TRAUMATIC BRAIN injury (TBI) is a predominant cause

Transcription

1 1621 Depression Assessment After Traumatic Brain Injury: An Empirically Based Classification Method Ronald T. Seel, PhD, Jeffrey S. Kreutzer, PhD, ABPP ABSTRACT. Seel RT, Kreutzer JS. Depression assessment after traumatic brain injury: an empirically based classification method. Arch Phys Med Rehabil 2003;84: Objectives: To describe the patterns of depression in patients with traumatic brain injury (TBI), to evaluate the psychometric properties of the Neurobehavioral Functioning Inventory (NFI) Depression Scale, and to classify empirically NFI Depression Scale scores. Design: Depressive symptoms were characterized by using the NFI Depression Scale, the Beck Depression Inventory (BDI), and the Minnesota Multiphasic Personality Inventory 2 (MMPI-2) Depression Scale. Setting: An outpatient clinic within a Traumatic Brain Injury Model Systems center. Participants: A demographically diverse sample of 172 outpatients with TBI, evaluated between 1996 and Interventions: Not applicable. Main Outcome Measures: The NFI, BDI, and MMPI-2 Depression Scale. The Cronbach, analysis of variance, Pearson correlations, and canonical discriminant function analysis were used to examine the psychometric properties of the NFI Depression Scale. Results: Patients with TBI most frequently reported problems with frustration (81%), restlessness (73%), rumination (69%), boredom (66%), and sadness (66%) with the NFI Depression Scale. The percentages of patients classified as depressed with the BDI and the NFI Depression Scale were 37% and 30%, respectively. The Cronbach for the NFI Depression Scale was.93, indicating a high degree of internal consistency. As hypothesized, NFI Depression Scale scores correlated highly with BDI (r.765) and MMPI-2 Depression Scale T scores (r.752). The NFI Depression Scale did not correlate significantly with the MMPI-2 Hypomania Scale, thus showing discriminant validity. Normal and clinically depressed BDI scores were most likely to be accurately predicted by the NFI Depression Scale, with 81% and 87% of grouped cases, respectively, correctly classified. Normal and depressed MMPI-2 Depression Scale scores were accurately predicted by the NFI Depression Scale, with 75% and 83% of grouped cases correctly classified, respectively. Patients NFI Depression Scale scores were mapped to the corresponding BDI categories, and 3 NFI score classifications emerged: minimally depressed (13 From the Defense and Veterans Brain Injury Center, McGuire Veterans Administration Medical Center, Richmond, VA (Seel); and Department of Physical Medicine and Rehabilitation, Medical College of Virginia, Richmond, VA (Seel, Kreutzer). Supported in part by the National Institute on Disability and Rehabilitation Research, US Department of Education (grant nos. H133B80029, H133P970003); the Defense and Veterans Brain Injury Center, Uniform Services University of the Health Sciences; and the Department of Veterans Affairs. No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit upon the author(s) or upon any organization with which the author(s) is/are associated. Reprint requests to Ronald T. Seel, PhD, McGuire VAMC Defense and Veterans Brain Injury Center, Dept of Physical Medicine and Rehabilitation (117), 1201 Broad Rock Blvd, ATT, Richmond, VA /03/ $30.00/0 doi: /s (03) ), borderline depressed (29 42), and clinically depressed (43 65). Conclusions: Our study provided further evidence that screening for depression should be a standard component of TBI assessment protocols. Between 30% and 38% of patients with TBI were classified as depressed with the NFI Depression Scale and the BDI, respectively. Our findings also provided empirical evidence that the NFI Depression Scale is a useful tool for classifying postinjury depression. Key Words: Brain Injuries; Depression; Outcome assessment (health care); Rehabilitation by the American Congress of Rehabilitation Medicine and the American Academy of Physical Medicine and Rehabilitation TRAUMATIC BRAIN injury (TBI) is a predominant cause of disability in the United States. More than 1.2 million Americans sustain a brain injury each year. 1 Individuals with TBI often face long-term disability as a consequence of physical, cognitive, and emotional sequelae. With regard to neurobehavioral sequelae, depression is among the most common consequences of TBI. Research findings regarding the frequency of post-tbi depression have reported incidence rates ranging from 6% to 77% Commonly reported symptoms include sadness, irritability, loss of interests, fatigue, sleep disturbance, psychomotor retardation, poor concentration, and memory dysfunction Historically, brain injury rehabilitation researchers have used a variety of methodologies to assess depression. 18 Some studies of neurobehavioral functioning and depression relied on semistructured interviews and symptom checklists with unknown psychometric properties. Other studies have measured depression by using assessment instruments that were developed for and standardized with psychiatric populations. In particular, instruments such as the Beck Depression Inventory 23 (BDI) and the Minnesota Multiphasic Personality Inventory 2 24 (MMPI-2) have been widely used by clinicians to classify patients with TBI as depressed or not depressed. A few researchers 2,7,9,11,25-27 have attempted to measure depression with the Diagnostic Statistical Manual for Mental Disorders, 4th edition 28 (DSM-IV). Few studies have used standardized neurobehavioral assessment instruments developed specifically for the TBI population. Several recent studies have used the Neurobehavioral Functioning Inventory 29 (NFI) to characterize postinjury depression symptoms. For example, Kreutzer et al 2 used the NFI to identify DSM-IV major depressive disorder symptoms. The investigators found that 42% of 722 outpatients averaging 35.3 months postinjury met the prerequisite number (5) of criterion A depression symptoms. Fatigue (46%), frustration (41%), and poor concentration (38%) were the most commonly cited manifestations of depression. Later, Seel et al 27 examined NFI responses from a demographically diverse sample of 666 outpatients with TBI enrolled in the National Institute on Disability and Rehabilitation Research Traumatic Brain Injury Model Systems (TBIMS) program. 30,31 Twenty-seven percent of patients with TBI met the DSM-IV prerequisite criteria for major

