Population vs Individual Prediction of Poor Health From Results of Adverse Childhood Experiences Screening

Key Points Question Can screening for adverse childhood experiences (ACEs) accurately predict individual risk for later health problems? Findings In 2 population-based birth cohorts (with a total of 2927 individuals) growing up 20 years and 20 000 km apart, ACE scores were associated with mean group differences in health problems independent of other information available to clinicians. However, ACE scores had low accuracy in predicting health problems at the individual level. Meaning ACE scores can forecast mean group differences in later health problems; however, ACE scores have poor accuracy in identifying individuals at high risk for future health problems.

shaken on a plate shaker for 30 min. hsCRP was quantified via Sandwich ELISA using the Human C-Reactive Protein ELISA Kit KHA0031 (Life Technologies, UK) according to manufacturers' instructions. The kit has a lower limit of detection <10 pg/mL. Standard deviations were calculated from the samples' duplicates giving a coefficient of variation of 3.3%. Paired collection of dried blood spot and serum in n = 98 Study members was performed to derive a within-study conversion equation for hsCRP levels, where serum CRP value = 6.51 * (blood spot CRP value) + 0.14. We excluded 64 Study members with serumequivalent CRP >10 mg/L as they were likely to have acute trauma, infections, or pathology. 21 We defined high inflammation levels as hsCRP>3 mg/L, in line with the Centers for Disease Control and Prevention (CDC)/American Heart Association (AHA) definition of high cardiovascular risk. 22 Asthma. Asthma since age 12 were assessed at age 18 in a private individual interview conducted by trained professionals. 23 Participants were asked "Since age 12, have you been told by a doctor that you have asthma?" Those who responded "yes" were coded as having asthma.
Sexually transmitted diseases (STDs). Participants answered questions about sexual health at age 18 via a private computer-administered questionnaire based on the 1990 British National Survey of Sexual Attitudes and Lifestyles. 24,25 Participants were first asked if they had ever had sexual intercourse, and if so, were subsequently asked whether they had been told by a doctor since age 12 that they had Chlamydia, Genital Warts of Human Papillomavirus, Gonorrhea, genital herpes, viral hepatitis, or any other STDs. Reports of any of these were taken to indicate that the participant had contracted an STD. Participants who had not had sexual intercourse were coded as not having an STD.
Sleep problems. We measured sleep quality at age 18 years using the Pittsburgh Sleep Quality Index (PSQI). 26  are used to derive scores for seven different components of sleep (subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleep medication and daytime dysfunction), each scored from 0 to 3. These were summed to produce a global score ranging from 0 to 21, with higher scores reflecting worse sleep quality. We defined sleep problems as a score higher than 5 on the PSQI, which has been proposed as a clinical cut-off. 26 Smoking. Daily cigarette smoking at age 18 was assessed by asking the participant if they had ever smoked a cigarette, followed by if and when they began smoking every day; current daily smokers were participants who endorsed daily smoking within the past year.
Any physical health problem. We derived a composite measure reflecting the presence of any physical health problem at age 18 (e.g., obesity, high inflammation levels, asthma, STD, sleep problems, and smoking). Participants who were missing data for all physical health problems at age 18 were excluded.

eMethods 6. The E-Risk Study: Clinically Available Childhood Risk Factors
To test whether prospectively ascertained ACE scores predicted health problems over and above other clinically available information, we adjusted for health risk factors in childhood that could be readily assessed by clinicians: namely, sex, family socioeconomic status, childhood mental health problems (for analyses on later mental health outcomes) and childhood physical health problems (for analyses on later physical health outcomes). We did not include ethnicity given the low ethnic heterogeneity of the E-Risk sample.
Family socioeconomic status. Family socioeconomic status was defined using a standardized composite of parents' income, education and social class ascertained at childhood phases of the study, which loaded significantly onto one latent factor. 27 The population-wide distribution of the resulting factor was divided in tertiles for analyses. Child mental health. At ages 5, 7, 10, and 12 years, parents and teachers completed the Child Behavior Checklist (CBCL) and teacher's report form (TRF), respectively. 28,29 Childhood internalizing problems were assessed through the withdrawn/depressed and somatic subscales, and childhood externalizing problems were assessed through the delinquency and aggression subscales. The total scores for internalizing and externalizing problems (respectively) were standardized and averaged across raters and assessments.
