The Effect of ACEs on Health Outcomes: Latent Class Analysis

J.T. Fuller| June 29, 2022

A significant body of research has built upon the ACEs study, and some researchers have hypothesized the type of ACE is related to specific adverse outcomes. Kim et al. (2019) claim specific groups of ACEs are associated with differential risk profiles for depression leading to SUDs in older adults. Lee et al. (2020) claim specific groups of ACEs are associated with differential risk profiles relating to community violence (CV) and depression, anxiety and PTSD in adults. Haar-Pedersen et al. (2020) claim specific groups of ACEs are associated with differential risk profiles for anxiety, depression, PTSD, CPTSD, emotional wellbeing (SWB) and social outcomes based on sex. Barboza (2017) claims specific groups of ACEs are associated with differential risk profiles for depression, HIV risk-taking and problem drinking in adults. Li et al. (2021) claim specific groups of ACEs are associated with differential risk profiles for suicidal behavior based on sex in Chinese students. Shin et al. (2018) claim specific groups of ACEs with differential risk profiles are associated with psychological distress, alcohol-related problems, and tobacco use but not drug dependence in younger adults.

Literature Review

What is the relationship between specific ACE domain patterns and risks of adverse outcomes? In all contained studies, researchers studied the effect of ACEs, the independent variable, on various adverse health outcomes, the dependent variable, using latent class analysis (LCA). In the Kim et al. (2019) study, the data was expected to reveal specific groups of ACEs with differential risk profiles for depression leading to SUDs in older adults. The purpose of this study was to identify ACE domain patterns in older adults associated with depression and SUDs. Survey data was expected to reveal specific groups of ACEs with differential risk profiles for depression leading to SUDs. Data was collected from the 2012-2013 National Epidemiological Survey on Alcohol and Related Conditions III, collected randomly from non-institutionalized US residents over age 55 (N=11,386). The data reflected oversampling of minorities to improve minority estimate reliability. ACEs indicators were adapted for consistency with the CDC-Kaiser ACE Study, and economic adversity was also measured. SUDs were assessed with the Alcohol Use Disorder and Associated Disabilities Interview Schedule based on DSM-V criteria for determining SUDs. Control data included age, sex, household income, education, race, marital status, economic stress, and physical health. LCA resulted in four ACEs patterns identified as Class 1: High Adversity (CPA, CPN, CEA, CEN, DV, SUD); Class 2: low adversity; Class 3: Child Abuse (CPA, CEA); Class 4: Parental Substance Use (SUD). All classes had higher risk of SUD as compared to the low adversity class when compared directly. Classes 1 and 3 had a significantly higher probability of depression than classes 2 and 4; there was no statistical difference between the rates of depression in classes 2 and 4.

This study suggests an increased likelihood of a depression diagnosis for the High Adversity and Child Abuse groups as compared to the SUD and Low Adversity groups. The SUD group showed a higher risk of SUDs as compared to the Low Adversity group but no increased risk for depression. In this study, CSA, EA, MI, and IFM were not relevant to the classification patterns. Limitations of this study include higher risk of recollection distortion due to older participant age, uncontrolled variables related to comorbidities, and potential for demographic differences in SUD development. 

In the Lee et al. (2020) study, data was expected to reveal specific groups of ACEs with differential risk profiles relating to CV and depression, anxiety and PTSD in adults. The purpose of this study was to identify ACE domain patterns including exposure to CV associated with internalized disorders in adults. Survey data was expected to reveal patterns relating to CV and mental disorders. Data was collected from the National Longitudinal Study of Adolescent and Adult Health Waves I (1994-1995), III (2001-2001) and IV (2008-2009) (N=10,686). Participants were in grades 7-12 in 1994-1995 and selected using stratified multistage school-based cluster sampling. Incomplete surveys were omitted from the study. The 10 CDC-Kaiser ACE domains were included with this data except for DV. MI and drug abuse was reported by proxy through adult suicide attempt and parental alcoholism. CV was included, as was foster care experience. Adult mental health disorders were measured from self-reported data on diagnosis and age for depression, anxiety, and PTSD. Control data included age, gender, races, education, household income, homelessness, health insurance, marital status, general health, public assistance, depressive and anxiety symptoms. LCA resulted in four ACE clustering patterns identified as Class 1: Child Maltreatment (CPN, CPA, CEA; CEN and CSA were reported at much lower rates but included in this category); Class 2: Household Dysfunction (SUD, DP, IFM); Class 3: Community Violence (CVW, CVE); Class 4: Low Adversity.

