Challenges Associated With Screening for Traumatic Brain Injury Among US Veterans Seeking Homeless Services

Abstract

We identified the prevalence of traumatic brain injury (TBI) among homeless veterans and assessed the TBI–4, a screening tool created to identify TBI history. Between May 2010 and October 2011, 800 US veterans from two hospitals, one eastern (n = 122) and one western (n = 678) completed some or all measures. Findings suggested that 47% of veterans seeking homeless services had a probable history of TBI (data for prevalence obtained only at the western hospital). However, psychometric results from the screening measure suggested that this may be an underestimate and supported comprehensive assessment of TBI in this population.

Eradicating homelessness and assessing and treating traumatic brain injury (TBI) are key areas of focus for the US Department of Veterans Affairs (VA).^1,2 Research from Canada has suggested that 53% of homeless individuals have a history of TBI, yet the prevalence of TBI among homeless veterans is unknown.³ Additionally, TBI screening measures for homeless veterans have yet to be validated. We aimed to determine the prevalence of TBI among veterans seeking homeless services, such as those provided by the Grant and Per Diem⁴ and the US Department of Housing and Urban Development’s VA Supportive Housing Programs,⁵ and hypothesized that the proportion of probable TBI in this group would be significantly higher than that found among homeless civilians (53%).³ Additionally, we evaluated the utility of a TBI screening tool (TBI–4)⁶ and hypothesized that the sensitivity and specificity of the TBI–4 compared with a gold-standard Ohio State University Traumatic Brain Injury–Identification Method (OSU TBI–ID)⁷ would be significantly greater than 0.75 and 0.80, respectively.

METHODS

We obtained data used to determine prevalence from 678 veterans at a western metropolitan VA hospital at which TBI–4 data were collected as part of standard procedures to obtain homeless services. The mean age at assessment was 51.9 years (SD = 9.8), and 642 (94.7%) were men.

Data used for analysis of criterion-related classification accuracy of the TBI–4 relative to the OSU TBI–ID were from 313 veterans who completed study measures. This cohort included 191 veterans from the same western VA hospital and 122 veterans from an eastern metropolitan VA hospital. The mean age at assessment was 52.2 years (SD = 8.6), and 302 (96.5%) were men.

Measures

The TBI–4⁶ is a 4-question brief screen for TBI history that was created for inclusion in a VA hospital’s mental health intake form. For our analyses, responses were coded in 2 ways. First, we coded a positive response to question 2 as a positive screen for probable TBI (Table 1). This question is most similar to that used in past prevalence research^3,8 and queries for an injury event with an associated loss of consciousness. Second, we coded a positive response to any of the 4 questions as a positive screen for possible TBI. The distinction between positive screens for probable versus possible TBI is necessary because positive responses to questions 1, 3, or 4 may indicate a history of injury without TBI. The OSU TBI–ID⁷ is a psychometrically sound structured clinical interview designed to elicit reports of TBI occurring over a person’s lifetime.

TABLE 1—

Positive Responses for Each TBI-4 Question: 1 Western and 1 Eastern US Metropolitan Veterans Affairs Hospital, May 2010-October 2011

Question	(1) Have you ever been hospitalized or treated in an emergency room following a head or neck injury?	(2) Have you ever been knocked out or unconscious following an accident or injury?	(3) Have you ever injured your head or neck in a car accident or from some other moving vehicle accident?	(4) Have you ever injured your head or neck in a fight or a fall?	Any 1 of 4
Positive response	249	317	216	251	400
Percentage positive (95% CI)	37 (33, 40)	47 (43, 51)	32 (28, 36)	37 (33, 41)	59 (55, 63)

Note. CI = confidence interval; TBI-4 = traumatic brain injury-4. The sample size was n = 678.

Procedures

At the western VA hospital, TBI–4 responses and demographic information for all veterans presenting for homeless services were gleaned via chart review. Of these veterans, 191 consented to complete additional study measures. At the eastern VA hospital, 122 veterans presenting for homeless services consented to complete the study measures, including the TBI–4; their demographic information was retrieved by chart review.

RESULTS

For the 678 veterans whose TBI–4 responses were collected at homeless intake evaluations, the prevalence of probable TBI, measured by question 2 of the TBI–4, was 47%.

This prevalence was significantly lower than hypothesized, with t(677) = −3.26; P = .001; and 95% confidence interval (CI) = 43%, 51%. The prevalence of possible TBI among veterans seeking homeless services was 59%, significantly higher than 53%, with t(677) = 3.17; P = .002; 95% CI = 55%, 63% (Table 1).

Using the OSU TBI–ID as the gold standard and a positive response to any question on the TBI–4 as the criterion for a possible TBI, sensitivity was 220 of 285, or 0.77 (97.5% CI = 0.71, 0.83) and specificity was 23 of 28, or 0.82 (97.5% CI = 0.60, 0.95). When using a positive response to question 2 as the criterion for probable TBI, sensitivity was 181 of 285, or 0.64 (97.5% CI = 0.57, 0.70) and specificity was 27 of 28, or 0.96 (97.5% CI = 0.79, 1.00; Table 2). The sensitivity and specificity of the TBI–4, on the basis of both possible and probable screens, was not significantly greater than the targeted values. Question 2 of the TBI–4 did not identify 36% of veterans who actually had a history of TBI on the basis of the gold-standard measure.

TABLE 2—

Criterion Validity of the TBI-4

	Yes to Any TBI-4 Question Is Considered Positive for TBI			Yes to TBI-4 Question 2 Is Considered Positive for TBI
TBI-4	Positive OSU, No.	Negative OSU, No.	Total No.	Positive OSU, No.	Negative OSU, No.	Total No.
Positive	220	5	225	181	1	182
Negative	65	23	88	104	27	131
Total	285	28	313	285	28	313

Note. OSU = Ohio State University TBI Identification Method; TBI-4 = Traumatic Brain Injury–4.

DISCUSSION

Nearly half of the veterans seeking homeless services at a western metropolitan VA reported a past injury resulting in loss of consciousness. A TBI prevalence of 47% is markedly high when compared with 12% reported in the general population.⁹ Although these results, in combination with past research on the negative sequelae associated with TBI,^10–12 clearly indicate that TBI is a significant health concern for homeless veterans, they do not initially appear to support the hypothesis that homeless veterans have a significantly higher prevalence of TBI than homeless civilians. However, because of the limited sensitivity and specificity of the TBI–4, the data obtained likely underestimated the true prevalence of TBI in this VA homeless population. Findings support comprehensive TBI assessment such as the OSU TBI–ID, as opposed to screening, for all veterans seeking homeless services.