Online screening assessment of lifetime exposure to traumatic brain injury: a pilot study of associations between exposure and health status

Background. This study aimed to conduct a pilot test of the feasibility and validity of administering an online screening measure of lifetime traumatic brain injury (TBI) exposure in Australia. Methods. One hundred and fifty six adults (aged 18–65 years) were recruited from the community via snowball sampling (convenience sample). A cross-sectional online survey was deployed that included the Ohio State University Traumatic Brain Injury Identification Method (OSU TBI-ID) short form. Secondary measures assessed post-concussion symptoms (the Rivermead Post-Concussion Symptoms Questionnaire), negative affective states (depression, anxiety, and stress scales), and behavioural health risks (items from the 2014 Ohio Behavioural Risk Factor Surveillance System). Results. Online OSU TBI-ID feasibility was high (minimal missing data and attrition, high completion rate). A TBI history was successfully recorded by approximately 60% of participants. Validity testing, via investigation of expected associations with risk factors controlled, found that selected indices [Worst TBI, Multiple TBIs] were positively associated with worse post-concussion symptoms: P ’s < 0.05, small–medium effects. Worst TBI was significantly related to one behavioural health risk, smoking. There were no other significant correlations between online OSU TBI-ID indices and secondary outcomes when accounting for covariates ( P ’s > 0.05). Conclusion. Initial support was found for the feasibility and validity of an online screening measure of lifetime TBI exposure (LTE) in an Australian sample. Cautious interpretation is warranted because of study limitations, especially the small unrepresentative sample. Further studies could increase confidence in the feasibility and validity of online LTE screening.