2 1622 DEPRESSION ASSESSMENT AFTER TRAUMATIC BRAIN INJURY, Seel depressive disorder. Fatigue (29%), distractibility (28%), irritability (28%), and rumination (25%) were the most frequently cited depressive symptoms. Feeling hopeless, feeling worthless, and having difficulty enjoying activities were the 3 symptoms that most differentiated depressed from nondepressed patients. Based on the initial promise of the NFI in identifying depression in persons with TBI, it appeared desirable to develop a classification scheme for NFI Depression Scale scores. Such a scheme could allow clinicians and researchers to reliably categorize depression levels. Similar classification schemes have been developed for the BDI and the MMPI-2 in their application to psychiatric populations. The NFI contains a 13-item Depression Scale that was identified through principal components factor analysis and was validated by using confirmatory factor analytic methods. 29 Kreutzer et al 32 found that the NFI Depression Scale showed high internal consistency (Cronbach.93). Comparison of scores with the MMPI-2 Depression Scale provided evidence of criterion-related validity (r.47). This investigation further examined the NFI Depression Scale s psychometric properties and provided a basis for categorizing scores. Specifically, the primary objectives of this investigation included (1) describing patterns of responses of brain-injured patients on measures including the NFI and the BDI; (2) examining the construct validity of the NFI Depression Scale by using the BDI and the MMPI-2; (3) examining the NFI Depression Scale s discriminant validity; and (4) developing an empirically based scheme for categorizing NFI Depression Scale scores. METHODS Participants The sample consisted of 172 patients with TBI seen for comprehensive evaluations at an outpatient clinic within a TBIMS 30,31 between 1996 and Patients completed the NFI and the BDI. When data from more than 1 evaluation for a patient were available, only the most recently acquired data were examined. The selection rule eliminated the bias of studying a subset of patients more than once and yielded a sample with widely varying postinjury times. Data from patients described in a previous psychometric investigation 32 were not included in this study. Information regarding age at evaluation, postinjury time, sex, ethnicity, and injury etiology was examined. Patients mean age standard deviation (SD) at the time of evaluation was years and ranged from 16 to 75 years. The average postinjury time for evaluation was months, with a median of 12.2 months, and ranged from 1 to 323 months. The sample was 62% male. With regard to ethnicity, 70% were white, 22% were African American, and 8% were of other ethnic backgrounds. Most patients (68%) were involved in vehicular collisions. Violence and falls accounted for 13% and 12% of injuries, respectively. With regard to injury severity, several indices were examined. The mean duration of unconsciousness was days. Half of the patients reported a loss of consciousness of 20 minutes or less. The mean number of days that patients experienced posttraumatic amnesia (PTA) was One third of patients experienced 7 or more days of PTA. The mean Glasgow Coma Scale (GCS) score at the time of system admission was , and 44% of patients were admitted with a GCS score of 8 or less. Data Collection Experienced clinicians collected injury and demographic data from medical records, health and history questionnaires, and patient and family interviews. An information package containing history questionnaires and the NFI was mailed to patients before their appointment. On arrival for evaluation, their responses were reviewed for completeness and compliance with instructions. Whenever possible, patients forms were corrected, and missing information was obtained. Clinicians conducted interviews, administered psychological tests, and derived scores for outcome measures. Instruments Neurobehavioral Functioning Inventory. The NFI 29 is composed of 76 items. Seventy items comprise 6 scales based on principal components and confirmatory factor analytic methodology: depression, somatic complaints, memory/attention difficulties, communication deficits, aggressive behaviors, and motor impairment. Six additional items, including inability to work, suicidal ideation, and seizures, address critical factors. The NFI Depression Scale is composed of 13 items that address themes including hopelessness, anhedonia, social isolation, frustration, and rumination. Subjects are asked to rate the frequency of problem occurrence based on a 5-point scale: 1, never; 2, rarely; 3, sometimes; 4, often; or 5, always. Research 32 has provided support for criterion-related validity and high internal consistency within NFI scales. The Cronbach for individual scales ranges from.86 to.95. Research 33 has shown high levels of agreement between patients and family members perceptions of difficulties. A comparison of patient and family responses on NFI items showed that between 93% and 96% either were in exact agreement or had a difference of 1 point. The NFI has been used to assess the frequency of depression-related symptoms. 2,27 In addition, scores for the NFI Depression Scale have been found to correlate positively with other measures of depression, including the MMPI-2 (depression). 32 Beck Depression Inventory. The BDI 23 is a 21-item selfreport instrument designed to screen for depression primarily cognitive and affective symptoms. Responses are made on a 4-point, minimally anchored scale, ranging from 0 to 3, with 3 being the most severe symptoms. An example of the 4-choice items includes the following: 0, I don t have thoughts of killing myself ; 1, I have thoughts of killing myself, but I would not carry them out ;2, I would like to kill myself ; and 3, I would kill myself if I had the chance. A total BDI score is calculated by adding all responses, with a possible range of scores from 0 to A range of 0 to 10 is considered normal; 11 to 16, minimal depression; 17 to 20, borderline clinical depression; 21 to 30, moderate depression; 31 to 40, severe depression; and more than 40, extreme depression. 23,35 Research has provided support for internal consistency. The Cronbach has ranged from.73 to.92, with test-retest reliability more than ,37 With regard to convergent validity, BDI scores have correlated positively with the Zung Self-Rating Depression Scale and the Hamilton Rating Scale. 36,38 Although the BDI was originally developed with a psychiatric population, the instrument has been used with neurologic populations, including those with severe closed head injury, stroke, and Parkinson s disease Minnesota Multiphasic Personality Inventory 2. The MMPI-2 24 is a self-administered questionnaire composed of 567 true-false questions intended to measure an individual s level of emotional adjustment and attitude toward test taking. The MMPI-2 comprises 10 clinical scales, including the De-

3 DEPRESSION ASSESSMENT AFTER TRAUMATIC BRAIN INJURY, Seel 1623 pression Scale, which contains 57 items. These items make up 5 subscales, including subjective depression (D1), psychomotor retardation (D2), physical malfunctioning (D3), mental dullness (D4), and brooding (D5). A high score on the clinical scales, indicating more difficulty, has varied considerably, with some researchers considering scores above 65 as significant and others defining a high score as in the upper quartile. 42,43 Low scores have also been shown to convey important diagnostic information. 42,43 Internal consistency has ranged from.23 to.85 for these subscales and from.59 to.64 for the Depression Scale as a whole. 24 Test-retest coefficients for the Depression Scale have ranged from.75 to Research 44 has also provided support for good discriminant validity between DSM-III diagnoses and the MMPI-2. There is evidence that the MMPI-2 has clinical utility in the assessment of a population with TBI. 45 Data Analyses Descriptive statistics were generated to describe the patient sample, mean scores on the 3 measures of mood functioning, and mean responses on NFI Depression Scale and BDI items. The percentage of patients reporting each item was also calculated. For the BDI, patients responses are typically calculated as symptomatic when rated as 1 or more. 23 To provide a description of patient responses equivalent to the BDI, NFI depression items were calculated as symptomatic when rated as occurring sometimes, often, or always. Quantitative analyses were performed to examine the relationship between demographic characteristics and patients self-reported depressive symptoms. For gender, ethnicity, and injury etiology, 1-way analyses of variance (ANOVAs) were conducted. 46 For age and number of months after injury at time of evaluation, Pearson r correlations were calculated. 47 To evaluate the distribution of scores and reliability of the NFI Depression Scale, 2 statistical tests were applied. First, a frequency analysis was conducted to test the extent to which the sample distribution conformed to a normal curve. Second, the reliability of the NFI Depression Scale was evaluated. Given the homogeneous nature of the scale and its design to measure the state of patients depressive symptoms, an analysis of internal consistency reliability was deemed most appropriate. 48,49 A multitrait-multimethod approach was used to evaluate the construct validity of the NFI Depression Scale. 50 This research method uses measurements from multiple assessment instruments and calculates correlation coefficients to evaluate convergent and discriminant validity. The BDI total score and the MMPI-2 Depression Scale T score were selected as measures of convergent validity. A significant positive correlation was predicted among the 3 instruments. Data for the MMPI-2 Hypomania Scale were selected to help evaluate discriminant validity. No significant correlation was expected between the Hypomania Scale and the NFI Depression Scale. Pearson correlation coefficients were calculated to establish the relationships between scores. To assess the predictive validity of the NFI Depression Scale, 2 canonical discriminant function analyses were conducted. 51,52 Canonical discriminant function analysis measures the extent to which a quantitative predictor variable correctly classifies group membership, by using a qualitative grouping variable. For the first analysis, the BDI was selected as the grouping variable. Patients group membership was categorized into 3 levels based on their BDI total score (0 10, normal; 11 20, mild to borderline depressed; 21, clinically depressed). For the second analysis, the MMPI-2 Scale 2 (depression) was selected as the grouping variable. Patients group membership was categorized into 2 levels based on their depression T score (34 64, not depressed; 65, depressed). The NFI Depression Scale was the sole predictor variable entered in both analyses. Eigenvalues and canonical correlations were calculated to analyze the relationships between predictor and grouping variables. The Wilks, a multivariate test of significance, was calculated to assess the significance of the main effect for the predictor variable (NFI Depression Scale). Wilks values range from 0 to 1. Values close to 0 indicate that group means differ; values close to 1 indicate that group means do not differ. A conservative level was set, minimizing the chance of a type I error. Considering familywise error rates, an level of P less than.01 was deemed acceptable. After instrument validity was examined, an attempt was made to enhance the