Adverse childhood experiences (ACEs)—preventable, potentially traumatic events, such as violence, abuse, and neglect, that occur during childhood—are associated with increased risk of adverse health and social consequences across the lifespan.^1–5 ACEs can have a cumulative effect, with multiple adverse experiences putting individuals at greater risk for poor health outcomes.^1,3,6,7 In the U.S., ACEs are common, with nearly 2 in 3 adults reporting at least 1 ACE and 1 in 6 adults reporting 4 or more ACEs.⁸ There are disparities in ACEs, with some groups experiencing higher rates.^3,8,9

Surveys and surveillance systems have been essential in assessing the magnitude of ACEs. The original Centers for Disease Control and Prevention (CDC)-Kaiser ACE study established ACEs as a concept, defined measurement, and examined the association between ACEs and health outcomes; subsequently, many instruments used in ACEs research and surveillance systems to date have been adapted from this study.^10–12 One such surveillance system is the Behavioral Risk Factor Surveillance System (BRFSS)’s ACEs module. Since 2009, the BRFSS ACEs Module has provided key data to measure the prevalence of ACEs among the U.S. adult population, to understand the relationship between ACEs and health consequences and behaviors later in life, and to inform the development of ACEs prevention strategies.¹⁰

BRFSS is a cross-sectional state-based telephone survey that collects data in all U.S. states, the District of Columbia, and 3 U.S. territories. BRFSS collects a range of self-reported information from non-institutionalized U.S. adults aged ≥18 years regarding chronic health conditions, health behaviors, and the use of preventive services.¹³ The BRFSS ACEs Module collects information on an individual’s exposure to abuse (emotional, physical, or sexual), neglect (emotional or physical), and household challenges (witnessing intimate partner violence, substance abuse in the household, mental illness in the household, or parental separation or divorce) during childhood (before the age of 18 years).¹⁴ In BRFSS, ACE questions were added either as an optional CDC module (2011–2012 and 2019 and onward) or as a state-added module (2013–2018).⁸ As a CDC-sponsored optional module, ACE questions are standardized, and states can choose to implement the entire module without modification. In state-added modules, by contrast, states can change the number, composition, and the language of ACE questions. Since 2009, all 50 states and the District of Columbia have included the ACEs module in their survey at least once, as either a CDC-sponsored optional module or state-added module, with many states including it for multiple years.¹⁰

Although the BRFSS ACEs data provide a wealth of information across states related to ACEs, declining response rates to BRFSS (e.g., 2016 median response rate: 47.1%; 2021 median response rate: 44.0%) may have implications for ACEs data.^15,16 In addition, item nonresponse (nonresponse to particular questions) can have implications for data quality, especially given the potential for nonresponse bias.^17–19 Furthermore, nonresponse can impact the representativeness of data and estimates of the health behavior or topic assessed.¹⁷ Previous studies examining nonresponse within BRFSS, including optional modules, have found differences in sociodemographic characteristics between responders and nonresponders.^18,20–24 In an analysis of individuals who declined to answer questions in the BRFSS Firearm Safety Module, men (compared with women) and White respondents (compared with Hispanic respondents) were more likely to refuse firearm-related questions.²¹ In studies examining nonresponse to the BRFSS ACEs Module, there were significant demographic differences between responders and nonresponders, with some groups (e.g., non-Hispanic Blacks, Asians, rural residents, those with riskier health behaviors, individuals with lower income) more likely to be nonresponders.^22,24 These studies underline that the underrepresentation of certain groups may bias estimates of the ACEs burden among U.S. adults and, thus, have important implications for the response activities to address the lasting impacts of ACEs.^22,24

To further understand nonresponse to the ACEs module to improve measurement of ACEs and related outcomes, this study builds upon previous research to understand sociodemographic characteristics of the participants across multiple states who did and did not respond to the ACEs module and to assess whether there were differences by year and question.

