Differential Use of Central Outcomes Center to Maximize Follow-up in a Study of Adolescents Seeking Care in the Emergency Department

Between 25–50% of adolescents with mild traumatic brain injuries (mTBIs) have substantial mental health sequelae after their injuries, the most common being depression and anxiety^1–3. While previous studies have identified risk factors for mental health sequelae after mTBIs, there is still a need for a robust clinical prediction tool to risk-stratify adolescents from diverse sociodemographic backgrounds recovering from mTBI. Screening Emotions in Adolescents Receiving Care at the Hospital after mTBI (SEARCH-mTBI) is a multicenter prospective observational cohort study enrolling adolescents who present to the emergency department (ED) with mTBIs⁴. SEARCH-mTBI is designed to derive and validate a clinical prediction tool to identify adolescents at risk for new or worsening mental health concerns and is enrolling English- and Spanish-speaking participants.

To minimize attrition, SEARCH-mTBI uses a central outcomes center (COC) staffed by English- and Spanish-speaking bilingual research staff dedicated to individualized participant follow-up. Few studies have reported on methods to optimize follow-up in emergency department research⁵. Our group has previously reported on the utility of the COC in promoting participant retention and reducing attrition⁶. We also found that caregiver-preferred language of Spanish was the factor most strongly associated with lower rates of compensation redemption, which may place these participants at risk for differential attrition⁷. Differences in attrition among participants from underrepresented groups, such as those whose caregivers prefer Spanish, can introduce systematic bias and result in a prediction tool that is not broadly applicable. However, differential use of COC support (e.g., telephone, text, or email reminders) among participants with completed follow-up data, stratified by caregiver language preference, has not been evaluated.

The aim of this study was to evaluate the differential utility of COC support to facilitate completion of adolescent-reported questionnaires required to detect new or worsening mental health concerns (the primary outcome for SEARCH-mTBI) according to caregiver-preferred language (English versus Spanish). A secondary aim was to identify key participant- and study-level variables associated with receipt of COC support.

We conducted a secondary analysis of the ongoing SEARCH-mTBI study (detailed study methods previously published)⁴. The SEARCH-mTBI clinical prediction tool will be derived based on detection of new or worsening mental health, measured by a clinically meaningful change on either the Generalized Anxiety Disorder-7 (GAD-7)⁸ or the Patient Health Questionnaire-8 (PHQ-8)⁹ completed by adolescents at 1 or 3 months post-injury relative to their pre-injury baselines⁴. Our study included adolescents (ages 11–17 years) enrolled in SEARCH-mTBI between 5/2024 and 2/2025 from four pediatric EDs in the United States who had complete PHQ-8 and GAD-7 questionnaires at baseline and 1 or 3 months.

Demographic data (e.g., age, sex) are obtained via the electronic medical record. Area Deprivation Index (ADI) is calculated based on home address census block data¹⁰. Language preference, child race/ethnicity, health insurance status, and any exposure to adverse childhood experience(s) (ACEs) are obtained via caregiver report (baseline survey). Race/ethnicity is included given known disparities in mTBI presentation and management¹¹. SEARCH-mTBI is approved by University of Utah’s Institutional Review Board.

Adolescent/caregiver dyads complete surveys at baseline (pre-injury functioning) in their enrolling ED or soon after, then 1 and 3 months post-enrollment. Caregivers who elect to complete follow-up surveys electronically receive up to three automated parent and child survey links/reminders via text/email. Adolescents do not directly receive automated survey reminders for their survey. Caregivers and/or adolescents who do not complete follow-up surveys independently after receiving the survey links, and those who prefer telephone follow-up, receive individualized COC contact via telephone, text, and/or email in their preferred language, categorized as English-preferred (EP) or Spanish-preferred (SP). The COC uses all available contacts, including adolescents’ direct cell phone numbers and email addresses.

We calculated descriptive statistics with R (RStudio version 2023.06.1+524; Posit Software, PBC, Boston, MA), including means and standard deviations for continuous variables (age and ADI) and frequencies and percentages for categorical variables (sex, language preference, health insurance, race/ethnicity, and exposure to ACEs). We explored differences in demographic variables between adolescents of EP and SP caregivers using t-tests for continuous variables (age and ADI) as well as two-proportion z-tests (language preference), chi-square tests (sex), or Fisher’s exact tests for categorical variables (race/ethnicity, health insurance, and ACEs). We performed a multivariable binary logistic regression analysis (using SPSS v.31.0.0) to predict survey completion with or without COC support based on caregiver language preference while controlling for other demographic characteristics. We excluded participants with missing data from this analysis.

Of the 407 enrolled adolescents, 340 completed the GAD-7 and PHQ-8 at baseline and 1 or 3 months after the index ED visit. Of those with completed survey data, 286 (84%) were adolescents of EP caregivers and 54 (16%) were adolescents of SP caregivers. A higher proportion of adolescents of SP caregivers required COC support than adolescents of EP caregivers (85% [46/54] vs. 60% [171/286], p < 0.001).

Adolescents of EP and SP caregivers did not differ in age (EP: mean=14.5 years, SD=1.9; SP: mean=14.6 years, SD=1.8), sex (EP=64% male; SP=63% male) or presence of ACEs (EP: Yes=14% vs. No=85%/No response=1%; SP: Yes=13% vs. No=83%/No response=4%). Significant differences were noted in caregiver-reported race/ethnicity (EP: non-Hispanic white=37%, non-Hispanic black=20%, non-Hispanic other/multiracial=11%, Hispanic=31%, no response=0.3%; SP: non-Hispanic white=2%, non-Hispanic black=2%, non-Hispanic other/multiracial=0%, Hispanic=94%, no response=2%, p<.001). There were also significant differences in insurance status (EP: Private=49%, Public=41%, No Insurance=7%, Unsure =2%, No response=1%; SP: Private=4%, Public=54%, No Insurance=19%, Unsure=19%, No response=6% p<.001) and ADI (EP (n=262): mean=49.84, SD=27.93; SP (n=51): mean=68.14, SD=22.77, p<.001).

Results of the multivariable binary logistic regression are presented in the Table. Adolescents of SP caregivers [OR=6.18 (95% CI:1.99, 19.15), p <.01] were more likely to require COC support than those of EP caregivers while controlling for other demographic characteristics. Older adolescent age [OR=1.2 (95% CI:1.05, 1.39), p <.01] was also found to increase the likelihood for requiring COC support. Sex, ADI, health insurance, race/ethnicity, and exposure to ACEs were not significantly associated with COC support.

Table.

Multivariable Logistic Regression Model of Predictors of Central Outcomes Center Support

Covariate	N	Odds Ratio	95% C.I.	P value
Sex				.78
Male	199	(Ref)
Female	104	.93	(.54, 1.58)
Age	303	1.21	(1.05, 1.39)	.008
ADIa	303	1.00	(.99, 1.01)	1.00
Caregiver Language Preference				.002
English	256	(Ref)
Spanish	47	6.18	(1.99, 19.15)
Health Insurance				.20
Private	132	(Ref)
Public	131	.89	(.48, 1.66)
No Insurance	26	.53	(.19, 1.44)
Unsure	14	.23	(.05,1.03)
Race & Ethnicity				.07
Hispanic	123	(Ref)
White, non-Hispanic	100	.52	(.27, 1.01)
Black, non-Hispanic	52	1.23	(.55, 2.77)
Other/Multiple, non-Hispanic	28	.42	(.17, 1.07)
ACEsb				.61
No	264	(Ref)
Yes	39	.82	(.39, 1.75)

^aNational Area Deprivation Index

^bAdverse Childhood Experiences.

In this study, we evaluated the differential utility of COC support to facilitate completion of adolescent-reported questionnaires. While adolescents of EP and SP caregivers differed on several key demographics, we found that, when controlling for these variables, caregiver-preferred language of Spanish was the factor most strongly associated with use of COC support for adolescent survey completion. Older adolescents overall were also more likely to utilize COC support. In contrast, sex, ADI, health insurance, race/ethnicity, and exposure to ACEs were not significantly associated with COC support.

To date, there is limited research on the retention of Spanish-speaking participants to reduce attrition bias in longitudinal pediatric studies. Our group previously found that COC support can be a useful and cost-effective model to promote retention and reduce attrition in a subset of participants enrolled in SEARCH-mTBI⁶. This study extends these findings by highlighting the differential utility of COC support to promote GAD-7 and PHQ-8 completion by adolescents of SP caregivers.

The finding that adolescents of SP caregivers are more likely to utilize COC support has important implications for research design. While study surveys and automated messages are sent in Spanish to SP caregivers, it is possible that confidence with online survey administration, access to online devices/data, and/or literacy skills may differ between EP and SP caregivers and/or adolescents, potentially accounting for increased support to maximize follow-up rates. Furthermore, our previous findings that caregivers who prefer Spanish are less likely to redeem their compensation earned for completed caregiver and adolescent surveys⁷ suggest another possible explanation that SP caregivers could be less incentivized to share the adolescent survey links with the participants because they may not be accessing their compensation following survey completion. Future research should investigate unique barriers experienced by Spanish-speaking families that contribute to our findings.

We also found that older adolescents, in general, were more likely to use COC support. Given that the automated survey links are shared with the child by the caregiver, caregivers may provide more scaffolding and support to younger versus older adolescents to ensure survey completion. If survey links had been delivered directly to older adolescents, it is possible that more would have completed surveys independently. That said, older adolescents may have greater demands on their time from academic, extracurricular, and social activities, which may interfere with independent survey completion. The benefit of sending survey links directly to adolescent contact information, especially when enrolling older adolescents with their own phones, needs further exploration.

Our study has several limitations. Our findings may be unique to children evaluated in the ED after mTBI and may not be generalizable to other ED populations. Additionally, adolescents of SP caregivers was a small portion of our sample, only 16%, limiting generalizability of these findings to Spanish-speaking populations more broadly. Furthermore, key variables that may be barriers to online survey completion such as online competence, equipment availability, and literacy were not evaluated.

In conclusion, COC support is useful to promote study retention among adolescent participants with mTBIs, especially those whose caregivers prefer Spanish. Researchers conducting longitudinal studies should consider incorporating COC support to maximize retention and minimize attrition bias, especially when enrolling caregivers who prefer Spanish and older adolescents in general.