Introduction
Traumatic brain injury (TBI) is associated with lifelong consequences and is considered the leading cause of death and disability worldwide (Langlois et al. 2006).Even 'mild' TBI (mTBI) has been linked to long-term complications and negative consequences (Langlois et al. 2006).Predicting outcomes from one, let alone multiple TBIs, is challenging (Dennis et al. 2022).Further, cumulative TBI exposure is a suggested mechanism or risk factor for late life health consequences, including chronic traumatic encephalopathy (Dams-O'Connor et al. 2013).The very strong potential for serious long-term negative consequences from multiple TBIs has inspired discussion about the best methods for measuring lifetime TBI exposure (LTE; Dennis et al. 2022).
There are well-documented challenges in single TBI assessment and additional issues for lifetime assessment.These issues include weak consensus for defining TBI (Menon et al. 2010); reliance on symptom-based criteria (Dennis et al. 2022); selection biases in TBI research due to unrecognised, unreported, or untreated injuries (Dams-OʼConnor et al. 2014); and parameter variation (e.g.selection criteria; Corrigan et al. 2003).TBI effects can be masked, misattributed, or otherwise altered by numerous factors, such as incentives to over- (Jurick et al. 2016) or under-report symptoms (Meier et al. 2015), use of interviews vs questionnaires (Edmed and Sullivan 2014), and cognitive distortions (Voormolen et al. 2020).For LTE, there is also the challenge of retrospective assessment over an extended period, including childhood (McKinlay et al. 2016).
Despite the challenges, several LTE assessment methods have been proposed.These methods range from a single item (e.g.'How many times, if ever…have you had… [a] head injury?')(Schofield et al. 2006b;Ilie et al. 2018); small-item sets (McKinlay et al. 2016); and structured, multi-item methods mimicking the clinical interview (e.g.Brain Injury Screening Questionnaire, Dams-OʼConnor et al. 2014); Ohio State University Traumatic Brain Injury Identification Method (OSU TBI-ID), Corrigan and Bogner 2007).A prior study (Dams-OʼConnor et al. 2014) reported that single item methods can miss >35% of cases identified by multi-item methods.Another study found that 15% of older adults (n = 559) screened positive for LTE using a multi-item method, about 80% of whom were previously unidentified (Schneider-Cline et al. 2019).Taken together, it seems that multi-item methods will be vital to advance understanding of LTE correlates for all TBIs (mild-severe).
The OSU TBI-ID short form is emerging as a leading LTE measure.This well-validated tool systematically elicits TBI history via a structured interview (Corrigan and Bogner 2007;Corrigan et al. 2013;McKinlay et al. 2017).The OSU TBI-ID can quantify TBI history in several ways via multiple indices, for example via a count of TBI exposures irrespective of severity.The OSU TBI-ID has been adopted in multi-centre, longitudinal TBI studies (Corrigan et al. 2013;Dams-O'Connor et al. 2013;Rabinowitz et al. 2020;Ohio State University Wexner Medical Centre 2022), and its usage is recommended by leading agencies (e.g. the National Institute of Neurological Disorder and Stroke 2019).
The findings from LTE studies that have used the OSU TBI-ID typically show that LTE is associated with increased behavioural health risks (BHRs), such as substance misuse, and poorer long-term health from a heightened risk of other conditions (Corrigan et al. 2013;Manchester et al. 2020).However, it remains unclear which indices are critical and whether increased BHRs are expected from exposure to any TBI or specific TBIs (e.g.severe injury or mild injury with loss of consciousness [LOC]; Rabinowitz et al. 2020).A recent study found that LTE metrics -with LOC--were associated with increased risks, such as heavy drinking and poor general mental health (Feiss et al. 2022).Further studies must clarify the relation between LTE and BHRs, including which injuries (and by extension which OSU TBI-ID indices) show these relations.
A potential limiting factor for LTE research is that the OSU TBI-ID was designed as an in-person interview (Lequerica et al. 2018).To address this issue, a handful of studies have deployed other OSU TBI-ID formats, including a computerassisted, telephone-administered version (Cuthbert et al. 2016) and online versions (Lequerica et al. 2018;Lequerica et al. 2021).These new formats could increase the ease of data collection from participants in distributed geographic regions (Lequerica et al. 2018).However, changing the OSU TBI-ID to other formats could introduce additional challenges.The interview can support conditional (followup/clarification) questions -a feature considered by some as critical for success (Ohio State University Wexner Medical Centre 2022) -and online surveys have specific biases (e.g. they may be susceptible to participant inattention and other data integrity threats; Oppenheimer et al. 2009;McKibben and Silvia 2017;Griffin et al. 2021).
To date, there have been a few studies on the feasibility and validity of an online form of the OSU TBI-ID (O-OSU TBI-ID).These studies, mostly from North America, have evaluated the completion rate (Lequerica et al. 2018(Lequerica et al. , 2021) ) and replicated expected associations (e.g. between LTE and postconcussion symptoms (PCS); Lequerica et al. 2018).In general, the findings have been favourable (Lequerica et al. 2018), but none have included the data integrity checks now commonly deployed in online surveys (Griffin et al. 2021).
In Australia, there is a paucity of research on LTE and its correlates.There are three prior studies with selected (criminal justice [CJ]; Perkes et al. 2011;Moore et al. 2014) or unselected (community) samples (Butterworth et al. 2004) and two with adults (cf.Moore et al. 2014).One study surveyed 200 men in the CJ system (Schofield et al. 2006a(Schofield et al. , 2006b) and a comparative sample (200 community-dwelling men; Perkes et al. 2011).Using a single-item LTE and followup questions for ≥five incidents, the study found a high prevalence of LTE with LOC (65% CJ vs 32% community) and all TBI exposures (82% CJ vs 72% community) as well as increased risk of neuropsychiatric sequalae when TBI exposed (CJ sample).Another cross-sectional survey (Anstey et al. 2004;Butterworth et al. 2004) in ~8000 randomly selected community members used a single-item LTE measure: '…ever had a serious head injury where you became unconscious for >15 minutes?'.Consistent with general population studies outside of Australia, LTE with LOC was associated with poorer general physical and mental health on screening tests (Anstey et al. 2004;Butterworth et al. 2004).Given that the Australian data are almost 20 years old and not obtained using a multiitem method or a current definition for 'serious' TBI, there is merit in revisiting this issue, including the methodology for LTE screening.
This study had two aims.First, to conduct an initial pilot investigation of O-OSU TBI-ID as a potential tool for LTE measurement in Australia via a small-scale replication and extension of the O-OSU TBI-ID feasibility and validity study by Lequerica et al. (2018).Thus, it was expected that the O-OSU TBI-ID would be feasible (i.e.data captured in all sections) and there would be a positive association between Worst TBI and PCS (Lequerica et al. 2018).The extension added data integrity checks and examined relations between other LTE indices (e.g.Multiple TBIs) and correlates (i.e.BHRs, PCS).