We defined poor child mental health as a score of ≥1 standard deviation above the cohort means for internalizing or/and externalizing problems (23.4%; N=518). Participants who were missing data for both internalizing and externalizing problems were excluded.
Childhood physical health. We measured children's physical health from parent reports and clinical ratings of illnesses and health conditions taken at assessments spanning birth to age 12 years. Long-term illnesses since birth were assessed at age 5 through mother's reports using an event history calendar. 30 Asthma was assessed at ages 5 and 10 through collecting information from mothers using an event history calendar. 23 Overweight was assessed at ages 10 and 12 through research workers' ratings of children's weights on a 7-point scale (with 1 being underweight and 7 being overweight). 31 Research worker ratings of children's weight at age 10 were correlated with their ratings at age 12 (r = 0.58). We defined overweight in childhood as a score of 6 or 7 at either age 10 or 12 years. Smoking in the past six months was assessed at age 12 through mother's report on the item "smokes tobacco" on the CBCL. We defined smoking as a response of "sometimes or somewhat true" or "very often true". From this information, we derived an overall measure of poor child physical health indexing the presence of any childhood long-term illness, asthma, overweight, or smoking. Participants who were missing data for all childhood physical health problems were excluded.  The Dunedin Study assessed five types of child harm (physical abuse, sexual abuse, emotional abuse, emotional neglect, and physical neglect) and five types of household dysfunction (household partner violence, household substance abuse, family mental illness, parental criminality, and parental separation) to correspond to the 10 categories of childhood adversity introduced by the CDC Adverse Childhood Experiences Study. 1 These adversities were assessed both prospectively in childhood and retrospectively in adulthood. eFigure 7 shows the prevalence of prospectively and retrospectively measured ACEs in the Dunedin cohort compared to the E-Risk cohort and CDC ACEs Study. 1

Prospective ACE measure:
Prospective ACE counts were generated from archival Dunedin Study records gathered during seven biennial assessments carried out from ages 3 to 15 years. The records include the following: social service contacts; structured notes from assessment staff who interviewed Study children and their parents; structured notes from pediatricians and psychometricians who observed mother-child interactions at the research unit; structured notes from nurses who recorded conditions witnessed at home visits; and notes of concern from teachers who were surveyed about the Study children's behavior and performance.
Separately, parental criminality was surveyed via postal questionnaire to the parents.
Archival Study data were reviewed in 2015 by four independent raters who were trained on the CDC definitions of ACEs. Individual ACEs were agreed upon by at least three of the four raters 80% of the time. The sole exception was emotional neglect where half the cases were identified by only two raters. Agreement across the full ACE count between the four raters ranged from kappa = .76 to .82, with an average inter-rater agreement kappa of .79. The completeness of archival Dunedin Study records of adversity varied by the type of ACE considered. Some ACEs (notably childhood sexual abuse) will have been underdetected to the extent that these experiences were not actively queried, reflecting assumptions in the 1970s that sexual abuse was exceedingly rare. 33 ACE summary score. We derived a prospective ACE score by summing the dichotomised physical neglect, household partner violence, household substance abuse, family mental illness, parental criminality, and parental separation). We truncated the ACE score at 4+ ACEs in line with conventions in research 1,2 and clinical practice. 16 Binary 4+ ACE cut-off score. We also derived a binary cut-off score indexing whether participants had prospective evidence of four or more ACEs or not (i.e., exposure to 0-3 ACEs), for sensitivity analyses.

Retrospective ACE measure:
ACEs were retrospectively assessed through structured interviews conducted when Dunedin Study participants were adults. Like the original CDC ACE Study, 1 we administered the Childhood Trauma Questionnaire (CTQ), 34 which ascertains physical, sexual, and emotional abuse, physical neglect, and emotional neglect; the CTQ was administered at age 38.
Following the CTQ manual, a specific category of harm was present if the Study member had a moderate to severe score. Study members were also interviewed about memories of exposure to family substance abuse, mental illness, and incarceration during childhood via the Family History Screen. 11 Exposure to partner violence was assessed by asking Study participants, 'Did you ever see or hear about your mother/father being hit or hurt by your father/mother/stepfather/stepmother?' We also interviewed participants about parental loss (due to separation, divorce, death, or removal from home).
ACE summary score. We derived a retrospective ACE score by summing the dichotomised component measures (namely, physical abuse, sexual abuse, emotional abuse and neglect, physical neglect, household partner violence, household substance abuse, family mental illness, parental criminality, and parental separation). We truncated the ACE score at 4+ ACEs, in line with conventions in research 1,2 and clinical practice. 16 Binary 4+ ACE cut-off score. We also derived a binary cut-off score indexing whether participants retrospectively reported four or more ACEs or not (i.e., exposure to 0-3 ACEs), for sensitivity analyses. Clinically trained interviewers conducted private interviews with the study members at 45 years of age using the Diagnostic Interview Schedule (DIS) 19,35 to assess the presence of psychiatric disorders over the previous 12 months. Diagnoses of major depression, generalized anxiety disorder, and ADHD were made according to the symptom and impairment criteria from the DSM-5. 36 Diagnoses of alcohol dependence, marijuana dependence, and drug dependence were made according to criteria from the DSM-IV, 37 and marijuana dependence and drug dependence were combined into one overall drug dependence measure. Self-harm and suicide attempts since age 38 years were also assessed at age 45 during structured interviews about self-harm and suicide. Interviewers differentiated between suicide attempts and nonsuicidal self-harm.
Any mental health problem. We derived a composite measure reflecting the presence of any mental health problem at age 45 (e.g., depression, anxiety, self-harm, suicide attempt, ADHD, alcohol dependence, and drug dependence). Participants who were missing data for all mental health problems at age 45 were excluded. Obesity. Individuals' height and weight were measured at age 45. Height was measured to the nearest millimeter using a Seca 264 Wireless Stadiometer. Weight was measured to the nearest 0.1 kg using calibrated scales. Individuals were weighed in light clothing. Obesity was defined as a BMI≥30.
Inflammation. Elevated systemic inflammation at age 45 was assessed using a highsensitivity immunoturbidimetric assay of C-reactive protein (hsCRP) in serum. HsCRP was measured on a Cobas c702 analyzer. Study members with serum-equivalent CRP values >10 mg/L were excluded as they were likely to have acute trauma, infections, or pathology. 21 The Centers for Disease Control and Prevention (CDC)/American Heart Association (AHA) definition of high cardiovascular risk (hsCRP>3 mg/L) was adopted to identify the risk group. 22 Asthma. Current asthma at age 45 was defined as a self-reported current diagnosis at least one of (1) recurrent wheeze, (2) asthma attack, or (3) asthma medication use in the past year.
Sexually transmitted diseases (STDs). At age 45, Study members were asked whether they had experienced one or more STDs since the previous assessment at age 38. 38 Conditions were identified from a list of the common STDs (chlamydia, non-specific urethritis [NSU], genital warts, herpes, gonorrhoea, trichomoniasis, syphilis), or specified as an "other STI".
Reports of any of these STDs were taken to indicate that the participant had contracted an STD between ages 38 and 45.
Sleep problems. We measured sleep quality at age 45 using the Pittsburgh Sleep Quality Index (PSQI). 26 The PSQI consists of 18 self-report items relating to individuals' sleep patterns and different forms of sleep impairment in the past month. These questions are used to derive scores for seven different components of sleep (subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleep medication and daytime dysfunction), each scored from 0 to 3. These were summed to produce a global score ranging from 0 to 21, with higher scores reflecting worse sleep quality. We defined sleep problems as a score higher than 5 on the PSQI, which has been proposed as a clinical cut-off. 26 Smoking. Current smoking was defined as smoking at least one cigarette daily for at least 1 month in the previous year based on reports by the participants at age 45.
Any physical health problem. We derived a composite measure reflecting the presence of any physical health problem at age 45 (e.g., obesity, high inflammation levels, asthma, STD, sleep problems, and smoking). Participants who were missing data for all physical health problems at age 45 were excluded.

eMethods 11. The Dunedin Study: Clinically Available Childhood Risk Factors
To test whether prospectively ascertained ACE scores predicted health problems over and above other clinically available information, we adjusted for health risk factors in childhood that could be readily assessed by clinicians: namely, sex, family socioeconomic status, childhood mental health problems (for analyses on later mental health outcomes) and childhood physical health problems (for analyses on later physical health outcomes). We did not include ethnicity given the low ethnic heterogeneity of the Dunedin sample. Child mental health. At ages 5, 7, 9, and 11, parents and teachers completed the Rutter Child Scale A and B, respectively, and additional items assessing inattention, impulsivity, and hyperactivity. Subscales, based upon factor analysis of all items, have been formed for measuring hyperactivity, antisocial behaviour, and worry-fearfulness. 39 We defined poor child mental health as a score of ≥1 standard deviation above the cohort mean for hyperactivity, antisocial behaviour, or worry-fearfulness (29.6%; N=304). Participants who were missing data for hyperactivity, antisocial behaviour, and worry-fearfulness were excluded.
Child physical health. We measured cohort members' childhood health from a panel of biomarkers and clinical ratings taken at assessments spanning birth to age 11 years. 40  Children's overall health at ages 3, 5, 7, 9, and 11 years was rated by two Unit staff members based on review of birth records and assessment dossiers including clinical assessments and reports of infections, diseases, injuries, hospitalizations, and other health problems collected from children's mothers during standardized interviews. Ratings were made on a five-point scale (inter-rater agreement=0.85). Body mass index was calculated from height and weight measurements taken at ages 5, 7, 9, and 11 years. In addition, tricep and subscapular skinfold thicknesses were measured at ages 7 and 9 years by trained anthropometrists. 44 (For calculation of the overall measure, tricep and subscapular skinfold thicknesses were averaged to create a single score.) Systolic and diastolic blood pressure were measured at ages 7, 9, and 11 years using a London School of Hygiene and Tropical Medicine blind mercury sphygmomanometer (Cinetronics Ltd., Mildenhall, United Kingdom). 45 Fixed expiratory volume in one second (FEV1) and the ratio of FEV1 to forced vital capacity (FVC) were measured at ages 9 and 11 using a Godart water spirometer. 46 To calculate the childhood health measure, assessments were standardized to have mean=0 and SD=1 within age and sex specific groups. Cross-age scores for each measure were   Abbreviations: ACE = adverse childhood experiences; ADHD = attention deficit hyperactivity disorder; SES = socioeconomic status. Note: Results are presented as Relative Risks and 95% confidence intervals for health problems per additional ACE experienced. We controlled for covariates measured at the time of ACE assessment to reflect information clinicians would have access to at the time of ACE screening. Specifically, analyses using prospective ACE measures adjusted for risk factors measured in childhood (e.g., family socioeconomic disadvantage; child mental health problems) whereas analyses using the retrospective ACE measure adjusted for risk factors in adulthood (e.g., socioeconomic disadvantage at age 38, self-reported health problems at age 38). We adjusted for sex in analyses based on both prospective and retrospective ACE measures.   Note. The Figure shows the prevalence of ACEs in the E-Risk and Dunedin cohorts compared to the CDC ACEs Study 1 . Small differences reflect slightly lower levels of ACEs in the CDC ACEs Study relative to in the E-Risk and Dunedin cohorts, suggesting that under-detection of ACEs is not likely to be an issue in the E-Risk and Dunedin cohorts. Differences between the prevalences likely reflect differences in the assessment methods (e.g., repeated interviews and observations vs. a single questionnaire; prospective vs. retrospective assessment), informant (e.g. self-reports vs. parent reports/records), geographical locations (e.g., the USA, UK, or New Zealand) and ages.