Several patterns emerged from comparing latent classes, demographics, and mental disorders. Depression and anxiety levels were highest in Class 1. Class 2 was characterized by low SES. Class 3 was overrepresented by males and Hispanics. Class 4 was characterized by higher SES, education level, income, and insurance. Class 1 was associated with higher rates of depression, anxiety, and PTSD as compared to Class 2 or 4. Class 3 was associated with higher rates of PTSD as compared to Class 4, but not for depression or anxiety. The rates of diagnosis in Class 2 were not significantly different from Class 4. 

The researchers found CV reported more frequently than CEA, CEN and PD. Because the diagnosis rates in Class 2 were not statistically significant from the low adversity group, the researchers hypothesized more support may have been available to that group. The rate of PTSD in the CV group was similar to that in Class 1, which is consistent with experiencing toxic stress. The diagnostic rate for minorities was inconsistent with the reported ACEs, suggesting underdiagnosis consistent with other research. Data for this study did not allow for direct comparison with other similar studies since the ACEs survey was not used to collect the data, but rather the data were extrapolated from other sources. Important ACEs like DV and MI were not directly reported. Nonetheless, the high reporting of CV suggests more attention to this ACE is warranted. 

In the Haar-Pedersen et al. (2020) study, the data was expected to reveal distinct groups of ACEs with differential risk profiles for anxiety, depression, PTSD, CPTSD, emotional wellbeing (SWB) and social outcomes based on sex. The purpose of this study was to identify ACE domain pattern differences between males and females, and gender variations associated with adverse psychosocial and emotional outcomes. The data was expected to reveal distinct groups of ACEs with differential risk profiles for anxiety, depression, PTSD, CPTSD, emotional wellbeing (SWB) and social outcomes based on sex. Data was collected in 2017 from a nationally representative household sample of adults aged 18-70 currently residing in the US using online self-report questionnaires (N = 1,839). Only individuals reporting at least one traumatic experience were included in this study. The survey design included oversampling of females and minorities because of the increased likelihood of trauma-related distress, therefore the data was weighted to better represent the US population. This sample population may include bias because internet access was required to complete the survey. ACEs were measured using the 10-item CDC-Kaiser ACEs survey. Psychosocial emotional outcomes were assessed based on survey responses. Mental health domains assessed include depression, anxiety, PTSD and CPTSD. Depression was measured with the PHQ-8; anxiety was measured with the Generalized Anxiety Disorder 7-item Scale; PTSD was measured with the International Trauma Questionnaire (ITQ) which is based on the ICD-11 PTSD and CPTSD symptoms and relates experiences to their most traumatic experience. Emotional wellbeing was assessed using the World Health Organization Wellbeing Index (WHO-5) to measure psychological health and the De Jong Gierveld Loneliness Scale, used to measure social and emotional loneliness. Social wellbeing was measured by collecting data on relationship, education, employment, and income. LCA resulted in different ACE clustering patterns for each gender. For males, two patterns were identified as Class 1: Low Adversity; and Class 2: Mixed Adversity (high CEA, moderate CPA, DP, SUD, CEN, DV). For females, four patterns were identified as Class 1: High Adversity (high CEA, CPA, CSA, CEN, CPN, SUD, DV, DP); Class 2: Child Abuse and Neglect (high CEA, moderate CSA, CPA, CEN); Class 3: Dysfunctional Home (high DP, moderate SUD, MI, CSA); Class 4: Low Adversity. 

Several patterns emerged from comparing psychosocial emotional outcomes, latent classes, and sex. The rate of multiple ACEs for females in this study was 39% and 21% for males. Females were more likely than males to report CSA, CPN, CEN, SUD, and MI. Females reported significantly higher levels of PTSD, CPTSD, depression, anxiety, and loneliness, and lower SWB. Females were more likely to be unemployed and have lower income. Within classes, males in Class 2 (mixed adversity) had significantly higher levels of PTSD, CPTSD, depression, anxiety, loneliness and significantly lower SWB, lower education and income levels as compared to Class 1 (low adversity). Females in Class 1 (high adversity) had significantly higher scores across all symptom domains compared to all other classes. Classes 2 and 3 had scores higher than Class 1 (low adversity) across all symptom domains. There was no difference between symptom scores in classes 2 and 3. SWB scores decreased as adversity increased, with Class 4 having the highest scores, then 3, 2 and 1 having scores lower than the previous. Notably, Low income and unemployment rates were significantly higher in Class 4. Also notable was a significantly lower rate of committed relationships and low income for Class 3 (dysfunctional home). 