Method Participants
Participants were recruited at a major metropolitan university, including via a for-credit, university-hosted research participation pool (SONA), and externally via electronic communication within the researchers' networks.A snowball method was employed, whereby the recipients of study information were encouraged to share it with their networks.Eligible participants were people aged ≥18 years with a valid protocol.

Measures Single-item TBI (SI-TBI) history
Based on prior research (Butterworth et al. 2004;Ilie et al. 2015;Osborn et al. 2018), a SI-TBI history was assessed: 'Have you ever had a Traumatic Brain Injury (TBI) also referred to as a head injury?'.

Ohio State University TBI identification method short form (OSU TBI-ID; Ohio State University Wexner Medical Centre 2022)
The OSU TBI-ID is a 3-5 min structured (three-step) tool to determine LTE (Corrigan and Bogner 2007).The interview form has strong inter-rater reliability and acceptable test/retest reliability (Corrigan and Bogner 2007).
Step 1 assesses LTE and injury cause (five yes/no questions; e.g.'In your lifetime, have you ever injured your head or neck in a car accident or…crashing…a bicycle…?'A yes at Step 1 triggers deployment of Step 2 (event details; e.g.'Were you knocked out…?').All participants complete Step 3 about a history of multiple TBIs.Three summary indices of LTE were used in this study: Worst TBI [WTBI]; Multiple TBIs [MultiTBI]; TBI number by severity [TBINS] (Corrigan and Bogner 2007).

Rivermead Post-Concussion Symptoms Questionnaire (RPQ; King, et al. 1995)
The RPQ assesses the frequency and severity of 16 PCS.The RPQ has good psychometric properties (King et al. 1995).An example RPQ item is: '…Do you now (i.e. over the last 24 hours) suffer from headaches?' Participants indicate symptom frequency and severity (0 = not experienced at all to 4 = a severe problem).The wording of the introduction to the RPQ was altered slightly to ensure participants without a TBI history would also record their experience of these symptoms.The total score range is 0-64, with higher scores indicating greater problems (King et al. 1995).In this study the RPQ was found to be reliable (α = 0.96).

The depression, anxiety, and stress scales (DASS; Lovibond and Lovibond 1995)
The DASS is a 42-item self-report measure of three negative emotional states with good to excellent internal consistency (Lovibond and Lovibond 1995).An example item is, 'I felt that I had nothing to look forward to.' Participants rate each item for personal applicability over the past week (0 = did not apply… to 3 = applied…very much, or most of the time).Prior OSU TBI-ID studies have used the Beck Depression Inventory (e.g.Konrad et al. 2011); however, since the DASS was developed and normed in Australia, it was chosen for this study.The DASS subscales are scored by summing the responses (total score range = 0-42).Higher scores indicate a more negative emotional state (Lovibond and Lovibond 1995).In this study, the Cronbach's alpha for each DASS subscale was >0.80.

Behavioural Risk Factor Surveillance System (BRFSS; Centres for Disease Control and Prevention 2014)
The BRFSS is a telephone survey of BHRs and chronic health conditions (Centres for Disease Control and Prevention 2014).Several studies (Bogner et al. 2020;Manchester et al. 2020;Feiss et al. 2022) have used specific BRFSS items to examine relationships between TBI exposure and BHRs (e.g.substance use, depression, and days of poor mental health; Bogner et al. 2020).The current study used selected 2014 BRFSS items from the following sections: Health Status, Healthy Days, Health-Related Quality of Life (HRQoL), Tobacco Use, Marijuana Use, and Alcohol Consumption, since they were previously used in LTE studies.An example HRQoL item is, 'How many days in the past 30 days was your physical health…not good?'.BRFSS items are individually interpreted.Higher scores on the BHR items indicate increased health risk.

Procedure
Project applications were lodged, reviewed, and approved by the Queensland University of Technology (ethical clearance no: 4145; health and safety clearance no: 1782).The survey was hosted on the online platform Qualtrics™ (Provo, UT) and opened from October 2021 to May 2022.An opt-in method (e.g.activating a link) provided access to study information (e.g.risks, benefits).Consent was probed with a forced-choice item (agree [to continue] or disagree [to exit]).Continuing participants completed a validity check (reCAPTCHA plugin 1 ) and the SI-TBI, O-OSU TBI-ID, RPQ, DASS, and BRFSS.Skip logic was programed for conditional questions, including O-OSU TBI-ID Step 2. The duration of LOC was presented with three options (i.e.no LOC, <30 min-24 h, and >24 h).The participants quit the study