The analytic sample size was 126,753 in 2019 and 92,238 in 2021. An estimated 1.2% (95% CI=1.1, 1.4) and 2.4% (95% CI=2.2, 2.5) of BRFSS participants were nonresponders to the ACEs module in 2019 and 2021, respectively, a statistically significant difference (p<.0001) (Table 1). In 2019, nonresponders were more likely to be aged 18–24 years (16.0% vs 11.6%, p=0.03), be non-Hispanic Black (18.1% vs 12.7%, p=0.01), and be non-Hispanic Asian (5.9% vs 2.8%, p=0.01); have less than a high school education (20.4% vs 11.7%, p<0.01); have income <$15,000 (17.0% vs 9.1%, p<0.01); be unemployed (53.0% vs 44.3%, p<0.01); or rate their health poor (9.7% vs 5.0%, p<0.01) than responders. Furthermore, in 2019, responders were significantly more likely to take a survey with fewer than 15 modules (43.0%) than nonresponders (35.9%, p<0.01). In 2021, a significantly greater proportion of nonresponders were male (52.7% vs 47.9%, p=0.01), were aged 45–54 years (20.1% vs 15.2%, p<0.01), were non-Hispanic Black (16.2% vs 13.0%, p=0.04), had less than a high school education (13.4% vs 10.6%, p=0.04), had an income of $25,000 −34,999 (17.3% vs 12.6%, p<0.01), or had an income of $35,000–49,999 (17.7% vs 13.4%, p=0.01) than respondents.

For both years, the patterns in the type of responder and nonresponder were similar, with most individuals—96.3% in 2019 and 94.5% in 2021—being complete responders to the ACEs module (Table 2). However, there was a significant increase in nonresponse from 2019 to 2021, with individuals more likely to be partial (2019, 0.5%; 2021, 1.0%; p<0.001) or complete (2019, 0.7%; 2021, 1.4%; p<0.001) nonresponders in 2021 than in 2019. Furthermore, when comparing module questions, nonresponse increased as the module progressed (Figure 2). Although the percentage of nonresponders by question was higher in 2021 (ranged from 2.0% to 3.8%) than in 2019 (ranged from 0.9% to 2.5%), for both years, the nonresponses were highest for sexual abuse questions, which occur toward the end of the module (Appendix Table 2).

There was variability in nonresponse across states, ranging from 0.5% in West Virginia to 8.1% in Nevada (Appendix Table 3, available online). When examining the characteristics of responders and nonresponders by state, in some instances (e.g., sex in Indiana), the differences exceeded 10 percentage points. At the same time, this absolute difference did not translate into statistically significant differences owing to small cell sizes for non-responders. For states that were included in both years, an estimated 1.3% (95% CI=1.1, 1.6) and 1.9% (95% CI=1.7, 2.1) of individuals were nonresponders to the ACEs module in 2019 and 2021, respectively. The proportion of participants who were break offs in response to ACE module questions was small (Appendix Table 4, available online). However, 2 questions on sexual abuse had the highest percentage of break offs (touched sexually: 0.5%; forced to have sex: 0.2%).

Across states and years, nonresponse for the BRFSS ACEs Module was low, at both the item and module level, suggesting that adults are willing to respond to questions assessing ACEs, consistent with findings of previous studies.^22,26,27 Furthermore, individuals were most likely to respond to the complete ACEs module rather than partially responding or not responding to the module, pointing to the acceptability of the module questions among BRFSS participants. Notably, from 2019 to 2021, the percentage of nonresponders approximately doubled, which may reflect overall declines in response to the BRFSS survey.