Consistent with expectations, female CSA reporting was higher than for males and female trauma profiles more complex, and could be related to higher rates of internalizing disorders. The researchers suggested there may be a relationship between household dysfunction and socioeconomic status, which is strongly associated with negative outcomes. The disparate rate in internalized disorders for females may be explained by higher rates of early life trauma, but the data for this study did not include details on ACE timing, severity, or intensity. Parental income and education status was also not accounted for, which may also be confounding variables. 

In the Barboza (2017) study, the data was expected to reveal specific groups of ACEs with differential risk profiles for depression, HIV risk-taking and problem drinking in adults. The purpose of this study was to identify ACE domain patterns associated with various psychosocial outcomes. Survey data was expected to reveal specific groups of ACEs with differential risk profiles for depressive symptoms, HIV risk-taking and problem drinking. Data was collected from the 2009-2012 BRFSS, collected from adults in 14 states (N=117,555). Incomplete surveys were excluded from the study. The BRFSS ACEs survey questions are consistent with the original 8-item CDC-Kaiser ACE study omitting CEN and CPN. This data may include bias related to partial state representation and only includes individuals with telephones. Psychosocial outcomes were assessed based on survey responses. Depressive symptoms were measured with the Patient Health Questionnaire (PHQ)-8, which is based on DSM-IV depression diagnosis criteria, and data on days of depressive symptoms in the prior 2 weeks were collected. Mental distress data was also collected. HIV risk-taking behavior was assessed based on whether, in the past year, the participant engaged in intravenous drug use, received treatment for a sexually transmitted disease, exchanged money or drugs for sex, or engaged in anal sex without using a condom; this data could not be disaggregated. Problem drinking was assessed based on frequency and quantity of consumption. Control data included age, gender, income, education, race, perceived life satisfaction, and social support. LCA resulted in five ACEs patterns identified as Class 1: Normative, Low Risk; Class 2: Emotionally Abusive Alcoholic Household (high CEA, SUD); Class 3: Emotionally Abusive Alcoholic Household With Parental Conflict (high DV, CEA, SUD, DP); Class 4: Sexual Abuse (high (CSA)); and Class 5: Highly Dysfunctional And Abusive (high DV, CEA, CSA, MI, SUD, DP). Several other factors were high within the groups, but there was a lack of homogeneity, particularly for CPA and IFM in classes 3 and 5. 

Several patterns emerged from comparing psychosocial outcomes and latent classes. Depressive symptoms were most prevalent in Class 5, and more prevalent in classes 2-4 as compared to low-risk Class 1. The depression risk between classes 2, 3 and 4 were indistinguishable. HIV risk-taking behavior was higher in high conflict classes 3 and 5. Surprisingly, having a history of sexual abuse was not an indicator of increased HIV risk in this study, which is inconsistent with other studies. Problem drinking behavior risk was high and indistinguishable for the two groups with alcoholic households, classes 2 and 3, which is consistent with research supporting a high risk of alcoholism for children of alcoholics. Problem drinking risk between classes 1, 4 and 5 were largely indistinguishable. The author of this study concluded evaluation of specific ACEs would potentially improve outcomes related to specific ACEs, particularly for children of incarcerated parents, and relying only on CR values could miss important nuances. Limitations on this study include self-reported data from those with telephones in specific states, who may not represent the general population and critical data on neglect was not collected. 

In the Li et al. (2021) study, the data was expected to reveal distinct groups of ACEs with differential risk profiles for suicidal behavior based on sex in Chinese students. The purpose of this study was to identify the ACE domain pattern differences between males and females associated with suicidal behavior in Chinese middle school students. Survey data was expected to reveal distinct groups of ACEs with differential risk profiles for suicidal behavior based on sex. Data was collected from a school-based health survey distributed to randomly selected schools in four Chinese provinces from 2017 to 2018 (N=14,500). Surveys more than 5% incomplete, fictitious or outside the age range were omitted from the study. Participants aged 10–20 years completed the Chinese version of the Childhood Trauma Questionnaire Short Form (CTQ-SF) questionnaires. Household dysfunction questions were adapted for this population based on the original CDC-Kaiser ACE questions. Suicidal behaviors were measured using the 2015 Youth Risk Behavior Surveillance System in America (Centers for Disease Control, 2015) identifying suicide ideation, suicide plan, and suicide attempt. Control data included age, sex, regional area (Shenzhen, Nanchang, Zhengzhou, and Guiyang), urban/rural residency, parents’ education level, family economic status, sibling status, and resident student status. LCA resulted in four ACEs patterns identified as Class 1: High ACEs (CEA, CPA, CSA, CEN, CPN, SUD, DP, DV, EA); Class 2: High Abuse And Neglect (CPA, CEA, CPN, CEN). Class 3: High Neglect (CPN, CEN); and Class 4: Low ACEs. There was no sex difference in the distribution between classes. 

Several patterns emerged from comparing latent classes, sex, and suicidal behaviors. The strongest predictor of suicide ideation and plan was EA; for suicide attempt was CSA. All classes had higher risk of suicide ideation, suicide plan and suicidal behavior as compared to the low ACEs class, with the high ACEs having the strongest association. Between sexes, CEN had a stronger effect on all suicidal behaviors for females as compared to males. This study found an important gender difference in CEN effect on suicidal behaviors where females have higher risk. A limitation of particular concern for this study is potential confounds including bullying and community violence. Since this is a study related to suicidal behavior, selection bias by death is also a concern. 

In the Shin et al. (2018) study, the data was expected to reveal distinct groups of ACEs with differential risk profiles for alcohol-related problems, tobacco use, drug dependence and psychological distress. The purpose of this study was to identify ACE domain patterns associated with common adverse outcomes in young adults. The data was expected to reveal patterns relating to alcohol-related problems, tobacco use, drug dependence and psychological distress. Data was collected from interviews of community-based participants ages 18–25, recruited online and offline (N=336). Phone screening resulted in volunteer exclusion when the age range, current physical health was rated as bad or very bad. A trained interviewer administered an hour-long, structured, in-person interview. ACEs were measured using the CTQ with additional questions to assess DV, SUD, IFM, MI, frequency of verbal abuse, CVE and CVW. Substance use and psychological symptoms were measured with various indicators. Alcohol problems were measured using the Rutgers Alcohol Problem Index (RAPI); a tobacco use questionnaire was developed for use in this study; drug use was measured using the substance section of the Composite International Diagnostic Interview (CIDI). The Brief Symptom Inventory (BSI) was used to measure recent feelings of destress including somatization, depression, and anxiety. Control data included age, gender, race/ethnicity, family income, and peer alcohol, tobacco, and drug use. LCA resulted in four ACE clustering patterns identified as Class 1: Low ACEs; Class 2: Household Dysfunction/CV (SUD, MI, property crime); Class 3: Emotional ACEs (CEA, CEM); and Class 4: High/Multiple Aces (CEA, CPA, CEN, SUD, DV, IFM, verbal aggression, property crime, gang violence). 

Several patterns emerged from comparing latent classes, demographics and adverse outcomes. Lower family income was related to membership in all groups other than Class 1 but not between classes 2, 3 or 4. Age and female gender were each related to an increased likelihood of Class 4 membership. Class 4 members were more likely to use tobacco and have alcohol problems as compared to Class 1. All groups had higher levels of psychological distress as compared to Class 1. There was no relationship between class membership and drug dependency. There were no differences in outcomes between classes 2 and 3. Class 4 members were more likely to have alcohol problems or psychological symptoms as compared to Class 3, and significantly higher than Class 2. There was not a significant difference between drug dependence or tobacco use in classes 2, 3 and 4.

Although this study identified differences between classes, the pattern was consistent with a CR model, suggesting quantity rather than type of ACE is a better predictor of substance use in this population. This study has some unique limitations by its design. Study data was collected through interviews rather than surveys, which has the potential for higher rates of social desirability bias. Further, the narrow age range may be missing periods of more significant differentiation. There is also a potential confound from resiliency, which is not controlled for in this study.  This study has a particularly high risk of sampling bias. The interview format and time commitment are potential sources of self-selection and non-response bias. There is also a risk of healthy participant bias as those in poor health were excluded from the study, thus resulting in less unhealthy participants. Pre-screening and advertising bias may have also affected this sample as both were used. As compared to other studies, this population was particularly small. 

Conclusion

The studies contained in this review support the hypothesis of a relationship between the type of ACEs and different outcomes except for the Shin et al.(2018). Although the class distribution of ACEs was not consistent, each study established a low ACEs/low risk class, a high ACEs/high risk class, and intermediate groups with variable risk. The variable risk groups show significant commonality but also particular risks depending on ACE. Of particular interest is the relationship between SUDs and depression found in Kim et al. (2019), where the association between parental substance use without significant other ACEs was associated with older adult SUD without depression. From the Lee et al. (2020) study, the underdiagnosis of minorities was striking, suggesting symptom inventory rather than self-reported diagnoses may be better to capture those without official diagnosis. Barboza (2017) identified HIV risk-taking behavior with high conflict ACEs, but not with CSA, contrary to other studies, which should be further investigated. Li et al. (2021) studied a different population with a different focus as compared to the other studies, but found higher risk for females who experienced CEN consistent with other studies and strong associations between EA, CSA and suicidal behavior. Although the Shin et al.(2018) study did not find class analysis informative beyond CR, given the particularly small sample size and significant risk of sampling bias, a similar study with a less biased sample could result in findings more consistent with other similar studies, so additional research is recommended.

Limitations

All the contained studies are limited based on the retrospective self-reporting nature of the data resulting in recall bias and an inability to conclusively determine temporal precedence. However, retrospective ACE recall has been established as reliable and valid in adult populations (Hardt and Rutter, 2004). Researchers studying this topic presume ACEs happen before adverse outcomes, but longitudinal studies would be more appropriate to establish causation. These studies do not account for ACE perpetrator, frequency, intensity, or age at the time of the ACE, which are potential confounds. Future studies could include expanded ACE data to assess possible confounds.

References

Barboza, G. E. (2017). Latent classes and cumulative impacts of adverse childhood experiences. Child Maltreatment, 23(2), 111–125. https://doi.org/10.1177/1077559517736628

Felitti, V. J., Anda, R. F., Nordenberg, D., Williamson, D. F., Spitz, A. M., Edwards, V., Koss, M. P., & Marks, J. S. (1998). Relationship of childhood abuse and household dysfunction to many of the leading causes of death in adults. American Journal of Preventive Medicine, 14(4), 245–258. https://doi.org/10.1016/s0749-3797(98)00017-8

Haahr-Pedersen, I., Perera, C., Hyland, P., Vallières, F., Murphy, D., Hansen, M., Spitz, P., Hansen, P., & Cloitre, M. (2020). Females have more complex patterns of childhood adversity: Implications for mental, social, and emotional outcomes in adulthood. European Journal of Psychotraumatology, 11(1), 1708618. https://doi.org/10.1080/20008198.2019.1708618

Kim, Y., Kim, K., Chartier, K. G., Wike, T. L., & McDonald, S. E. (2019). Adverse childhood experience patterns, major depressive disorder, and substance use disorder in older adults. Aging & Mental Health, 25(3), 484–491. https://doi.org/10.1080/13607863.2019.1693974

Lee, H., Kim, Y., & Terry, J. (2020). Adverse childhood experiences (ACEs) on mental disorders in young adulthood: Latent classes and community violence exposure. Preventive Medicine, 134, 106039. https://doi.org/10.1016/j.ypmed.2020.106039

Li, S., Wang, S., Gao, X., Jiang, Z., Xu, H., Zhang, S., Sun, Y., Tao, F., Chen, R., & Wan, Y. (2021). Patterns of adverse childhood experiences and suicidal behaviors in adolescents: A four-province study in China. Journal of Affective Disorders, 285, 69–76. https://doi.org/10.1016/j.jad.2021.02.045

Shin, S. H., McDonald, S. E., & Conley, D. (2018). Patterns of adverse childhood experiences and substance use among young adults: A latent class analysis. Addictive Behaviors, 78, 187–192. https://doi.org/10.1016/j.addbeh.2017.11.020