Data analysis
The

Data screening
Nine participants were excluded because of extreme missing data, an invalid protocol (failed integrity check), or an implausible survey completion time of <3 min (a value we could only obtain with indiscriminate responding).One underage (17 years) and one older participant (outlier aged >95% trimmed sample mean age) were also excluded.The selection of participants is shown in Fig. 1.The completion rate for the OSU TBI-ID was examined with controls applied (e.g.where skip logic was used).Questions about TBI from all causes were answered (see Fig. 1).The average survey completion time (OSU TBI-ID plus other measures) was approximately 148 min (5% trimmed mean, range ~4-5398 min).The lengthy completions are likely due to the failure to close the survey browser.

Discussion
This study had two aims.The primary aim was to conduct an initial pilot investigation of O-OSU TBI-ID as a potential tool for LTE screening in Australia.This was done via a smallscale replication and extension of Lequerica et al. (2018).We also performed an extension of Lequerica et al. (2018) by adding data integrity checks and examining novel relations between LTE indices (e.g.MultiTBIs) and the outcomes used in other LTE studies (e.g.BHRs).In broad terms, this study found initial support for the O-OSU TBI-ID.
We examined the feasibility of the O-OSU TBI-ID via the inspection of completion rates as per Lequerica et al. (2018).The O-OSU TBI-ID prompted participants to recall and record details about TBIs from all causes.Further, a data verification check was trialled with no issues reported.Although this study could not determine if this inclusion improved the data integrity, it did prevent bot completion.Adding this type of feature in future O-OSU TBI-ID applications would reflect best practice in online data collection (McKibben and Silvia 2017;Griffin et al. 2021).
We also tested for known associations between LTE indices and correlates, such as PCS.We found a positive association between WTBI and PCS, as previously demonstrated (Lequerica et al. 2018).Compared to people with no history of TBI, when covariates were controlled (age, sex, at risk status) a history of mild-moderate TBI (with LOC) was associated with worse PCS.Further, in the extension with the original controls (Lequerica et al. 2018), a lifetime history of TBI (2+ TBIs) was associated with worse PCS.However, lifetime TBINS was not significantly related to PCS, and no indices were significantly related to negative emotional states (depression, anxiety, or stress).A history of mild-moderate TBI (with LOC) (WTBI) was associated with one BHR (smoking).This analysis reveals further support for selected O-OSU TBI-ID indices (WTBI, MultiTBI), since they were correlated as expected with PCS.
To our knowledge, this study was the first deployment of the OSU TBI-ID in any format in Australia.We showed that about 1 in 10 people (13.5%) experienced at least one TBI with LOC (up to 24 h) in their lifetime.Prior North American OSU TBI-ID studies in larger samples have reported LTE rates that are higher (1 in 5 adults;  Our rate is higher than the prior Australian study with community members (i.e. between 5 and 6% prevalence; Anstey et al. 2004;Butterworth et al. 2004) and lower than the selected men-only sample (~32%; Perkes et al. 2011)  compare the LTE estimate from an online-vs interviewadministration of the OSU TBI-ID, and the self-reported LTE estimate could also be cross-checked with LTE data from another source (e.g.medical history, if available).These comparisons are suggested for future research.The LTE data are self-reported, conferring a risk of error and inaccuracy (e.g.recall bias, uncorroborated injury).Some measures referred to 'head injury' when the intent was to measure TBI, including our single-item measure and parts of the original OSU-TBI.The wording of our single-item measure could have been confusing for some participants, and this may have reduced the validity of some measures.This study used an online version of the BRFSS, which to our knowledge has not been tried before.This change in the BRFSS administration format from the original telephone interview could affect our results.Cross-sectional studies cannot answer questions of causation; thus, the findings do not show that LTE caused poorer mental health, only that it was associated with it.Some O-OSU TBI-ID scores will differ from those in other studies (i.e.we could not distinguish between mild vs a moderate TBI (with LOC) because of administrative error; and our WTBI measure did include severe TBI).The small sample size and composition (largely female university students) is a limitation, and we excluded a handful of responses as 'invalid attempts' (e.g.extremely quick completions or mostly blank protocols), whereas these responses could in fact signal a problem with feasibility (i.e. that the measure could not actually be completed by all of the participants).Whilst the sample may be sufficient for an initial Australian pilot test of the O-OSU TBI-ID, including its feasibility, it does not support generalisable conclusions about LTE nor does it show how the test would function in specific groups with a heightened TBI exposure (e.g.adolescent contact-sports players or unsteady, older persons).Future studies with larger, representative samples are now needed to determine if the O-OSU TBI-ID can be used as a population screening tool for LTE.This study suggests that estimates of LTE prevalence depend on assessment method (single-vs multi-item).Whichever method is used, the implementation should address the potential for expectancy bias or iatrogenesis (Dams-OʼConnor et al. 2014).A suggested strategy is to provide participants with information about this risk, including that taking the survey does not prove causation (Dams-OʼConnor et al. 2014).
There is strong interest in understanding LTE effects on health and functioning.The current study finds initial support for the feasibility of the O-OSU TBI-ID, including with added data validation, and in a new context (Australian sample).We successfully replicated some expected associations (Lequerica et al. 2018) and recommend further studies to identify key indices.However, the study limitations must also be kept in mind when considering the implications, including that this study was a small pilot investigation.The results of this pilot trial of this tool in Australia gives some confidence that the O-OSU TBI-ID could be successfully deployed in a larger population survey.In the longer term, this could eventually support improved LTE screening and understanding of LTE correlates, which in turn could improve resource planning and supports for people with TBI exposure.

