Adverse childhood experiences increase the long-term accumulation of morbidity in women

The study sample was repurposed from an existing cohort study, the Mayo Clinic Cohort Study of Oophorectomy and Aging –2 (MOA-2) 12 , 13 . However, the sampling strategy was modified specifically for the objectives of the current study. We sampled randomly from a complete enumeration of the population a cohort of women of age 21–45 years who resided in Olmsted County, Minnesota, between 1988 and 2007. Contrary to the sampling for the MOA-2 study, women were included in the current sample regardless of any gynecologic disease or any gynecologic surgery that occurred before the inclusion in the study. Therefore, our final sample of 1026 women can be considered a random sample of the complete Olmsted County female population in the defined age range over a 20-year time window (1988–2007). The date and age at which the women were passively included in the study was considered the index date and index age. All research activities were approved by the Mayo Clinic and Olmsted Medical Center Institutional Review Boards. An informed consent specific for this study was not required because the women were not contacted in person; however, all women in the study provided a general authorization for the use of their medical records information for research (Minnesota Research Authorization) 14 . We used the 10 categories of childhood adversity introduced by the Centers for Disease Control and Prevention 1 , 15 , 16 . We considered 5 types of child harm (physical abuse, verbal or emotional abuse, sexual abuse, emotional neglect, and physical neglect) and 5 types of household dysfunction (parental separation, domestic violence, substance abuse, mental illness, and incarceration of a family member). Any abuse included physical abuse, verbal or emotional abuse, and sexual abuse. The items were also grouped as an overall ACE score assigning the same weight to each item (0–10 points). In addition, we also studied abuse (physical, verbal or emotional, or sexual) that happened between age 19 and the index age (age at inclusion into the study). Finally, we combined abuse during childhood and abuse between age 19 years and index age into a cumulative lifetime experience of abuse from birth to the index age. Information on ACE and about abuse between age 19 years and index age was obtained using the medical records-linkage system of the Rochester Epidemiology Project (REP), which includes complete records (inpatient and outpatient) from all major medical care providers in Olmsted County, Minnesota 14 , 17 , 18 . Demographic and clinical characteristics as well as the individual ACE items were collected via abstraction of medical records 13 , 19 . The narrative psychiatric and medical notes were manually abstracted by a physician (LGR) to identify any reports of the adverse experiences included in the ACE questionnaire from birth through age 18 years. Examples of the statements found in the medical records that corresponded to the 10 items of the ACE questionnaire are reported in Supplementary Table  1 19 . Women without medical record documentation of ACE were considered free of adverse experiences. The medical records were informative because the median length of contact with the system before the date of recruitment for this study was ~20 years. In addition, our method of ACE measurement was previously assessed for reliability. As part of two previous studies, we reported an intra-rater agreement of 90% with a kappa of 80% (95% CI 54–100) and an inter-rater agreement of 94% with kappa of 83% (95% CI 76–92) for the medical records abstraction 19 , 20 . We studied the accumulation of 18 chronic conditions (morbidity) between the date of passive inclusion in the study (index date) and December 31, 2021. We defined morbidity using the 20 chronic conditions and the corresponding International Classification of Diseases codes recommended by the Department of Health and Human Services (DHHS) 21 , 22 . However, we excluded non-melanoma skin cancer codes from the definition of cancer. From the DHHS list we also excluded human immunodeficiency virus (HIV) and autism spectrum disorder because these conditions were rare in our sample (total of 18 conditions considered). The diagnostic codes were extracted electronically from the REP records-linkage system. A woman was required to have received two or more codes separated by more than 30 days from the list of codes defining a condition to be counted as having the condition 23 . We considered several indicators of socioeconomic status. Education and employment status were abstracted from medical records. In addition, the geolocation of women was calculated using their address 24 . Geocoded women were assigned a census block group (114 census block groups in Olmsted County), and publicly available data were obtained for area deprivation index (ADI) at the census block group level 25 , 26 . Information about income was derived from the 2000 United States Census 27 . Each woman was assigned the median household income for the census block group in which she lived at the index date. In addition, we considered 12 events or characteristics that were present at the index date and were hypothesized to intervene in the chain of causality between ACE and the accumulation of morbidity (viz., never married, pregnancy before age 18 years, no live births, bilateral oophorectomy, hysterectomy with ovarian conservation, body mass index (BMI) ≥30 kg/m 2 , cigarette smoking, substance abuse disorders, eating disorders, depression, anxiety, and suicidal ideation). Our primary analyses considered the contrast of ACE score ≥2 versus 0. However, secondary analyses were performed using other cut-off points of the ACE score, history of child harm, history of household dysfunction, and any abuse before the index date. We studied the cumulative incidence of the 18 chronic conditions separately in women with ACE score ≥2 versus 0 and computed hazard ratios (HRs) using Cox regression models. The starting point for the cumulative incidence curves was the time of passive inclusion in the study (index date). The accumulation of morbidity (number of chronic conditions considered as a score with equal weighting of each condition) was represented graphically using Aalen–Johansen curves (a multistate generalization of cumulative incidence curves), and we computed HRs using Andersen–Gill regression models with age as the time scale 28 – 30 . The Aalen–Johansen curves were shown graphically with age starting at 35 years because earlier ages had small numbers. None of the models violated the proportional hazards assumption. We also conducted analyses to explore the possible confounding effect of socioeconomic characteristics (education, occupation, income, and ADI). Finally, we explored the possible mediation effect of 12 events or characteristics and reported their indirect effects individually and jointly. All tests of statistical significance were conducted at the two-tailed alpha level of 0.05 in primary analyses. However, we also conducted a set of secondary analyses for individual chronic conditions using one-tailed tests (assuming adequate a priori evidence of directionality). Analyses were performed using R version 4.2.2 (R Foundation for Statistical Computing, Vienna, Austria) and SAS version 9.4 (SAS Institute Inc., Cary, NC). Further information on research design is available in the  Nature Portfolio Reporting Summary linked to this article.

We studied 1026 women with a median age of 41.0 years (interquartile range [IQR], 38.0–44.0 years) at the time of inclusion into the study (index date and index age) in the time window 1988–2007 and followed them historically through their medical records for a median of 21.0 years (IQR, 16.1–25.8 years). Table  1 shows the demographic and clinical characteristics of women at inclusion into the study overall and in three strata with ACE scores 0, 1, or ≥2. The majority of women were white (94.7%). An ACE score of ≥1 was observed in 38.5% of women, a score of ≥2 in 17.8%, and a score of ≥4 in 4.8%. Five chronic conditions were more common in women with an ACE score ≥2 compared to 0 or 1 at index date, viz., arthritis, chronic obstructive pulmonary disorder, depression, schizophrenia or psychosis, and substance abuse disorders. Women with an ACE score ≥2 were also more likely to have ≥3 chronic conditions compared to women with an ACE score of 0 or 1. Table 1 Characteristics of women at recruitment for the study (index date) overall and by ACE score Characteristic, n (%) Overall ( n  = 1026) ACE Score 0 ( n  = 631) ACE Score 1 ( n  = 212) ACE Score ≥ 2 ( n  = 183) P value a Age (years), median (IQR) 41.0 (38.0, 44.0) 41.0 (38.0, 44.0) 42.0 (39.0, 44.0) 41.0 (37.0, 44.0) 0.06 Age (years) 0.08 <35 132 (12.9) 86 (13.6) 17 (8.0) 29 (15.8) 35–39 211 (20.6) 137 (21.7) 40 (18.9) 34 (18.6) 40–45 683 (66.6) 408 (64.7) 155 (73.1) 120 (65.6) Index year 0.79 1988–1992 208 (20.3) 136 (21.6) 42 (19.8) 30 (16.4) 1993–1997 251 (24.5) 149 (23.6) 51 (24.1) 51 (27.9) 1998–2002 349 (34.0) 215 (34.1) 72 (34.0) 62 (33.9) 2003–2007 218 (21.2) 131 (20.8) 47 (22.2) 40 (21.9) Race 0.51 Asian 36 (3.5) 27 (4.3) 6 (2.8) 3 (1.6) Black 15 (1.5) 9 (1.4) 2 (0.9) 4 (2.2) White 972 (94.7) 593 (94.0) 203 (95.8) 176 (96.2) Other 3 (0.3) 2 (0.3) 1 (0.5) 0 (0.0) Hispanic or Latino ethnicity 15 (1.5) 10 (1.6) 5 (2.4) 0 (0.0) 0.12 Years of education 0.002 ≤12 280 (27.6) 161 (26.1) 49 (23.1) 70 (38.3) 13–16 552 (54.5) 333 (53.9) 129 (60.8) 90 (49.2) >16 181 (17.9) 124 (20.1) 34 (16.0) 23 (12.6) Unknown 13 13 0 0 Marital status 0.0002 Married 759 (74.0) 494 (78.3) 152 (71.7) 113 (61.7) Widowed 11 (1.1) 7 (1.1) 2 (0.9) 2 (1.1) Divorced or separated 142 (13.8) 64 (10.1) 38 (17.9) 40 (21.9) Never married 114 (11.1) 66 (10.5) 20 (9.4) 28 (15.3) Employment status 0.0003 Full-time work 823 (80.4) 496 (78.7) 175 (82.9) 152 (83.1) Part-time work 92 (9.0) 60 (9.5) 22 (10.4) 10 (5.5) Homemaker 84 (8.2) 65 (10.3) 9 (4.3) 10 (5.5) Unemployed 25 (2.4) 9 (1.4) 5 (2.4) 11 (6.0) Unknown 2 1 1 0 Income quartiles 0.44 <$42,000 253 (24.7) 142 (22.5) 55 (25.9) 56 (30.6) $42,000–56,999 259 (25.2) 162 (25.7) 52 (24.5) 45 (24.6) $57,000–71,999 254 (24.8) 159 (25.2) 52 (24.5) 43 (23.5) ≥$72,000 260 (25.3) 168 (26.6) 53 (25.0) 39 (21.3) ADI national percentiles 0.36 1–25 194 (18.9) 130 (20.6) 37 (17.4) 27 (14.8) 26–50 496 (48.3) 304 (48.2) 106 (50.0) 86 (47.0) 51–75 284 (27.7) 166 (26.3) 61 (28.8) 57 (31.1) 76–100 52 (5.1) 31 (4.9) 8 (3.8) 13 (7.1) Length of medical record >20 years 501 (48.8) 289 (45.8) 107 (50.5) 105 (57.4) 0.02 Medical record history To birth year 185 (18.0) 104 (16.5) 32 (15.1) 49 (26.8) 0.003 To age 10 years 294 (28.7) 166 (26.3) 61 (28.8) 67 (36.6) 0.03 To age 15 years 336 (32.7) 185 (29.3) 68 (32.1) 83 (45.4) 0.0002 Smoking status <0.0001 Current 204 (19.9) 99 (15.7) 48 (22.6) 57 (31.1) Former 217 (21.2) 119 (18.9) 48 (22.6) 50 (27.3) Never 605 (59.0) 413 (65.5) 116 (54.7) 76 (41.5) Smoking pack-years <0.0001 0 605 (60.0) 413 (66.8) 116 (55.5) 76 (41.8) 0.1–2.9 85 (8.4) 55 (8.9) 14 (6.7) 16 (8.8) ≥3.0 319 (31.6) 150 (24.3) 79 (37.8) 90 (49.5) Unknown 17 13 3 1 Alcohol abuse 94 (9.2) 24 (3.8) 21 (9.9) 49 (26.8) <0.0001 Drug abuse 69 (6.7) 20 (3.2) 12 (5.7) 37 (20.2) <0.0001 Eating disorder 43 (4.2) 15 (2.4) 9 (4.2) 19 (10.4) <0.0001 Body mass index (kg/m 2 ) 0.046 <25.0 447 (44.1) 287 (46.1) 87 (41.6) 73 (40.1) 25.0–29.9 298 (29.4) 178 (28.6) 73 (34.9) 47 (25.8) ≥30.0 268 (26.5) 157 (25.2) 49 (23.4) 62 (34.1) Unknown 13 9 3 1 Number of pregnancies 0.12 0 149 (14.5) 94 (14.9) 33 (15.6) 22 (12.0) 1 117 (11.4) 65 (10.3) 21 (9.9) 31 (16.9) 2 299 (29.1) 193 (30.6) 53 (25.0) 53 (29.0) ≥3 461 (44.9) 279 (44.2) 105 (49.5) 77 (42.1) Number of live births 0.19 0 185 (18.0) 112 (17.7) 40 (18.9) 33 (18.0) 1 142 (13.8) 74 (11.7) 35 (16.5) 33 (18.0) 2 382 (37.2) 244 (38.7) 69 (32.5) 69 (37.7) ≥3 317 (30.9) 201 (31.9) 68 (32.1) 48 (26.2) Bilateral oophorectomy 14 (1.4) 6 (1.0) 4 (1.9) 4 (2.2) 0.29 Hysterectomy with ovarian conservation b 74 (7.2) 31 (4.9) 19 (9.0) 24 (13.1) 0.0004 Number of chronic conditions (0–18) <0.0001 0 590 (57.5) 417 (66.1) 108 (50.9) 65 (35.5) 1 243 (23.7) 138 (21.9) 53 (25.0) 52 (28.4) 2 122 (11.9) 57 (9.0) 33 (15.6) 32 (17.5) ≥3 71 (6.9) 19 (3.0) 18 (8.5) 34 (18.6) Presence of chronic conditions Hypertension 60 (5.8) 38 (6.0) 7 (3.3) 15 (8.2) 0.11 Congestive heart failure 1 (0.1) 0 (0.0) 0 (0.0) 1 (0.5) – Coronary artery disease 4 (0.4) 2 (0.3) 1 (0.5) 1 (0.5) – Cardiac arrhythmias 30 (2.9) 15 (2.4) 7 (3.3) 8 (4.4) 0.35 Hyperlipidemia 86 (8.4) 44 (7.0) 21 (9.9) 21 (11.5) 0.10 Stroke 4 (0.4) 1 (0.2) 0 (0.0) 3 (1.6) – Arthritis 43 (4.2) 15 (2.4) 18 (8.5) 10 (5.5) 0.0004 Asthma 52 (5.1) 27 (4.3) 10 (4.7) 15 (8.2) 0.10 Cancer (all types) 59 (5.8) 28 (4.4) 17 (8.0) 14 (7.7) 0.07 Chronic kidney disease 8 (0.8) 6 (1.0) 0 (0.0) 2 (1.1) 0.41 COPD c 149 (14.5) 71 (11.3) 40 (18.9) 38 (20.8) 0.0007 Dementia 2 (0.2) 0 (0.0) 0 (0.0) 2 (1.1) – Depression d 151 (14.7) 49 (7.8) 37 (17.5) 65 (35.5) <0.0001 Diabetes mellitus 20 (1.9) 10 (1.6) 4 (1.9) 6 (3.3) 0.36 Hepatitis 13 (1.3) 4 (0.6) 5 (2.4) 4 (2.2) 0.051 Osteoporosis 6 (0.6) 2 (0.3) 1 (0.5) 3 (1.6) 0.10 Schizophrenia or psychosis 9 (0.9) 0 (0.0) 3 (1.4) 6 (3.3) <0.0001 Substance abuse disorders 31 (3.0) 4 (0.6) 4 (1.9) 23 (12.6) <0.0001 ACE adverse childhood experiences, ADI area deprivation index, COPD chronic obstructive pulmonary disease, IQR interquartile range (25 th percentile, 75 th percentile). a P values were obtained from Kruskal–Wallis tests, chi-square tests, or Fisher’s exact tests comparing women across three score levels. P values were not reported when based on small numbers. b Hysterectomy performed with conservation of 1 or both ovaries. c The diagnostic codes for COPD recommended by the Department of Health and Human Services included a code for “bronchitis not specified as acute or chronic”. Of the 149 women with COPD, 98 women (65.8%) had only this non-specific code. Therefore, this code may have caused an overcounting of COPD in these relatively young women. d The diagnostic codes for depression recommended by the Department of Health and Human Services included all levels of severity (mild, moderate, or severe), acute or chronic duration, and initial or recurrent episodes. Therefore, the specific characteristics of depression could not be distinguished using diagnostic codes alone. Characteristics of women at recruitment for the study (index date) overall and by ACE score Hysterectomy with ovarian conservation b ACE adverse childhood experiences, ADI area deprivation index, COPD chronic obstructive pulmonary disease, IQR interquartile range (25 th percentile, 75 th percentile). a P values were obtained from Kruskal–Wallis tests, chi-square tests, or Fisher’s exact tests comparing women across three score levels. P values were not reported when based on small numbers. b Hysterectomy performed with conservation of 1 or both ovaries. c The diagnostic codes for COPD recommended by the Department of Health and Human Services included a code for “bronchitis not specified as acute or chronic”. Of the 149 women with COPD, 98 women (65.8%) had only this non-specific code. Therefore, this code may have caused an overcounting of COPD in these relatively young women. d The diagnostic codes for depression recommended by the Department of Health and Human Services included all levels of severity (mild, moderate, or severe), acute or chronic duration, and initial or recurrent episodes. Therefore, the specific characteristics of depression could not be distinguished using diagnostic codes alone. Table  2 shows the frequency of each one of the 10 items of the ACE score and of several cut-off categories overall and in two strata by length of the medical record information (≤20 vs. >20 years). In addition, the table shows the frequency of abuse between age 19 years and index age, or cumulatively across life (birth to the index age). Any abuse (physical, verbal or emotional, or sexual) before age 19 years was observed in 13.2% of women. Any abuse at any age before index age was observed in 24.6% of women. Table 2 ACE and abuse characteristics of women at recruitment for the study (index date) overall and by length of medical record Characteristic, n (%) Overall ( n  = 1026) Medical record ≤ 20 years ( n  = 525) Medical record > 20 years ( n  = 501) P value a Childhood (age <19 years) Q1: verbal or emotional abuse 101 (9.8) 42 (8.0) 59 (11.8) 0.04 Q2: physical abuse 49 (4.8) 24 (4.6) 25 (5.0) 0.75 Q3: sexual abuse 61 (5.9) 33 (6.3) 28 (5.6) 0.64 Q4: emotional neglect 47 (4.6) 18 (3.4) 29 (5.8) 0.07 Q5: physical neglect 18 (1.8) 4 (0.8) 14 (2.8) 0.01 Q6: parental separation 82 (8.0) 47 (9.0) 35 (7.0) 0.25 Q7: witnessed domestic violence 16 (1.6) 5 (1.0) 11 (2.2) 0.11 Q8: substance abuse in household 204 (19.9) 93 (17.7) 111 (22.2) 0.07 Q9: mental illness in household 171 (16.7) 75 (14.3) 96 (19.2) 0.04 Q10: incarceration 1 (0.1) 1 (0.2) 0 (0.0) – Any abuse b 135 (13.2) 60 (11.4) 75 (15.0) 0.09 Child harm c 160 (15.6) 69 (13.1) 91 (18.2) 0.03 Household dysfunction d 345 (33.6) 159 (30.3) 186 (37.1) 0.02 ACE score ≥1 395 (38.5) 183 (34.9) 212 (42.3) 0.01 ACE score ≥2 183 (17.8) 78 (14.9) 105 (21.0) 0.01 ACE score ≥3 87 (8.5) 38 (7.2) 49 (9.8) 0.14 ACE score ≥4 49 (4.8) 26 (5.0) 23 (4.6) 0.79 ACE score ≥5 23 (2.2) 11 (2.1) 12 (2.4) 0.75 ACE score 0.03 e 0 631 (61.5) 342 (65.1) 289 (57.7) 1 212 (20.7) 105 (20.0) 107 (21.4) 2 96 (9.4) 40 (7.6) 56 (11.2) 3 38 (3.7) 12 (2.3) 26 (5.2) 4 26 (2.5) 15 (2.9) 11 (2.2) ≥5 23 (2.2) 11 (2.1) 12 (2.4) Adulthood (age 19 to index date) Verbal or emotional abuse 174 (17.0) 64 (12.1) 110 (22.0) <0.0001 Physical abuse 71 (6.9) 36 (6.9) 35 (7.0) 0.94 Sexual abuse 23 (2.2) 15 (2.9) 8 (1.6) 0.17 Any abuse b 187 (18.2) 73 (13.9) 114 (22.8) 0.0002 Cumulative experience (birth to index) Verbal or emotional abuse 224 (21.8) 93 (17.7) 131 (26.1) 0.001 Physical abuse 106 (10.3) 54 (10.3) 52 (10.4) 0.96 Sexual abuse 77 (7.5) 43 (8.2) 34 (6.8) 0.39 Any abuse before index b 252 (24.6) 110 (21.0) 142 (28.3) 0.006 ACE score ≥1 or any abuse f 451 (44.0) 204 (38.9) 247 (49.3) 0.0008 ACE score ≥2 or any abuse f, g 307 (34.8) 132 (29.1) 175 (40.8) 0.0003 ACE score ≥3 or any abuse f, h 258 (31.0) 111 (25.7) 147 (36.7) 0.0006 ACE adverse childhood experiences. a P values were obtained from chi-square tests or Fisher’s exact tests comparing women with medical record length ≤20 years and >20 years. P values were not reported when based on small numbers. b Any abuse includes verbal or emotional, physical, or sexual abuse. c Child harm includes verbal or emotional abuse, physical abuse, sexual abuse, emotional neglect, and physical neglect. d Household dysfunction includes parental separation, domestic violence, substance abuse, mental illness, and incarceration of a family member. e The p value for the categorical ACE score was obtained from an overall chi-square test across all categories. f Compared to women with ACE score of 0 and no abuse. g Women with ACE score of 1 and no abuse were excluded from these analyses ( n  = 144 overall, 72 with medical record ≤20 years, 72 with medical record >20 years). h Women with ACE score of 1 or 2 and no abuse were excluded from these analyses ( n  = 193 overall, 93 with medical record ≤20 years, 100 with medical record >20 years). ACE and abuse characteristics of women at recruitment for the study (index date) overall and by length of medical record ACE adverse childhood experiences. a P values were obtained from chi-square tests or Fisher’s exact tests comparing women with medical record length ≤20 years and >20 years. P values were not reported when based on small numbers. b Any abuse includes verbal or emotional, physical, or sexual abuse. c Child harm includes verbal or emotional abuse, physical abuse, sexual abuse, emotional neglect, and physical neglect. d Household dysfunction includes parental separation, domestic violence, substance abuse, mental illness, and incarceration of a family member. e The p value for the categorical ACE score was obtained from an overall chi-square test across all categories. f Compared to women with ACE score of 0 and no abuse. g Women with ACE score of 1 and no abuse were excluded from these analyses ( n  = 144 overall, 72 with medical record ≤20 years, 72 with medical record >20 years). h Women with ACE score of 1 or 2 and no abuse were excluded from these analyses ( n  = 193 overall, 93 with medical record ≤20 years, 100 with medical record >20 years). Figure  1 and Supplementary Table  2 show the HRs and confidence intervals obtained for each one of the 18 conditions separately. The risks of 10 of the 18 conditions were significantly increased among women with ACE ≥ 2, viz., depression, schizophrenia or psychosis, substance abuse disorders, arrhythmias, coronary artery disease, congestive heart failure, asthma, chronic obstructive pulmonary disease, arthritis, and chronic kidney disease. In secondary analyses using one-tailed statistical tests, two additional chronic conditions, hypertension and hyperlipidemia, became statistically significant. Fig. 1 Risk of the 18 chronic conditions considered separately in women with ACE ≥ 2 versus ACE 0. The risk of the chronic conditions was represented graphically using a forest plot. The hazard ratio was significantly increased for 10 of the 18 conditions, namely, depression, schizophrenia or psychosis, substance abuse disorders, cardiac arrhythmias, coronary artery disease, congestive heart failure, asthma, chronic obstructive pulmonary disease, arthritis, and chronic kidney disease. The results for mental health disorders are shown in blue-gray, cardiovascular and metabolic conditions in dark red, and other somatic conditions in light blue. The error bars in the forest plot show the 95% confidence intervals for a maximum of n  = 814 women (detailed numbers for each chronic condition are reported in Supplementary Table  2 ). ACE adverse childhood experiences. The risk of the chronic conditions was represented graphically using a forest plot. The hazard ratio was significantly increased for 10 of the 18 conditions, namely, depression, schizophrenia or psychosis, substance abuse disorders, cardiac arrhythmias, coronary artery disease, congestive heart failure, asthma, chronic obstructive pulmonary disease, arthritis, and chronic kidney disease. The results for mental health disorders are shown in blue-gray, cardiovascular and metabolic conditions in dark red, and other somatic conditions in light blue. The error bars in the forest plot show the 95% confidence intervals for a maximum of n  = 814 women (detailed numbers for each chronic condition are reported in Supplementary Table  2 ). ACE adverse childhood experiences. Women with higher ACE scores, a history of child harm, a history of household dysfunction, and any abuse before the index date were at increased risk of accumulating morbidity during follow-up (Table  3 ). Figure  2 shows the cumulative incidence of chronic conditions for women with ACE score 1, 2, or ≥3 compared to women with a score of 0. The curves for women with higher ACE scores had higher morbidity scores at age 35 years (intercept with the y axis) and a more rapid accumulation during follow-up (panel b). There was a significant dose-effect trend of increasing HR with increasing number of ACE ( P  < 0.0001; panel a). Figure  3 shows the cumulative incidence of morbidity for different cut-offs of ACE scores compared to women with score of 0. Figure  3 also shows results for child harm, household dysfunction, any abuse before index age, and for ACE score ≥1, ≥2, or ≥3 or any abuse before index age compared to women with score of 0 and no abuse. Table 3 Accumulation of morbidity during follow-up by ACE score and by abuse before index age a ACE score and abuse N at risk Person-years N of events HR (95% CI) b P value ACE score cut-offs ACE 0 631 12,443 1420 -reference- ACE ≥ 1 395 8046 1139 1.23 (1.11–1.37) <0.0001 ACE ≥2 c 183 3702 595 1.41 (1.24–1.61) <0.0001 ACE ≥ 3 d 87 1667 309 1.64 (1.37–1.96) <0.0001 Child harm e None 866 17,163 2044 -reference- Any child harm 160 3326 515 1.31 (1.15–1.51) 0.0001 Household dysfunction f None 681 13,594 1566 -reference- Any household dysfunction 345 6895 993 1.24 (1.12–1.38) <0.0001 Any abuse before index g No abuse 774 15,324 1735 -reference- Any abuse 252 5165 824 1.41 (1.26–1.57) <0.0001 ACE score cut-off and abuse before index g ACE 0 and no abuse 575 11,348 1246 -reference- ACE ≥ 1 or any abuse 451 9142 1313 1.30 (1.17–1.43) <0.0001 ACE ≥ 2 or any abuse h 307 6303 971 1.39 (1.25–1.56) <0.0001 ACE ≥ 3 or any abuse i 258 5286 839 1.44 (1.28–1.62) <0.0001 ACE adverse childhood experience, HR hazard ratio, CI confidence interval. a Morbidity was defined using 18 chronic conditions considered as a score, with equal weighting. b Hazard ratios were calculated using Andersen–Gill regression models with age as the time scale. c Women with ACE score of 1 ( n  = 212) were excluded from this analysis. d Women with ACE score of 1 or 2 ( n  = 308) were excluded from this analysis. e Child harm includes verbal or emotional abuse, physical abuse, sexual abuse, emotional neglect, and physical neglect. f Household dysfunction includes parental separation, domestic violence, substance abuse, mental illness, and incarceration of a family member. g Includes verbal or emotional, physical, or sexual abuse during childhood (to age 19 years) or at any time before the index date. h Women with ACE score of 1 and no abuse ( n  = 144) were excluded from this analysis. i Women with ACE score of 1 or 2 and no abuse ( n  = 193) were excluded from this analysis. Fig. 2 Accumulation of the 18 chronic conditions considered as a score (with equal weighting) in four strata of women by ACE score (dose-effect trend). The accumulation of morbidity was represented graphically using ( a ) a forest plot and ( b ) Aalen–Johansen curves. Age started at 35 years because earlier ages had small numbers. The error bars in the forest plot show the 95% confidence intervals for n  = 1026 women. ACE adverse childhood experiences. Fig. 3 Accumulation of the 18 chronic conditions considered as a score (with equal weighting) in strata of women by ACE score and by occurrence of any abuse before the index age (abuse before or after age 19 years). The accumulation of morbidity was represented graphically using Aalen–Johansen curves for ( a ) ACE score ≥1 versus ACE score 0, ( b ) ACE score ≥2 versus ACE score 0, ( c ) ACE score ≥3 versus ACE score 0, ( d ) child harm versus none, ( e ) household dysfunction versus none, ( f ) any abuse before index versus no abuse, ( g ) ACE score ≥1 or any abuse before index versus ACE score 0 and no abuse, ( h ) ACE score ≥2 or any abuse before index versus ACE score 0 and no abuse, ( i ) ACE score ≥3 or any abuse before index versus ACE score 0 and no abuse. Age started at 35 years because earlier ages had small numbers. ACE adverse childhood experiences, HR hazard ratio. Accumulation of morbidity during follow-up by ACE score and by abuse before index age a ACE adverse childhood experience, HR hazard ratio, CI confidence interval. a Morbidity was defined using 18 chronic conditions considered as a score, with equal weighting. b Hazard ratios were calculated using Andersen–Gill regression models with age as the time scale. c Women with ACE score of 1 ( n  = 212) were excluded from this analysis. d Women with ACE score of 1 or 2 ( n  = 308) were excluded from this analysis. e Child harm includes verbal or emotional abuse, physical abuse, sexual abuse, emotional neglect, and physical neglect. f Household dysfunction includes parental separation, domestic violence, substance abuse, mental illness, and incarceration of a family member. g Includes verbal or emotional, physical, or sexual abuse during childhood (to age 19 years) or at any time before the index date. h Women with ACE score of 1 and no abuse ( n  = 144) were excluded from this analysis. i Women with ACE score of 1 or 2 and no abuse ( n  = 193) were excluded from this analysis. The accumulation of morbidity was represented graphically using ( a ) a forest plot and ( b ) Aalen–Johansen curves. Age started at 35 years because earlier ages had small numbers. The error bars in the forest plot show the 95% confidence intervals for n  = 1026 women. ACE adverse childhood experiences. The accumulation of morbidity was represented graphically using Aalen–Johansen curves for ( a ) ACE score ≥1 versus ACE score 0, ( b ) ACE score ≥2 versus ACE score 0, ( c ) ACE score ≥3 versus ACE score 0, ( d ) child harm versus none, ( e ) household dysfunction versus none, ( f ) any abuse before index versus no abuse, ( g ) ACE score ≥1 or any abuse before index versus ACE score 0 and no abuse, ( h ) ACE score ≥2 or any abuse before index versus ACE score 0 and no abuse, ( i ) ACE score ≥3 or any abuse before index versus ACE score 0 and no abuse. Age started at 35 years because earlier ages had small numbers. ACE adverse childhood experiences, HR hazard ratio. Table  4 shows the result of analyses exploring possible confounding variables. The association of ACE ≥ 2 with accelerated accumulation of morbidity remained significant after adjustment for 4 socioeconomic indicators, education, occupation, income, and area deprivation index (ADI), considered individually. Consistent with the lack of a confounding effect, the correlations of years of education, household income, and ADI with the ACE score were weak (Supplementary Fig.  1 ). Table  5 shows the result of analyses exploring possible mediation variables. Five variables, hysterectomy with ovarian conservation, BMI, cigarette smoking, eating disorders, and suicidal ideation, had small but significant indirect effects (mediation effects). Premenopausal bilateral oophorectomy did not have a significant indirect effect. The overall mediation effect including all 12 variables was 47.2%. However, the direct effect of ACE remained significant after including in the model the 12 possible mediation variables considered one at a time or combined into an overall multivariable model. Table 4 Accumulation of morbidity by ACE score including possible socioeconomic confounder variables in the models a Confounder variable Confounder variable ACE score ≥ 2 vs. ACE score 0 HR (95% CI) b P value HR (95% CI) b P value Univariable model – – 1.41 (1.24–1.61) <0.0001 Low education (≤12 years) 1.21 (1.07–1.38) 0.003 1.39 (1.22–1.58) <0.0001 Low income (<$42,000; Q1 vs. Q2–Q4) 1.17 (1.03–1.34) 0.02 1.40 (1.23–1.59) <0.0001 Unemployed 1.44 (1.05–1.98) 0.02 1.39 (1.22–1.59) 75 th national percentile) c 1.33 (1.09–1.61) 0.005 1.40 (1.23–1.60) <0.0001 ACE adverse childhood experience, HR hazard ratio, CI confidence interval, ADI area deprivation index. a Morbidity was defined using 18 chronic conditions considered as a score, with equal weighting. b Hazard ratios were calculated using Andersen–Gill regression models with age as the time scale. c High ADI corresponds to greater deprivation. Table 5 Accumulation of morbidity by ACE score including possible mediation variables in the models a ACE score ≥ 2 vs. ACE score 0 Mediation variable b Mediation variable Direct effect Indirect effect d HR (95% CI) c P value HR (95% CI) c P value HR (95% CI) P value Percent Never married 1.31 (1.10–1.56) 0.003 1.39 (1.22–1.59) <0.0001 1.01 (0.99–1.04) 0.12 3.8% Pregnancy before age 18 0.98 (0.70–1.36) 0.89 1.41 (1.24–1.61) <0.0001 0.99 (0.99–1.01) 0.91 – No live births 1.23 (1.05–1.43) 0.009 1.41 (1.24–1.61) <0.0001 0.99 (0.98–1.02) 0.78 – Bilateral oophorectomy 1.42 (0.91–2.23) 0.12 1.40 (1.23–1.60) <0.0001 1.01 (0.99–1.03) 0.36 2.2% Hysterectomy with ovarian conservation e 1.35 (1.09–1.66) 0.006 1.38 (1.21–1.56) <0.0001 1.02 (1.00–1.06) 0.03 7.0% Body mass index ≥30 kg/m 2 1.59 (1.41–1.78) <0.0001 1.34 (1.18–1.52) <0.0001 1.06 (1.01–1.10) 0.01 15.5% Cigarette smoking (current or former) 1.24 (1.10–1.39) 0.0004 1.34 (1.18–1.53) <0.0001 1.05 (1.02–1.09) <0.0001 14.0% Substance abuse disorders f 1.36 (0.95–1.94) 0.09 1.37 (1.20–1.56) <0.0001 1.03 (0.99–1.08) 0.13 9.4% Eating disorder 1.34 (1.07–1.68) 0.01 1.38 (1.21–1.57) <0.0001 1.02 (1.00–1.06) 0.03 6.9% Depression f 1.02 (0.86–1.21) 0.83 1.40 (1.22–1.62) <0.0001 1.01 (0.96–1.05) 0.81 1.4% Anxiety 1.14 (0.92–1.40) 0.23 1.39 (1.22–1.58) <0.0001 1.02 (0.99–1.05) 0.27 4.5% Suicidal ideation 1.32 (1.04–1.66) 0.02 1.32 (1.15–1.53) 0.0001 1.07 (1.01–1.14) 0.03 18.4% Full multivariable – – 1.20 (1.04–1.38) 0.01 1.18 (1.08–1.29) <0.0001 47.2% ACE adverse childhood experience, HR hazard ratio, CI confidence interval. a Morbidity was defined using 18 chronic conditions considered as a score, with equal weighting. b All mediation variables were assessed at the index date. c Hazard ratios were calculated using Andersen–Gill regression models with age as the time scale. d The total effect for ACE score ≥2 was an HR of 1.41 (95% CI 1.24–1.61; P  < 0.0001). The indirect effect was calculated by subtracting the direct effect beta coefficient from the total effect beta coefficient, and was expressed as an HR with bootstrapped 95% CI. We also computed a percent of mediation (indirect effect beta coefficient divided by the total effect beta coefficient). For pregnancy before age 18 and for no live births there was no indirect effect and no percent of mediation. e Hysterectomy performed with conservation of 1 or both ovaries. f Substance abuse disorders and depression were considered as possible mediation variables when present at the index date. However, they were also part of the 18 conditions included in the morbidity score. Accumulation of morbidity by ACE score including possible socioeconomic confounder variables in the models a ACE adverse childhood experience, HR hazard ratio, CI confidence interval, ADI area deprivation index. a Morbidity was defined using 18 chronic conditions considered as a score, with equal weighting. b Hazard ratios were calculated using Andersen–Gill regression models with age as the time scale. c High ADI corresponds to greater deprivation. Accumulation of morbidity by ACE score including possible mediation variables in the models a ACE adverse childhood experience, HR hazard ratio, CI confidence interval. a Morbidity was defined using 18 chronic conditions considered as a score, with equal weighting. b All mediation variables were assessed at the index date. c Hazard ratios were calculated using Andersen–Gill regression models with age as the time scale. d The total effect for ACE score ≥2 was an HR of 1.41 (95% CI 1.24–1.61; P  < 0.0001). The indirect effect was calculated by subtracting the direct effect beta coefficient from the total effect beta coefficient, and was expressed as an HR with bootstrapped 95% CI. We also computed a percent of mediation (indirect effect beta coefficient divided by the total effect beta coefficient). For pregnancy before age 18 and for no live births there was no indirect effect and no percent of mediation. e Hysterectomy performed with conservation of 1 or both ovaries. f Substance abuse disorders and depression were considered as possible mediation variables when present at the index date. However, they were also part of the 18 conditions included in the morbidity score.

Adverse childhood experiences and abuse in childhood or in early adulthood are associated with an accelerated accumulation of morbidity later in life. The association is not due to confounding by socioeconomic status and is only partly explained by a series of intermediate events or characteristics. Future studies should further investigate the mechanisms underlying the association. However, protection of children and adults from abuse and other adverse events will reduce morbidity regardless of the intervening mechanisms.

In this population-based sample, women with an ACE score ≥2 had higher incidence of 10 of the 18 chronic conditions considered separately and an accelerated accumulation of morbidity compared to women with ACE score of 0. In addition, women exposed to abuse in childhood or early adulthood had accelerated accumulation of morbidity. These associations were not explained by confounding by socioeconomic status and were only partially mediated by a series of 12 possible mediation events or characteristics. The World Mental Health Surveys conducted across 21 countries found little variation in ACE prevalence regardless of country income group (high, medium, or low income). An ACE score ≥1 was present in 38–39% of participants and a score of ≥4 in 2–3% 31 . In our study, a score ≥1 was present in 38.5% of women and a score of ≥4 in 4.8%. This comparison provides evidence that our assessment for ACE was reasonably complete and that our sample was representative of a general population. By contrast, the 37 studies included in the 2017 systematic review and meta-analysis of long-term sequelae of ACE were overall biased toward higher risk groups because a score ≥1 was present in 57% of women and a score of ≥4 in 13% 2 . The 37 studies in the 2017 review were performed in the United States, United Kingdom, and in several other countries with high-income or middle-income. None were from low-income countries. Twenty-six of the studies had a cross-sectional design and only 11 were cohort studies. In addition, all 37 studies used retrospective self-reported ACE collected at a single point in time. There was a wide variability in the number and list of ACEs considered, and in the number and list of outcomes considered. Men and women were most commonly considered together, and the cut-off for the ACE score was more extreme (ACE ≥ 4) 2 . Therefore, the comparability of our study with the overall results of the 2017 review is limited. Nevertheless, the 2017 review reported significant associations of ACE with five somatic chronic diseases (liver and digestive diseases, diabetes, cardiovascular diseases, cancer, and respiratory diseases), and with several mental health outcomes (depression, anxiety, and substance abuse disorders) 2 . Our results are consistent for cardiovascular disease (arrhythmias, coronary artery disease, and congestive heart failure), respiratory disease (asthma and chronic obstructive pulmonary disease), and mental health disorders (depression, and substance abuse disorders). Our sample size and statistical power was limited compared to the 2017 review, and we used a less extreme cut-off of ACE ≥ 2. Our numbers were too small for using the cut-off of ACE ≥ 4. Our study focused specifically on women because there is some evidence that ACE are more common in women and may have a stronger impact on women compared to men 7 , 8 . In addition, ACE have been associated with an increased risk of premenopausal bilateral oophorectomy, and premenopausal bilateral oophorectomy has been associated with a higher risk of accumulation of chronic conditions 12 , 19 , 32 , 33 . However, in our mediation analyses, the association of ACE with morbidity was not attenuated significantly when bilateral oophorectomy present at index date was included in the model, thus suggesting that bilateral oophorectomy was not a major mediation event. On the other hand, the UK Biobank study reported an association of a cumulative adversity score with acceleration of phenotypic aging that was stronger among men than women 34 . Therefore, more research is needed on sex and gender differences. This study was designed to measure associations and to explore possible confounding or mediation effects at the population level. However, we did not collect data that can address the underlying mechanisms. Evidence from mechanistic research at the cellular, tissue, organ, or system level suggests that the association may be mediated by neurological, hormonal, immunological, and epigenetic changes occurring in persons who experienced ACE 34 , 35 . Some studies have associated early trauma with acceleration of aging measured using deoxyribonucleic acid (DNA) methylation markers. Hamlat and colleagues reported that early life abuse, but not early life neglect, was associated with accelerated aging measured using an aging clock (DNAm GrimAge). This clock in turn predicted time-to-death and time-to-coronary heart disease and was associated with age at menopause. It remains unclear whether the epigenetic age acceleration observed was a true mediator of the effect of early trauma that caused poor health and premature death or whether it was simply a marker of other aging processes 35 . If epigenetic aging is a direct biological mediation mechanism, interventions to slow DNA methylation may prevent the harmful long-term effects of early life trauma on morbidity and mortality. In the UK Biobank study, childhood adversity was significantly associated with accelerated aging measured using a phenotypic age acceleration marker. Childhood adversity was defined as physical neglect, emotional neglect, sexual abuse, physical abuse, and emotional abuse and corresponds to our definition of child harm. In addition, the association was partly mediated by unhealthy lifestyle. They used an unhealthy lifestyle score combining BMI, smoking status, alcohol consumption, physical activity, and diet 34 . In our study, BMI and smoking that were present at the index date had small but statistically significant mediation effects, whereas substance abuse disorders (primarily alcohol in this Minnesota population) did not have a significant mediation effect. However, we did not have information about physical activity and diet, and we did not combine the behaviors into a score. This study has several strengths. In particular, the sample can be considered representative of a defined population of the United States (US) upper Midwest (population-based study) 36 . