The ACE Study is based at Kaiser Permanente’s San Diego Health Appraisal Clinic. More than 45,000 adults undergo standardized examinations there each year, making this clinic one of the nation’s largest free-standing medical evaluation centers. All enrollees in the Kaiser Health Plan in San Diego are advised through sales literature about the services (free for members) at the clinic; after enrollment, members are advised again of its availability through new-member literature. Most members obtain appointments by self-referral; 20% are referred by their health care provider. A recent review of membership and utilization records among Kaiser members in San Diego continuously enrolled between 1992 and 1995 showed that 81% of those 25 years and older had been evaluated in the Health Appraisal Clinic.

Health appraisals include completion of a standardized medical questionnaire that requests demographic and biopsychosocial information, review of organ systems, previous medical diagnoses, and family medical history. A health care provider completes the medical history, performs a physical examination, and reviews the results of laboratory tests with the patient.

The ACE Study protocol was approved by the Institutional Review Boards of the Southern California Permanente Medical Group (Kaiser Permanente), the Emory University School of Medicine, and the Office of Protection from Research Risks, National Institutes of Health. All 13,494 Kaiser Health Plan members who completed standardized medical evaluations at the Health Appraisal Clinic between August–November of 1995 and January–March of 1996 were eligible to participate in the ACE Study. Those seen at the clinic during December were not included because survey response rates are known to be lower during the holiday period.[38[38]

In the week after visiting the clinic, and hence having their standardized medical history already completed, members were mailed the ACE Study questionnaire that included questions about childhood abuse and exposure to forms of household dysfunction while growing up. After second mailings of the questionnaire to persons who did not respond to the first mailing, the response rate for the survey was 70.5% (9,508/13,494).

A second survey wave of approximately the same number of patients as the first wave was conducted between June and October of 1997. The data for the second survey wave is currently being compiled for analysis. The methods for the second mail survey wave were identical to the first survey wave as described above. The second wave was done to enhance the precision of future detailed analyses on special topics and to reduce the time necessary to obtain precise statistics on follow-up health events. An overview of the total ACE Study design is provided in Fig. 1Fig. 1.

Fig. 1

ACE Study design. ∗After exclusions, 59.7% of the original wave I sample (8,056/13,494) were included in this analysis.

View Large Image | Download PowerPoint Slide

We abstracted the completed medical evaluation for every person eligible for the study; this included their medical history, laboratory results, and physical findings. Respondent (n = 9,508) and nonrespondent (n = 3,986) groups were similar in their percentages of women (53.7% and 51.0%, respectively) and in their mean years of education (14.0 years and 13.6 years, respectively). Respondents were older than nonrespondents (means 56.1 years and 49.3 years) and more likely to be white (83.9% vs. 75.3%) although the actual magnitude of the differences was small.

Respondents and nonrespondents did not differ with regard to their self-rated health, smoking, other substance abuse, or the presence of common medical conditions such as a history of heart attack or stroke, chronic obstructive lung disease, hypertension, or diabetes, or with regard to marital status or current family, marital, or job-related problems (data not shown). The health appraisal questionnaire used in the clinic contains a single question about childhood sexual abuse that reads “As a child were you ever raped or sexually molested?” Respondents were slightly more likely to answer affirmatively than nonrespondents (6.1% vs. 5.4%, respectively).

We used questions from published surveys to construct the ACE Study questionnaire. Questions from the Conflicts Tactics Scale[39[39]were used to define psychological and physical abuse during childhood and to define violence against the respondent’s mother. We adapted four questions from Wyatt[40[40]to define contact sexual abuse during childhood. Questions about exposure to alcohol or drug abuse during childhood were adapted from the 1988 National Health Interview Survey.[41[41]All of the questions we used in this study to determine childhood experiences were introduced with the phrase “While you were growing up during your first 18 years of life …”

Questions about health-related behaviors and health problems were taken from health surveys such as the Behavioral Risk Factor Surveys[42[42]and the Third National Health and Nutrition Examination Survey,[43[43]both of which are directed by the Centers for Disease Control and Prevention. Questions about depression came from the Diagnostic Interview Schedule of the National Institute of Mental Health (NIMH).[44[44]Other information for this analysis such as disease history was obtained from the standardized questionnaire used in the Health Appraisal Clinic. (A copy of the questionnaires used in this study may be found at www.elsevier.com/locate/amepre.)