Table 1 .
Average scores for RPQ, DASS, and BRFSS are shown in Table 2. Table 3 displays BHR frequency data by three measures of LTE.selection and study flow showing the participant exclusions (and reasons), completion of OSU TBI-ID sections, and study exit choices.
a = significant missing data, i.e., no response to almost all questions; 3.10% missing data); b = rushed response, survey duration <3 minutes (minimum pilot response time); c = study summary 16.6%, 2022; d = reimbursement with either course credit (n = 81) or entry (n = 62) for a prize draw 91.6%, 2022.Sixty-two participants entered the prize draw, which was drawn on 20 June 2022; 81 participants were awarded course credit (29 April 2022), and the study summary was provided to twenty-six participants (20 June 2022).

Table 1 .
(Continued) A Other, gender, n = 3; Other, primary occupation, n = 4. B 3-year degree n = 21; 4-year degree n = 17.C Master degree, n = 8; PhD/Doctorate, n = 2. D SI-TBI item: 'Have you ever had a Traumatic Brain Injury also referred to as a head injury?'.E Combat item: '… have you been nearby when an explosion or blast occurred?If you served in the military, think about any combat-or trainingrelated incidents'; Violence item: '… have you ever injured your head or neck in a fight, from being hit by someone, or from being shaken violently?Have you ever been shot in the head?'.F Subgroup n varies as reported.

Table 2 .
Descriptive statistics for TBI correlates.Rivermead Post-concussion Symptoms Questionnaire.DASS = depression, anxiety, and stress scales.BRFSS = 2014 Behavioural Risk Factor Surveillance System.Item stem: *No. of days in the last 30 days….**How many times in the past 30 days did you have… ***Item response scale: 0 = not at all, 1 = some days, 2 = every day.

Table 3 .
Frequencies and percentages for the number of days with poor behavioural health outcomes by three indices of LTE (O-OSU TBI-ID).

Table 4 .
Griffin et al. 2021)e replicated, it could indicate that Australia has a higher potential burden from LTE than previously modelled.This could be due to changes in the injury rate, reporting practices, or a research artefact (ie.use of a multiitem vs single item LTE measure).This study has several limitations.Online surveys carry a specific risk of bias.This survey deployed bot detection; but other validation techniques should be trialled (e.g.'honey pots';Griffin et al. 2021).A true equivalence test would Full ANCOVA predicting RPQ total score in three models of LTE models (Model 1, Worst TBI; Model 2, Multiple TBIs; Model 3, TBI number by severity).Analysis of Covariance; RPQ, Rivermead Post-Concussion Symptoms Questionnaire; d.f., degrees of freedom; TBI, traumatic brain injury; Worst TBI and Multiple TBIs are index scores from the O-OSU TBI-ID = Online Ohio State University TBI Identification Method.

Table 5 .
Multivariable binomial logistic regressions of relations between Worst TBI, Multiple TBIs, TBI number by severity, and BHRs.