)	50 (8.01)	33 (5.29)	16 (2.56)
12. Thoughts of ending one’s life	502 (80.32)	59 (9.44)	22 (3.52)	28 (4.48)	14 (2.24)
Anxiety
13. Nervousness/shake	319 (51.53)	142 (22.94)	86 (13.89)	51 (8.24)	21 (3.39)
14. Feeling tense	362 (58.11)	110 (17.66)	77 (12.36)	52 (8.35)	22 (3.53)
15. Suddenly scared	449 (71.61)	72 (11.48)	50 (7.97)	39 (6.22)	17 (2.71)
16. Terror/panic spell	516 (82.69)	51 (8.17)	28 (4.49)	21 (3.37)	8 (1.28)
17. Feeling restless	396 (63.36)	104 (16.64)	64 (10.24)	40 (6.40)	21 (3.36)
18. Feeling fearful	410 (64.92)	105 (16.88)	42 (6.75)	39 (6.27)	26 (4.18)

Note. BSI = Brief Symptom Inventory.

Less than 2% missing on all items.

Table 2.

Mean and Standard Deviation of BSI-18 Items (N = 629).

Symptom	M	SD
Somatization	4.17	4.56
1. Faintness	0.65	0.96
2. Heart pain	0.68	1.02
3. Nausea/upset stomach	0.88	1.16
4. Trouble breathing	0.62	1.09
5. Numbness	0.58	0.97
6. Feeling weak	0.80	1.11
Depression	4.20	4.74
7. Feeling no interest	1.22	1.17
8. Feeling lonely	0.92	1.25
9. Feeling blue	0.64	1.07
10. Feeling worthless	0.51	1.00
11. Feeling hopeless	0.56	1.01
12. Thoughts of ending one’s life	0.39	0.92
Anxiety	3.86	4.58
13. Nervousness/shake	0.89	1.13
14. Feeling tense	0.82	1.15
15. Suddenly scared	0.57	1.05
16. Terror/panic spell	0.32	0.82
17. Feeling restless	0.70	1.10
18. Feeling fearful	0.66	1.11
General Severity Index	11.88	12.42

Note. BSI = Brief Symptom Inventory.

Bold indicates overall mean and standard deviation.

CFA

We have tested four measurement models. The fit indices are reported in Table 3. The second-order three-factor model had a closer fit to the data compared with the competing models: χ²(132) = 444.027, p = .000, CFI = .974, TLI = .970, RMSEA = .06, and WRMR = 1.121.

Table 3.

Confirmatory Factor Analysis of BSI-18 (N = 629).

Mode	χ²	df	CFI	TLI	RMSEA	WRMR
Unidimensional model	500.291	135	.898	.885	.120	2.235
Second-order three-factor model	444.027	132	.974	.970	.061	1.121
Bifactor model	1,115.540	118	.917	.893	.116	1.926

Note. BSI = Brief Symptom Inventory; CFI = Comparative Fit Index, good fit >.95; TLI = Tucker–Lewis Index; RMSEA = root mean square error of approximation, good fit < .06; SRMSR = standardized root mean square residual, good fit < .08; WRMR = weighted root mean square residual (Hu & Bentler, 1998).

Factor loadings are presented in Table 4. In the second-order three-factor model, all items loaded significantly on their specific factors. Correlations between specific factors are also presented in Table 4. The range of correlation between specific factors is between .242 and .348.

Table 4.

Factor Loadings of the Second-Order Three-Factor of BSI-18 (N = 629).

Symptom	Estimate(b)	SE	Estimate/SE
Somatization
1. Faintness	1.000
2. Heart pain	1.264	0.108	11.684***
3. Nausea/upset stomach	1.564	0.129	12.125***
4. Trouble breathing	1.421	0.120	11.859***
5. Numbness	1.242	0.108	11.486***
6. Feeling weak	1.669	0.132	12.664***
Depression
7. Feeling no interest	1.000
8. Feeling lonely	1.540	0.133	11.597***
9. Feeling blue	1.333	0.114	11.682***
10. Feeling worthless	1.383	0.115	12.035***
11. Feeling hopeless	1.302	0.112	11.638***
12. Thoughts of ending one’s life	1.034	0.095	10.928***
Anxiety
13. Nervousness/shake	1.000
14. Feeling tense	1.022	0.076	13.504***
15. Suddenly scared	1.077	0.071	15.254***
16. Terror/panic spell	0.751	0.054	13.912***
17. Feeling restless	0.998	0.072	13.841***
18. Feeling fearful	1.111	0.075	14.735***
Correlation between SOM and DEP	0.242	0.029	8.206***
Correlation between SOM and ANX	0.348	0.037	9.405***
Correlation between DEP and ANX	0.326	0.044	9.122***

Note. BSI = Brief Symptom Inventory; SOM = somatization; DEP = depression; ANX = anxiety.

***

p <.001.

Multidimensional Rating Scale IRT Models

Based on the CFA results, we used the three-factor model for all Multidimensional Rasch Rating Scale IRT analyses. In the sample, all items had good fit (MNSQ between 0.5 and 1.50; Table 5). Difficulty parameters suggested a wide range of symptoms along the psychological distress continuum (or range) with the item of feeling no interest with the lowest difficulty (b = −0.63) and terror/panic spell with the highest difficulty (b = 0.58; Table 6). In other words, participants with low levels of psychological distress were likely to endorse the presence of feeling no interest, while only participants with severe levels of psychological distress were likely to endorse the presence of terror/panic spell. Along the latent continuum, items from each of the dimensions were adequately represented in each level of severity with little redundancy (see Table 6).

Table 5.

Multidimensional Rating Scale IRT Model for Sample Along BSI Continuum (N = 629).

Item	Difficulty Parameter (b)	Infit MNSQ^a	Outfit MNSQ
1. Faintness	0.050	1.35	1.33
2. Heart pain	0.021	0.80	1.60
3. Nausea/upset stomach	−0.206	0.90	0.77
4. Trouble breathing	0.099	1.43	1.62
5. Numbness	0.153	1.02	0.79
6. Feeling weak	−0.115	0.80	0.84
7. Feeling no interest	−0.634	1.37	1.84
8. Feeling lonely	−0.307	1.11	1.30
9. Feeling blue	0.046	0.72	0.56
10. Feeling worthless	0.250	0.71	0.49
11. Feeling hopeless	0.168	0.96	0.69
12. Thoughts of ending one’s life	0.477	0.99	0.53
13. Nervousness/shake	−0.322	1.10	1.21
14. Feeling tense	−0.239	0.95	0.73
15. Suddenly scared	0.101	0.92	0.64
16. Terror/panic spell	0.581	1.21	0.48
17. Feeling restless	−0.078	0.85	0.72
18. Feeling fearful	−0.033	1.05	0.97

Note. IRT = item response theory; BSI = Brief Symptom Inventory; MNSQ = mean square error.

MNSQ infit statistic acceptable range above 0.50 and below 1.50.

Table 6.

BSI-18 Items Along the Latent Continuum From Least Difficulty to Most Difficulty.

Item	Difficulty Parameter
Somatization
Faintness	0.050
Heart pain	0.021
Nausea/upset stomach	−0.206
Trouble breathing	0.099
Numbness	0.153
Feeling weak	−0.115
Depression
Feeling no interest	−0.634
Feeling lonely	−0.307
Feeling blue	0.046
Feeling worthless	0.250
Feeling hopeless	0.168
Thoughts of ending one’s life	0.477
Anxiety
Nervousness/shake	−0.322
Feeling tense	−0.239
Suddenly scared	0.101
Terror/panic spell	0.581
Feeling restless	−0.078
Feeling fearful	−0.033

Note. BSI = Brief Symptom Inventory.

Discussion

The primary aim of this study was to examine the reliability and validity of the BSI-18 scale (Derogatis, 2000) classifying distress symptoms and assessing for item endorsements among a sample of low-income African American youth. Our study significantly contributes to our understanding of the utility of the BSI-18 among this vulnerable population through both classical test theory and IRT methods. This study was based on a purposive sample given that attempts were made to recruit youth from low-income neighborhoods within the same geographical clusters that were affected by high levels of exposures to community violence. As such, this is one of the few studies to examine the reliability and validity at the item level of the widely used BSI measure for psychological distress among a youth sample of low-income African American youth exposed to high levels of community violence.

Overall findings indicated that levels of exposures to community violence were relatively high among the overall sample. Although the distribution of responses to the BSI-18 symptom questions indicated the majority of respondents endorsed not at all “0” experiencing the symptoms, the total scale score average (or general severity index) was 11.88 (SD = 12.42), slightly above the established cutoff for psychological distress. This is consistent with findings among other trauma-affected populations (Asner-Self et al., 2006; Contractor et al., 2014; Wang et al., 2010) that have found average scale scores ranging from 12.08 among Indian youth following the Mumbai terrorist attacks (Contractor et al., 2014) to 19.28 among drug-using populations (Wang et al., 2010). The high level of psychological distress among our sample indicates the urgent need for treatment addressing psychosocial problems related to community violence.

Our study was also consistent with previous literature, which supports the three-factor structure of the BSI-18 (Derogatis, 2000; Wang et al., 2010; Weisner et al., 2010), suggesting the utility of the three factors among African American youth who have been exposed to community violence. In addition, results from the IRT analyses indicated adequate fit (little measurement error) across all items demonstrating reliability of the measure among this population. Furthermore, the range of items with little redundancy across the three domains along the latent construct of psychological distress provides supporting evidence that the BSI-18 is an appropriate measure for capturing and assessing a broad range of severity, differentiating participants with varying levels of psychological distress among low-income African American youth exposed to community violence. We found that items with low- and midrange difficulty were representative of all three domains, which is promising for the BSI-18 as a screening tool among this population. Items were primarily from the depression and anxiety domain in the high difficulty range (e.g., feeling worthless, hopeless, terror/panic spell, and thoughts of ending one’s life), suggesting that symptom types from the depression and anxiety domain are more likely endorsed among African American youth with more severe levels of psychological distress. Although suicidal thoughts would indicate severe distress in all circumstances, these findings are especially useful in that the more severe symptoms of depression and anxiety that may not typically indicate acute severe psychological impairment (i.e., worthlessness, hopelessness, and a sense of panic) can be used for assessing youth who may be more at risk and in need of immediate or crisis treatment referral.

Notwithstanding the contributions of this study, there are several limitations that warrant mention. Overall findings may have limited generalizability for other youth populations and those residing outside of similar geographical and economic clusters. Given that some of the previous literature suggested gender differences in psychological distress severity among African American youth exposed to community violence (Schiff & McKay, 2003; Voisin & Neilands, 2010), future studies with a special focus on gender should evaluate measurement invariance assessment of the BSI-18 across gender using multigroup CFA models among African American youth. In addition, there are other factors that might influence measurement findings of the BSI-18 among African American youth. For instance, prior findings among African American youth have documented that mothers’ parenting practices and family relationships were associated with youths’ externalizing behavior problems (Schiff & McKay, 2003). We did not explore other variables that could be responsible for lack of observed Differntial Item Funtioning (DIF), such as age, sexual orientation, household composition and stress, or other unmeasured variables. Future studies should evaluate DIF related to these and other potential sources of response bias. Notwithstanding these limitations, overall findings suggest a good reliability and validity of the BSI-18 in a sample of low-income African American youth, indicating that it may be valuable to assess psychological distress among African American youth exposed to community violence. In addition, the findings of this study serve to broaden our knowledge with respect to the factorial structure of the BSI-18. The valid factorial structure of the BSI-18 enables one to model Structional Equation Model (SEM) using low-income African American samples without the influence of measurement errors in the observed items.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the STI/HIV Intervention Network and Center for Health Administration Studies at the University of Chicago. This manuscript was also made possible with help from the Third Coast Center for AIDS Research (P30 AI 117943).

Author Biographies

Dong Ha Kim is an assistant professor at Chungwoon University. A major focus of her scholarship is on risk and protective factors among adolescents with a special focus research methodology.

Lynn Murphy Michalopoulos is an associate professor at Columbia University. A major focus of her scholarship is on the effects of trauma among national and international populations especially those who are at disproportionate risk for exposures to interpersonal and community trauma.

Dexter R. Voisin is a professor at the University of Chicago. A major focus of his research is on the effects of exposures to community violence on racial and sexual minority youth and the factors that protect adolescents from behavioral health-related risk factors in the face of such exposures.

References

Abraham

D. S.

Gruber-Baldini

A. L.

Harrington

Shulman

L. M.

(2017). The factor structure of the Brief Symptom Inventory-18 (BSI-18) in Parkinson disease patients. Journal of Psychosomatic Research, 96, 21-26.

Adams

R. J.

M. L.

Wilson

M. R.

(2012). ACER ConQuest 3.0.1. Camberwell: Australian Council for Educational Research.

Al-Krenawi

Graham

J. R.

Kanat-Maymon

(2009). Analysis of trauma exposure, symptomatology and functioning in Jewish Israeli and Palestinian adolescents. The British Journal of Psychiatry, 195, 427-432. doi:10.1192/bjp.bp.108.050393

Andreu

Galdón

M. J.

Dura

Ferrando

Murgui

García

Ibáñez

(2008). Psychometric properties of the Brief Symptoms Inventory-18 (BSI-18) in a Spanish sample of outpatients with psychiatric disorders. Psicothema, 20, 844-850.

Asner-Self

K. K.

Schreiber

J. B.

Marotta

S. A.

(2006). A cross-cultural analysis of the Brief Symptom Inventory-18. Cultural Diversity & Ethnic Minority Psychology, 12, 367-375. doi:10.1037/1099-9809.12.2.367

Bond

Fox

C. M.

(2015). Applying the Rasch model: Fundamental measurement in the human sciences. Mahwaw, NJ: Lawrence Erlbaum Associates.

Brown

T. A.

(2006). Confirmatory factor analysis for applied research. New York, NY: The Guildford Press.

Centers for Disease Control and Prevention. (2010). Web-based Injury Statistics Query and Reporting System (WISQARS). National Center for Injury Prevention and Control, Centers for Disease Control and Prevention. Retrieved from www.cdc.gov/injury/wisqars/fatal.html

City-Data. (n.d.). bbb.org . Retrieved from http://www.city-data.com (accessed 11 December 2015)

10.

Contractor

A. A.

Mehta

Tiamiyu

M. F.

Hovey

J. D.

Geers

A. L.

Charak

. . . Elhai

J. D.

(2014). Relations between PTSD and distress dimensions in an Indian child/adolescent sample following the 2008 Mumbai terrorist attacks. Journal of Abnormal Child Psychology, 42, 925-935. doi:10.1007/s10802-013-9846-5

11.

Cooley-Quille

Boyd

R. C.

Frantz

Walsh

(2001). Emotional and behavioral impact of exposure to community violence in inner-city adolescents. Journal of Clinical Child & Adolescent Psychology, 30, 199-206. doi:10.1207/S15374424JCCP3002_7

12.

De Ayala

R. J

. (2013). The IRT tradition and its applications. In Little

T. D.

(Ed.), The Oxford handbook of quantitative methods: Foundations (Vol. 1, pp. 144-169). Oxford, UK: Oxford University Press.

13.

De Beurs

. (2011). Brief Symptom Inventory (BSI) en BSI 18. Handleiding 2011. Leiden, The Netherlands: Psychologische Instrumenten, Tests en Services (PITS).

14.

Derogatis

L. R.

(1977). The SCL-90 Manual I: Scoring, administration and procedures for the SCL-90. Baltimore, MD: Johns Hopkins University School of Medicine, Clinical Psychometrics Unit.

15.

Derogatis

L. R.

(1993). Brief Symptoms Inventory (BSI): Administration, scoring and procedures manual (3rd ed.). Minneapolis, MN: NCS Pearson.

16.

Derogatis

L. R.

(1994). Symptom Checklist-90-R (SCL-90-R): Administration, scoring and procedures manual (3rd ed.). Minneapolis, MN: NCS Pearson.

17.

Derogatis

L. R.

(2000). Brief Symptom Inventory-18. Minneapolis, MN: NCS Pearson.

18.

Duncan

P. W.

Bode

R. K.

Lai

S. M.

Perera

, & Glycine Antagonist in Neuroprotection Americas Investigators. (2003). Rasch analysis of a new stroke-specific outcome scale: The Stroke Impact Scale. Archives of Physical Medicine and Rehabilitation, 84, 950-963. doi:10.1016/S0003-9993(03)00035-2

19.

Embretson

S. E.

Reise

S. P.

(2013). Item response theory. Hillsdale, NJ: Erlbaum.

20.

Farrell

A. D.

Bruce

S. E.

(1997). Impact of exposure to community violence on violent behavior and emotional distress among urban adolescents. Journal of Clinical Child Psychology, 26, 2-14. doi:10.1207/s15374424jccp2601_1

21.

Gaylord-Harden

N. K.

Cunningham

J. A.

Zelencik

(2011). Effects of exposure to community violence on internalizing symptoms: Does desensitization to violence occur in African American youth? Journal of Abnormal Child Psychology, 39, 711-719. doi:10.1007/s10802-011-9510-x

22.

Harrell

Langton

Berzofsky

Couzens

Smiley-McDonald

(2014). Household poverty and nonfatal violent victimization, 2008–2012. Washington, DC: Bureau of Justice Statistics.

23.

L. T.

Bentler

P. M.

(1998). Fit indices in covariance structure modeling: Sensitivity to underparameterized model misspecification. Psychological Methods, 3, 424-453.

24.

Krug

E. G.

Mercy

J. A.

Dahlberg

L. L.

Zwi

A. B.

(2002). The world report on violence and health. The Lancet, 360, 1083-1088. Available from www.thelancet.com

25.

Lancaster

M. A.

McCrea

M. A.

Nelson

L. D.

(2016). Psychometric properties and normative data for the Brief Symptom Inventory-18 (BSI-18) in high school and collegiate athletes. The Clinical Neuropsychologist, 30, 321-333. doi:10.1080/13854046.2016.1138504

26.

Linacre

J. M.

(2002). What do infit and outfit, mean-square and standardized mean. Rasch Measurement Transactions, 16(2), 878.

27.

Linacre

J. M.

Wright

B. D.

(1994). Chi-square fit statistics. Rasch Measurement Transactions, 8(2), 350.

28.

Margolin

Gordis

E. B.

(2000). The effects of family and community violence on children. Annual Review of Psychology, 51, 445-479. doi:10.1146/annurev.psych.51.1.445

29.

McGee

Z. T.

Davis

B. L.

Brisbane

Collins

Irving

Nuriddin

. . . Martin

(2001). Urban stress and mental health among African-American youth: Assessing the link between exposure to violence, problem behavior, and coping strategies. Journal of Cultural Diversity, 8(3), 94-104.

30.

Meijer

R. R.

de Vries

R. M.

van Bruggen

(2011). An evaluation of the Brief Symptom Inventory–18 using item response theory: Which items are most strongly related to psychological distress? Psychological Assessment, 23(1), 193-202.

31.

Merckelbach

Langeland

de Vries

Draijer

(2014). Symptom over reporting obscures the dose-response relationship between trauma severity and symptoms. Psychiatry Research, 217, 215-219. doi:10.1016/j.psychres.2014.03.018

32.

Muthén

L. K.

Muthén

B. O.

(2014). Mplus 7.2. Los Angeles, CA: Author.

33.

Recklitis

C. J.

Parsons

S. K.

Shih

M. C.

Mertens

Robison

L. L.

Zeltzer

(2006). Factor structure of the Brief Symptom Inventory—18 in adult survivors of childhood cancer: Results from the childhood cancer survivor study. Psychological Assessment, 18, 22-32.

34.

Schiff

McKay

M. M.

(2003). Urban youth disruptive behavioral difficulties: Exploring association with parenting and gender. Family Process, 42, 517-529. doi:10.1111/j.1545-5300.2003.00517.x

35.

Self-Brown

S. R.

LeBlanc

Kelley

M. L.

Hanson

Laslie

Wingate

(2006). Effects of community violence exposure and parental mental health on the internalizing problems of urban adolescents. Violence and Victims, 21, 183-198. doi:10.1891/vivi.21.2.183

36.

Gaylord-Harden

N. K.

Voisin

D. R.

Scott

(2018). Future orientation as a protective factor for African American adolescents exposed to community violence. Youth & Society, 50(6), 734-757.

37.

Spano

Bolland

(2013). Disentangling the effects of violent victimization and violent behavior on gun carrying for minority inner city youth living in extreme poverty. Crime & Delinquency, 59, 191-213. doi:10.1177/0011128710372196

38.

StataCorp. (2013). Stata Statistical Software: Release 13. College Station, TX: Author.

39.

Stein

M. B.

Walker

J. R.

Hazen

A. L.

Forde

D. R.

(1997). Full and partial posttraumatic stress disorder: Findings from a community survey. American Journal of Psychiatry, 154, 1114-1119. doi:10.1176/ajp.154.8.1114

40.

Torres

Miller

M. J.

Moore

K. M.

(2013). Factorial invariance of the Brief Symptom Inventory-18 (BSI-18) for adults of Mexican descent across nativity status, language format, and gender. Psychological Assessment, 25, 300-305.

41.

Voisin

D. R.

Neilands

T. B.

(2010). Community violence and health risk factors among adolescents on Chicago’s southside: Does gender matter? Journal of Adolescent Health, 46, 600-602. doi:10.1016/j.jadohealth.2009.11.213

42.

Voisin

D. R.

Neilands

T. B.

Hunnicutt

(2011). Mechanisms linking violence exposure and school engagement among African American adolescents: Examining the roles of psychological problem behaviors and gender. American Journal of Orthopsychiatry, 81, 61-71. doi:10.1111/j.1939-0025.2010.01072.x

43.

Wang

Kelly

B. C.

Booth

B. M.

Falck

R. S.

Leukefeld

Carlson

R. G.

(2010). Examining factorial structure and measurement invariance of the Brief Symptom Inventory (BSI)-18 among drug users. Addictive Behaviors, 35, 23-29. doi:10.1016/j.addbeh.2009.08.003

44.

Wang

Kelly

B. C.

Liu

Zhang

Hao

(2013). Factorial structure of the Brief Symptom Inventory (BSI)-18 among Chinese drug users. Drug and Alcohol Dependence, 133, 368-375. doi:10.1016/j.drugalcdep.2013.06.017

45.

Wiesner

Chen

Windle

Elliott

M. N.

Grunbaum

J. A.

Kanouse

D. E.

Schuster

M. A.

(2010). Factor structure and psychometric properties of the Brief Symptom Inventory–18 in women: A MACS approach to testing for invariance across racial/ethnic groups. Psychological Assessment, 22(4), 912-922.

46.

Wright

B. D.

Linacre

J. M.

Gustafson

J. E.

Martin-Lof

(1994). Reasonable mean-square fit values. Rasch Measurement Transactions, 8(3), 370.

47.

Zahn-Waxler

Shirtcliff

E. A.

Marceau

(2008). Disorders of childhood and adolescence: Gender and psychopathology. Annual Review of Clinical Psychology, 4, 275-303. doi:10.1146/annurev.clinpsy.3.022806.091358