interpretability of NFI Depression Scale scores. To identify logical cutoff scores for the NFI Depression Scale, the following procedure was used. The BDI was selected as the criterion variable. Each patient was classified into 1 of 6 BDI categories (normal, mild mood disturbance, borderline clinical depression, moderate depression, severe depression, very severe depression) based on their BDI total score. A matrix was then designed in which patients NFI Depression Scale scores were mapped to their corresponding BDI category. The matrix was then examined, and NFI cutoff scores were derived based on a decision rule that emphasized the minimization of false negatives and false positives. Due to the clinical relationship between depression and suicidal behaviors, as well as the critical importance of accurately assessing suicidality, NFI Depression Scale scores were compared with self-reported suicidal ideation and intent. BDI item 9, which addresses 4 levels of suicidal behavior (ranging from no suicidal thoughts to high suicidal intent), was selected as the criterion variable. A second matrix was then designed in which patients NFI Depression Scale scores were mapped to their corresponding BDI suicidal behavior level. The matrix was then examined to identify the rate of false negatives that is, the proportion of patients with low NFI Depression Scale scores who reported suicidal intent. RESULTS Self-Reported Depressive Symptoms NFI Depression Scale items were examined, and the proportion of symptomatic patients was identified (see table 1). Frustration was the most predominant problem (81%), and feeling scared or frightened was least predominant (38%). Two thirds or more of patients reported problems with boredom, sadness, rumination, and restlessness. The mean total score for the Depression Scale was A similar process was used to identify symptomatic patients with the BDI (see table 2). Fatigability was the most predominant problem (80%), and suicidal ideation was the least predominant (21%). Two thirds or more of patients were symptomatic with regard to dissatisfaction, irritability, indecisiveness, work difficulty, and insomnia. The mean BDI score was The percentages of patients classified as moderately, severely, or extremely depressed were 24%, 12%, and 2%, respectively. Relation Between Demographic Characteristics and Self- Reported Depression Quantitative analyses were performed to examine the relation between demographic characteristics and patients selfreported depressive symptoms. Age at injury was correlated

4 1624 DEPRESSION ASSESSMENT AFTER TRAUMATIC BRAIN INJURY, Seel Table 1: NFI Depression Scale Items of Symptomatic Patients Depression Items Symptomatic (%) Mean SD Frustrated Restless Can t get mind off certain thoughts Bored Sad, blue Difficulty enjoying activities Uncomfortable around others Lonely Feels hopeless No confidence Sits with nothing to do Feels worthless Scared or frightened Total score Cronbach.93 NOTE. Symptomatic is defined as occurring sometimes, often, or always. positively with NFI depression scores (r.19, P.015). Time after injury, days in coma, days in PTA, and GCS at the time of system admission did not correlate significantly with NFI depression scores (P.05). ANOVAs were calculated to examine the effect of categorical demographic variables on NFI depression scores. No main effects were identified for sex, ethnicity, or injury etiology (P.05). Score Distribution and Reliability A frequency analysis was conducted to test the extent to which the sample distribution of NFI depression scores conformed to a normal curve (see fig 1). The distribution of NFI depression scores closely conformed to a normal curve. The Table 2: BDI Items of Symptomatic Patients BDI Item Content Symptomatic Mean SD Fatigability 80% Irritability 76% Work difficulty 75% Indecisiveness 68% Insomnia 68% Dissatisfaction 68% Self-accusations 62% Somatic preoccupation 58% Self-dislike 55% Sadness 52% Pessimism 48% Loss of libido 46% Crying 45% Social withdrawal 45% Sense of failure 40% Body-image change 39% Loss of appetite 39% Punishment 34% Weight loss 33% Guilt 32% Suicidal ideas 21% Total score Cronbach.91 NOTE. Symptomatic is used to describe items with ratings 1. Fig 1. Distribution of NFI Depression Scale total scores (N 172.0; mean SD, ). mean NFI depression score was , with a range of 13 to 65. The distribution neither was skewed (.05) nor showed kurtosis (.50). To examine the reliability of the NFI Depression Scale, an analysis of internal consistency was calculated. The Cronbach for the NFI Depression Scale was.93, indicating a high degree of internal consistency. Construct Validity To examine construct validity, a multitrait-multimethod approach was applied (see table 3). As hypothesized, the NFI Depression Scale scores correlated highly with the BDI scores (r.765) and the MMPI-2 Depression Scale T scores (r.752). The strength of the correlation among the 3 measures suggested that the NFI Depression Scale showed convergent validity. The NFI Depression Scale did not correlate significantly with the MMPI-2 Hypomania Scale, thus showing discriminant validity. Of interest, the BDI correlated positively with the MMPI-2 Hypomania Scale. This finding suggests that the NFI Depression Scale may show superior discriminant validity compared with the BDI. Predictive Validity With Discriminant Function Analysis Two canonical discriminant function analyses were conducted to assess the predictive validity of the NFI Depression Scale. For the first analysis, BDI category was selected as the Table 3: NFI Depression Scale: Multitrait-Multimethod Analysis of Convergent and Discriminant Validity With the MMPI-2 and BDI Instrument/Scale NFI Depression BDI MMPI-2 Depression MMPI-2 Hypomania NFI Depression Scale BDI total score.765 MMPI-2 Depression Scale MMPI-2 Hypomania Scale *.093 NOTE. Boldface indicates convergent tests; positive correlations are hypothesized. Italics indicate discriminant tests; no correlations are hypothesized. *P.05; P.001.