We used three categories of childhood abuse: psychological abuse (2 questions), physical abuse (2 questions), or contact sexual abuse (4 questions). There were four categories of exposure to household dysfunction during childhood: exposure to substance abuse (defined by 2 questions), mental illness (2 questions), violent treatment of mother or stepmother (4 questions), and criminal behavior (1 question) in the household. Respondents were defined as exposed to a category if they responded “yes” to 1 or more of the questions in that category. The prevalence of positive responses to the individual questions and the category prevalences are shown in Table 1Table 1.

We used these 7 categories of childhood exposures to abuse and household dysfunction for our analysis. The measure of childhood exposure that we used was simply the sum of the categories with an exposure; thus the possible number of exposures ranged from 0 (unexposed) to 7 (exposed to all categories).

Using information from both the study questionnaire and the Health Appraisal Clinic’s questionnaire, we chose 10 risk factors that contribute to the leading causes of morbidity and mortality in the United States.[6, 7, 36, 37]The risk factors included smoking, severe obesity, physical inactivity, depressed mood, suicide attempts, alcoholism, any drug abuse, parenteral drug abuse, a high lifetime number of sexual partners (≥50), and a history of having a sexually transmitted disease.

We also assessed the relationship between childhood exposures and disease conditions that are among the leading causes of mortality in the United States.[6[6]The presence of these disease conditions was based upon medical histories that patients provided in response to the clinic questionnaire. We included a history of ischemic heart disease (including heart attack or use of nitroglycerin for exertional chest pain), any cancer, stroke, chronic bronchitis, or emphysema (COPD), diabetes, hepatitis or jaundice, and any skeletal fractures (as a proxy for risk of unintentional injuries). We also included responses to the following question about self-rated health: “Do you consider your physical health to be excellent, very good, good, fair, or poor?” because it is strongly predictive of mortality.[45[45]

We defined severe obesity as a body mass index (kg/meter²) ≥35 based on measured height and weight; physical inactivity as no participation in recreational physical activity in the past month; and alcoholism as a “Yes” response to the question “Have you ever considered yourself to be an alcoholic?” The other risk factors that we assessed are self-explanatory.

Of the 9,508 survey respondents, we excluded 51 (0.5%) whose race was unstated and 34 (0.4%) whose educational attainment was not reported. We also excluded persons who did not respond to certain questions about adverse childhood experiences. This involved the following exclusions: 125 (1.3%) for household substance abuse, 181 (1.9%) for mental illness in the home, 148 (1.6%) for violence against mother, 7 (0.1%) for imprisonment of a household member, 109 (1.1%) for childhood psychological abuse, 44 (0.5%) for childhood physical abuse, and 753 (7.9%) for childhood sexual abuse. After these exclusions, 8,056 of the original 9,508 survey respondents (59.7% of the original sample of 13,494) remained and were included in the analysis. Procedures for insuring that the findings based on complete data were generalizable to the entire sample are described below.

The mean age of the 8,506 persons included in this analysis was 56.1 years (range: 19–92 years); 52.1% were women; 79.4% were white. Forty-three percent had graduated from college; only 6.0% had not graduated from high school.

We used the Statistical Analysis System (SAS)[46[46]for our analyses. We used the direct method to age-adjust the prevalence estimates. Logistic regression analysis was employed to adjust for the potential confounding effects of age, sex, race, and educational attainment on the relationship between the number of childhood exposures and health problems.

To test for a dose-response relationship to health problems, we entered the number of childhood exposures as a single ordinal variable (0, 1, 2, 3, 4, 5, 6, 7) into a separate logistic regression model for each risk factor or disease condition.

To determine whether our results were influenced by excluding persons with incomplete information on any of the categories of childhood exposure, we performed a separate sensitivity analysis in which we included all persons with complete demographic information but assumed that persons with missing information for a category of childhood exposure did not have an exposure in that category.

The level of positive responses for the 17 questions included in the seven categories of childhood exposure ranged from 3.0% for a respondent’s mother (or stepmother) having been threatened with or hurt by a gun or knife to 23.5% for having lived with a problem drinker or alcoholic (Table 1Table 1). The most prevalent of the 7 categories of childhood exposure was substance abuse in the household (25.6%); the least prevalent exposure category was evidence of criminal behavior in the household (3.4%). More than half of respondents (52%) experienced ≥1 category of adverse childhood exposure; 6.2% reported ≥4 exposures.

The probability that persons who were exposed to any single category of exposure were also exposed to another category is shown in Table 2Table 2. The relationship between single categories of exposure was significant for all comparisons (P < .001; chi-square). For persons reporting any single category of exposure, the probability of exposure to any additional category ranged from 65%–93% (median: 80%); similarly, the probability of ≥2 additional exposures ranged from 40%–74% (median: 54.5%).

The number of categories of childhood exposures by demographic characteristics is shown in Table 3Table 3. Statistically, significantly fewer categories of exposure were found among older persons, white or Asian persons, and college graduates (P < .001). Because age is associated with both the childhood exposures as well as many of the health risk factors and disease outcomes, all prevalence estimates in the tables are adjusted for age.

Both the prevalence and risk (adjusted odds ratio) increased for smoking, severe obesity, physical inactivity, depressed mood, and suicide attempts as the number of childhood exposures increased (Table 4Table 4). When persons with 4 categories of exposure were compared to those with none, the odds ratios ranged from 1.3 for physical inactivity to 12.2 for suicide attempts (Table 4Table 4).

Similarly, the prevalence and risk (adjusted odds ratio) of alcoholism, use of illicit drugs, injection of illicit drugs, ≥50 intercourse partners, and history of a sexually transmitted disease increased as the number of childhood exposures increased (Table 5Table 5). In comparing persons with ≥4 childhood exposures to those with none, odds ratios ranged from 2.5 for sexually transmitted diseases to 7.4 for alcoholism and 10.3 for injected drug use.

We found a strong relationship between the number of childhood exposures and the number of health risk factors for leading causes of death in adults (Table 6Table 6). For example, among persons with no childhood exposures, 56% had none of the 10 risk factors whereas only 14% of persons with ≥4 categories of childhood exposure had no risk factors. By contrast, only 1% of persons with no childhood exposures had four or more risk factors, whereas 7% of persons with ≥4 childhood exposures had four or more risk factors (Table 6Table 6).

When persons with 4 or more categories of childhood exposure were compared to those with none, the odds ratios for the presence of studied disease conditions ranged from 1.6 for diabetes to 3.9 for chronic bronchitis or emphysema (Table 7Table 7). Similarly, the odds ratios for skeletal fractures, hepatitis or jaundice, and poor self-rated health were 1.6, 2.3, and 2.2, respectively (Table 8Table 8).

In logistic regression models (which included age, gender, race, and educational attainment as covariates) we found a strong, dose-response relationship between the number of childhood exposures and each of the 10 risk factors for the leading causes of death that we studied (P < .001). We also found a significant (P < .05) dose-response relationship between the number of childhood exposures and the following disease conditions: ischemic heart disease, cancer, chronic bronchitis or emphysema, history of hepatitis or jaundice, skeletal fractures, and poor self-rated health. There was no statistically significant dose-response relationship for a history of stroke or diabetes.

In the sensitivity analysis where missing information for a category of childhood exposure was considered as no exposure, the direction and strength of the associations between the number of childhood exposures and the risk factors and disease conditions were nearly identical (data not shown). Thus, the results we present appear to be unaffected by our decision to exclude persons for whom information on any category of childhood exposure was incomplete.