There are important sociodemographic differences between responders and nonresponders. Although the BRFSS weighting helps to adjust for nonrepresentative participation in the survey overall, the weighting does not necessarily adjust for item nonresponse and break offs for optional modules.²³ For both years, participants who were non-Hispanic Black and had less than a high school education were significantly more likely to be nonresponders, which is consistent with the finding of Crouch et al.²² The BRFSS ACEs Module was adapted from the original CDC-Kaiser ACE study, which assessed ACEs most relevant to a White, middle-class community.^10,11 Participants could have chosen not to respond to specific questions because they may have felt that it was irrelevant,²⁸ or they may have been more likely to experience ACEs and upset by the sensitive questions. The inclusion of expanded ACE questions,²⁹ which capture additional constructs of childhood adversity, may have more relevancy to diverse populations and potentially increase response. Furthermore, in 2021, nonresponders were more likely to be male than female. The sensitivity of the questions, particularly related to sexual abuse, could have potentially led to item nonresponse among males.^12,30,31 In 2019, individuals who rated their health poor were almost twice as likely to be nonresponders as to be responders, indicating a possible self-selection bias among people with poor health. Grigsby and colleagues²⁴ also found that individuals rating their health fair or poor and those with different health conditions (e.g., depression, diabetes) were more likely to decline to respond to ACE items. Given that ACEs are a risk factor for poor health outcomes, higher nonresponse to ACE questions among individuals rating their health as poor may result in underestimation of the ACEs burden.¹ In 2021, nonresponders were older, had higher education and income, and were more likely to be employed than in 2019. Future research may be warranted to investigate the changing demographic patterns of nonresponders to determine whether this is consistent across all ACE questions and other questionnaire modules and how these demographic patterns changed in the context of the COVID-19 pandemic.

When examining nonresponse by question, for both years, nonresponse increased as the module progressed, resulting in sexual abuse questions, which are located near the end of the survey, having the highest nonresponse. The sensitivity of these questions may have further contributed to item nonresponse, as indicated by higher break-off rates and corroborated by previous studies.^32,33 In addition, in 2019, a significantly greater proportion of participants who took a shorter BRFSS survey (<15 modules) were more likely to be responders. However, it is unclear whether the comparatively high levels of nonresponse and break offs for these questions are due to their sensitivity or being later in the survey or both. Previous studies examining break off in BRFSS found that survey break off often occurred at the end of the survey after the core questionnaire.^17,34 These findings may suggest participant survey fatigue, which warrants additional investigation.

Assessing the magnitude of ACEs is essential for informing the development of ACEs prevention and response activities. For this study, 1 form of nonresponse to ACE questions was examined, item nonresponse, which could potentially lead to biased estimates. Although item nonresponse does not automatically result in biased estimates, strategies to address nonresponse in existing surveillance systems can improve estimates of ACEs burden. For instance, placing the ACEs module earlier in the BRFSS survey may improve response to ACE questions and address potential survey fatigue at the time they are asked. However, although cumulative break off during the ACEs module is low, it may be higher than that in other modules, which could have implications for other survey sections if the ACEs module is moved earlier. Furthermore, adding text to the beginning of the module to describe the purpose of sensitive questions may improve participants’ willingness to answer sensitive ACE questions. In addition, findings provide insight into understanding how to interpret missing data in the context of nonresponse and sociodemographic of nonresponse. Regarding unacceptability of questions, a qualitative study among nonrespondents could be conducted to examine why individuals are not responding to ACE questions or dropping off. Nevertheless, for researchers and public health professionals considering adding ACE questions to surveys and surveillance systems, findings indicate that respondents are willing to answer ACE questions, and asking such questions in surveys are likely to be acceptable to most survey participants. The favorable BRFSS response rates compared with those in similar surveys¹⁶ and the overall low nonresponse to the ACEs module highlight the usefulness of the BRFSS ACEs Module in providing information about ACEs in the U.S.

Although nonresponse to the ACEs module was overall low, there was an estimated overall increase in nonresponse from 2019 to 2021, with substantial differences across sociodemographic groups. Among ACE questions, the highest nonresponse was documented for sexual abuse questions, either hinting at sensitivity to those questions or indicating their proximity to the end of the module. Strategies to address nonresponse, for example, adding the ACEs module earlier in the BRFSS survey or as part of the core questionnaire, could improve the quality and representativeness of ACEs data. Overall, findings demonstrate that individuals are willing to respond to the ACEs module questions, highlighting the importance of the BRFSS ACEs module in measuring their burden in the U.S.