5 DEPRESSION ASSESSMENT AFTER TRAUMATIC BRAIN INJURY, Seel 1625 Table 4: Discriminant Function Analysis: NFI Depression Scale Scores Predicting BDI Group Membership Actual BDI category Predicted Group Membership Normal Mild/Borderline Depressed Normal 81% 8% 11% Mild/borderline 32% 27% 41% Depressed 9% 4% 87% NOTE. A total of 70.3% of original grouped cases were correctly classified. grouping variable. The model yielded a significant predictive relationship for NFI Depression Scale score on BDI category (Wilks.490, P.001). A canonical correlation of.714 indicated that there was a significant relationship between patients BDI categories and NFI Depression Scale scores. A significant factor loading (eigenvalue 1.041) indicated that the NFI Depression Scale contributed a significant proportion of the model variance. A review of the classification matrix indicated that 70.3% of grouped cases were correctly classified (see table 4). Normal and clinically depressed BDI scores were most likely to be accurately predicted by the NFI Depression Scale, with 81% and 87% of grouped cases, respectively, correctly classified. Patients with mild and borderline clinically depressed BDI scores were less likely to be correctly classified by NFI Depression Scale score. For the second canonical discriminant function analysis, the MMPI-2 Depression Scale category was selected as the grouping variable. The second model also yielded a significant relationship, with NFI Depression Scale scores predicting MMPI-2 Depression Scale categories (Wilks.589, P.001). A canonical correlation of.641 indicated that there was a significant relationship between patients MMPI-2 Depression Scale categories and NFI Depression Scale scores. A significant factor loading (eigenvalue.699) indicated that the NFI Depression Scale contributed a significant proportion of the model variance. A review of the classification matrix indicated that 80% of grouped cases were correctly classified (see table 5). Normal and depressed MMPI-2 Depression Scale scores were accurately predicted by the NFI Depression Scale, with 75% and 83% of grouped cases, respectively, correctly classified. NFI Depression Scale Score Classifications and BDI Total Scores Patients NFI Depression Scale scores were mapped to corresponding BDI categories (see table 6), and the resulting distribution was examined. With an emphasis on reducing the rate of false negatives and false positives, 3 groups of patients were identified. The minimally depressed group had scores ranging from 13 to 28 and a group mean of More than 88% of patients who were minimally depressed based on NFI scores were also categorized as not depressed based on Table 5: Discriminant Function Analysis: NFI Depression Scale Scores Predicting MMPI-2 Depression Group Membership Actual MMPI-2 Depression Category Predicted Group Membership Not Depressed Depressed Not depressed 75% 25% Depressed 17% 83% NOTE. A total of 80% of original grouped cases were correctly classified. Table 6: NFI Depression Scores Mapped to BDI Depression Category NFI Depression Score BDI Depression Category Normal Mild/Borderline Depressed BDI total scores. Only 1 patient was minimally depressed with the NFI Depression Scale and scored moderately to severely depressed on the BDI. The classification scheme yielded a relatively low rate of false negatives. Scores for the borderline depressed group ranged from 29 to 42, with a mean NFI depression score of Fortyone percent of patients were also classified as borderline depressed based on their BDI total score. False-negative and false-positive classifications relative to the BDI were identified. Twenty-six percent of the patients who were rated as borderline depressed on the NFI Depression Scale were moderately to severely depressed based on BDI total scores. Conversely, 33% of patients who were rated as borderline depressed on the NFI Depression Scale were rated as minimally depressed with the BDI. The clinically depressed group had scores of 43 or greater, with a mean NFI depression score of The rate of false positives was very low. Only 1 patient classified as clinically depressed with the NFI was classified as not depressed based on the BDI total score. Nearly 90% of patients who were classified as clinically depressed based on their NFI Depression Scale scores were also classified as moderately to severely depressed based on their BDI total scores. The percentage of patients who were classified as clinically depressed based on NFI Depression Scale scores was 30%. NFI Depression Scale Score Classifications and Suicidal Risk NFI Depression Scale scores were mapped to scores for the BDI suicidal thoughts item (see table 7). The resulting matrix identified the proportion of patients classified by NFI scores as minimally depressed who reported suicidal ideation or intent. Only 1 patient with a minimally depressed NFI score reported that he would like to kill himself. None of the patients with minimally depressed scores reported that they would kill (themselves) if (they) had the chance. Nearly 98% of patients with nondepressed NFI scores reported minimal to no suicidal intent. Indeed, the classification scheme yielded a low rate of false negatives. DISCUSSION In the past several decades, brain injury researchers have focused their efforts on studying physical, functional, and cognitive outcomes. More recently, researchers have identified emotional distress and depression as a major problem for survivors. In a series of studies, the NFI was used to charac-

6 1626 DEPRESSION ASSESSMENT AFTER TRAUMATIC BRAIN INJURY, Seel Table 7: NFI Depression Scores Mapped to BDI Suicidal Thoughts Item NFI Depression Score No Suicidal Thoughts Suicidal Thoughts BDI Suicidal Thoughts Item Would Like to Kill Self Would Kill Self If Had Chance terize postinjury depressive symptoms. The NFI was developed and standardized with a large brain injury sample and has served as a primary outcome measure for the TBIMS program. 29,30 Questions have arisen regarding the application of the NFI Depression Scale as a screening instrument for postinjury depressive disorders. This study further delineated the patterns of depression in brain injury survivors, evaluated the psychometric properties of the NFI Depression Scale, and empirically classified NFI Depression Scale scores. Patterns of Depression After TBI Across assessment instruments, most patients reported having at least some problems with depressive symptoms, including frustration, rumination, sadness, negative self-image, and dissatisfaction. Based on BDI scores, 38% of the sample was categorized as moderately or severely depressed. With the NFI Depression Scale, 30% of the sample was classified as clinically depressed. The depression rates in this TBI sample are consistent with existing studies, in which rates have typically ranged between 27% and 42%. 2,9-11,24,25 Our study provides further evidence that depression is a common occurrence among persons with TBI. The findings support recent recommendations that screening for clinical depression should be a standard component of any postinjury assessment and treatment protocol. 25 Reliability and Validity of the NFI A second objective of this investigation was to examine the psychometric properties of the NFI Depression Scale. The results from this study showed that internal consistency was quite high (Cronbach.93). This high reliability level was consistent with findings from the investigation of Kreutzer et al, 29 who reported an of.93 for the NFI Depression Scale. The internal consistency findings from these 2 studies provide strong evidence that the 13 NFI Depression Scale items are a homogeneous symptom set that define a single construct: depression. The normal distribution of depression scores suggested that the scale s item content and response format allowed patients with low, medium, and high levels of depression to be differentiated. Evidence of convergent validity was established by robust positive correlations between the NFI Depression Scale scores and total scores on the BDI and the MMPI-2 Depression Scale. The results provided strong evidence that the NFI Depression Scale measures the general construct of depression. Preliminary evidence suggested that the NFI Depression Scale might possess better discriminant validity than the BDI. Scores on the NFI Depression Scale were unrelated to a measure of hypomania, whereas BDI scores were positively correlated. A third objective of the study was to establish the predictive validity of NFI Depression Scale scores by using discriminant function analysis. Groups of patients were identified as depressed by using BDI and MMPI-2 classifications. NFI Depression Scale scores accurately identified nearly 90% of depressed patients. Additionally, NFI depression scores accurately identified approximately 80% of patients with little or no depression. Less predictive accuracy was observed for borderline levels of clinical depression. In summary, the NFI Depression Scale was found to be a valid and accurate predictor of clinical depression. Classification Levels and Clinical Use of NFI Depression Scale Scores Empirical analysis provided a means to further classify depression levels based on NFI scores. Persons with NFI Depression Scale scores of 28 or less (minimal depression) were highly unlikely to score as depressed on the BDI or to report suicidal ideation and intent. In measuring depression, one first and foremost aspires to avoid false-negative classifications. Our empirical classification of minimally depressed patients showed an acceptably low rate ( 3%) of false negatives. Conversely, persons with NFI depression scores of 43 or greater (clinical depression) were highly likely to score as moderately or severely depressed based on BDI responses. In addition, persons with NFI depression scores of 43 or greater were highly likely to report suicidal ideation and were at significant risk for suicidal intent. The scale s empirical classification of clinical depression also showed acceptably low levels of false positives. Our research findings provided support for using the NFI Depression Scale as a tool for identifying postinjury depression. Table 8 presents the range of scores for the 3 NFI Depression Scale classification levels, presents an interpretation of each range of scores, and provides clinical recommendations based on depression severity. Persons who score as clinically depressed have a high likelihood of having a major depressive disorder and require immediate clinical attention. Persons whose scores range between 29 and 42 (borderline depression) are likely experiencing a mood disturbance that should alert clinicians to the need for additional assessment to establish an accurate diagnosis. The high false-positive and false-negative prediction rates for NFI borderline depression scores further highlight the need for additional assessment. Such an assessment might include clinical observation and a DSM-IV based 26 interview with the patient and perhaps a family member. Persons who are minimally depressed likely will not require clinical treatment. Regardless of depression level, clinicians should review patient responses to NFI items that measure hopelessness and self-harm. The NFI appears to be a valuable screening tool for depression within the TBI population. Conversely, clinicians should note that the use of any self-report depression instrument does not allow for the fine distinctions necessary to make a formal diagnosis of a major depressive disorder. For example, NFI Depression Scale items do not succinctly address symptom duration or the full range of DSM-IV criterion A symptoms for depression. As with all self-report depression instruments, a structured diagnostic interview would be necessary to fully identify the DSM-IV inclusion and exclusion criteria.

7 DEPRESSION ASSESSMENT AFTER TRAUMATIC BRAIN INJURY, Seel 1627 Table 8: NFI Depression Scale Scores, Interpretation, and Clinical Recommendations Scale Score Range Classification and Interpretation Assessment and Treatment Recommendations Minimal depression. The patient rarely experiences depressive symptoms or suicidal intent. There is a less than 3% falsenegative rate for clinical depression and suicidal intent based on BDI ratings Borderline depression. The patient sometimes or often experiences some symptoms associated with a depressive disorder. There is a high likelihood (65%) that the patient has suicidal ideation, with a moderate risk (13%) for plan and intent. High false-positive and false-negative prediction rates for NFI borderline depression scores further warrant the need for additional assessment Clinical depression. The patient often or always experiences a wide range of symptoms consistent with a major depressive episode. There is a very high likelihood (86%) that the patient has suicidal ideation, with a significant risk (37%) for plan and intent. There is a less than 2% false-positive rate for minimal depression based on BDI ratings. Review NFI items related to hopelessness and self-harm to identify areas of concern. Monitoring is likely to be helpful, and immediate clinical intervention is not likely to be required. Review Depression Scale and other NFI items to identify risk factors, comorbidities, and areas requiring further assessment. Conduct structured interview for differential diagnosis; assess suicidal history, ideation, plan, and intent. Develop and implement treatment plan. Review Depression Scale and other NFI items to identify risk factors, comorbidities, and areas requiring further assessment. Conduct structured interview for differential diagnosis; assess suicidal history, ideation, plan, and intent. Develop and implement treatment plan. For extreme scores ( 58), review all NFI items; rule out response set bias, cry for help, and secondary gain issues. Limitations and Future Research Several limitations should be noted. First, this investigation studied patients drawn from a single center. Future research should examine the applicability of these findings to a larger sample drawn from multiple settings. Second, a clinical interview and DSM-IV diagnostic data were not evaluated as part of this study. Future research using the NFI Depression Scale should examine construct validity by using the results from DSM-IV interviews for the full range of depressive and adjustment disorders. Third, the validity of the NFI depression classification system is still questionable for select patient populations with specific cognitive impairments. For example, patients with aphasic impairments were not included in this sample. Future research should examine whether the NFI depression classification system is appropriate for the full range of injury severity and cognitive impairment types. Fourth, data on patients premorbid depression was not available for analysis. When interpreting the results from our study, readers should not make inferences regarding a causal relationship between TBI and depression. Future research should examine the extent to which TBI might either cause depression or exacerbate patients predisposition for depression, as well as whether depression might be a risk factor for sustaining TBI. Last, future research should compare NFI depression data derived from patients and family members. CONCLUSIONS Our study provided further evidence that screening for depression should become a standard component of TBI assessment protocols. Between 30% and 38% of patients with TBI were classified as depressed with the NFI Depression Scale and the BDI, respectively. Findings also provided evidence that the NFI Depression Scale is a useful tool for classifying postinjury depression. References 1. Holsinger T, Steffens DC, Phillips C, et al. Head injury in early adulthood and the lifetime risk of depression. Arch Gen Psychiatry 2002;59: Kreutzer JS, Seel RT, Gourley E. The prevalence and symptom rates of depression after traumatic brain injury: a comprehensive examination. Brain Inj 2001;15: Rutherford WH, Merrett JD, McDonald JR. Sequelae of concussion caused by minor head injuries. Lancet 1977;1: Brooks DN, McKinlay W. Personality and behavioral change after severe blunt head injury a relative s view. J Neurol Neurosurg Psychiatry 1983;46: Atteberry-Bennett J, Barth JT, Loyd BH, et al. The relationship between behavioral and cognitive deficits, demographics and depression in patients with minor head injuries. Int J Clin Neuropsychol 1986;8: Lezak MD. Relationship between personality disorders, social disturbances, and physical disability following traumatic brain injury. J Head Trauma Rehabil 1987;2(1): Varney NR, Martzke JS, Roberts RJ. Major depression in patients with closed head injury. Neuropsychology 1987;1: Kinsella G, Moran C, Ford B, Ponsford J. Emotional disorder and its assessment within the severe head injured population. Psychol Med 1988;118: Federoff JP, Starkstein SE, Forrester AW, et al. Depression in patients with acute traumatic brain injury. Am J Psychiatry 1992; 149: Jorge RE, Robinson RG, Starkstein SE, Arndt SV. Depression and anxiety following traumatic brain injury. J Neuropsychiatry Clin Neurosci 1993;5: Jorge RE, Robinson RG, Arndt S. Are there symptoms that are specific for depressed mood in patients with traumatic brain injury? J Nerv Ment Dis 1993;181: Bowen A, Neumann V, Conner M, Tennant A, Chamberlain MA. Mood disorders following traumatic brain injury: identifying the extent of the problem and the people at risk. Brain Inj 1998;12:

8 1628 DEPRESSION ASSESSMENT AFTER TRAUMATIC BRAIN INJURY, Seel 13. Deb S, Lyons I, Koutzoukis C, Ali I, McCarthy G. Rate of psychiatric illness 1 year after traumatic brain injury. Am J Psychiatry 1999;156: Oddy M, Humphrey M, Uttley D. Subjective impairment and social recovery after closed head injury. J Neurol Neurosurg Psychiatry 1978;41: McKinlay WW, Brooks DN, Bond MR, Martinage DP, Marshall MM. The short-term outcome of severe blunt head injury as reported by the relatives of the injured person. J Neurol Neurosurg Psychiatry 1981;44: Cavallo MM, Kay T, Ezrachi O. Problems and changes after traumatic brain injury: differing perceptions within and between families. Brain Inj 1992;6: Thomsen IV. Late outcome of very severe blunt head trauma: a year second follow-up. J Neurol Neurosurg Psychiatry 1984;47: Rosenthal M, Christensen BK, Ross TP. Depression following traumatic brain injury. Arch Phys Med Rehabil 1998;79: Dikmen S, Reitan RA. Emotional sequelae of head injury. Ann Neurol 1977;2: Leininger BE, Kreutzer JS, Hill MR. Comparison of minor and severe head injury emotional sequelae using the MMPI. Brain Inj 1991;5: Schoenhuber R, Gentilini M. Anxiety and depression after mild head injury: a case control study. J Neurol Neurosurg Psychiatry 1988;51: Lykouras L, Oulis P, Adrachta D, et al. Beck Depression Inventory in the detection of depression among neurological inpatients. Psychopathology 1998;31: Beck A, Steer R. Beck Depression Inventory manual. San Antonio: The Psychological Corp; Butcher JN, Dahlstrom WG, Graham JR, Tellegen A, Kaemmer B. Minnesota Multiphasic Personality Inventory-2 (MMPI-2) manual for administration and scoring. Minneapolis: Univ Minnesota Pr; Alexander M. Neuropsychiatric correlates of persistent postconcussive syndrome. J Head Trauma Rehabil 1992;7(2): Fann J, Katon W, Uomoto J, Esselman P. Psychiatric disorders and functional disability in outpatients with traumatic brain injury. Am J Psychol 1995;152: Seel RT, Kreutzer JS, Rosenthal M, Hammond FM, Corrigan JD, Black K. Depression after traumatic brain injury: a National Institute on Disability and Rehabilitation Research Model Systems multicenter investigation. Arch Phys Med Rehabil 2003;84: American Psychiatric Association. Diagnostic and statistical manual of mental disorders. 4th ed. Washington (DC): APA; Kreutzer J, Seel R, Marwitz J. The Neurobehavioral Functioning Inventory. San Antonio: The Psychological Corp; Dahmer E, Shilling M, Hamilton C, et al. A model systems database for traumatic brain injury. J Head Trauma Rehabil 1993; 8(2): Harrison-Felix C, Newton N, Hall K, Kreutzer J. Descriptive findings from the traumatic brain injury model systems national database. J Head Trauma Rehabil 1996;11(5): Kreutzer J, Marwitz J, Seel R. Validation of a Neurobehavioral Functioning Inventory for adults with traumatic brain injury. Arch Phys Med Rehabil 1996;77: Seel R, Kreutzer J, Sander A. Concordance of patients and family members ratings of neurobehavioral functioning after traumatic brain injury. Arch Phys Med Rehabil 1997;78: Beck A, Beck R. Screening depressed patients in family practice. Postgrad Med 1972;52: Beck A, Steer R, Garbin M. Psychometric properties of the Beck Depression Inventory: twenty-five years of evaluation. Clin Psychol Rev 1988;8: Reynolds W, Gould J. A psychometric investigation of the standard and short form Beck Depression Inventory. J Consult Clin Psychol 1981;49: Steer R, Beck A, Brown G. Sex differences on the revised Beck Depression Inventory for outpatients with affective disorders. J Pers Assess 1989;53: Brown C, Schulberg H, Madonia M. Assessing depression in primary care practice with the Beck Depression Inventory and the Hamilton Rating Scale for Depression. Psychol Assess 1995;7: Garske G, Thomas K. Self-reported self-esteem and depression: indexes of psychosocial adjustment following severe traumatic brain injury. Rehabil Counsel Bull 1992;36: Levin B, Llabre M, Weiner W. Parkinson s disease and depression: psychometric properties of the Beck Depression Inventory. J Neurol Neurosurg Psychiatry 1988;51: Ng K, Chan K, Straughan P. Depression and functional impairment in stroke victims. Acta Psychiatr Scand 1995;92: Graham J. MMPI-2: assessing personality and psychopathology. 2nd ed. New York: Oxford Univ Pr; Green R. MMPI-2/MMPI: an interpretive manual. 2nd ed. Boston: Allyn & Bacon; Patrick J. Concordance of the MCMI and the MMPI in the diagnosis of three DSM-III Axis I disorders. J Clin Psychol 1988;44: Arbisi P, Ben-Porath Y. The use of the Minnesota Multiphasic Personality Inventory 2 in the psychological assessment of persons with TBI: correction factors and other clinical caveats and conundrums. Neurorehabilitation 1999;13: Keppel G. Design and analysis: a researcher s handbook. Englewood Cliffs: Prentice Hall; Porter J, Hamm R. Statistics: applications for the behavioral sciences. Monterey: Brooks/Cole Publishing; Dawes R. Scale construction. J Counsel Psychol 1987;34: Anastasi A. Psychological testing. New York: MacMillan Publishing; Campbell D, Fiske D. Convergent and discriminant validation by the multitrait-mutlimethod matrix. Psychol Bull 1959;56: Tabachnick B, Fidell L. Using multivariate statistics. New York: HarperCollins Publishers; Betz N. Use of discriminant analysis in counseling psychology research. J Counsel Psychol 1987;34: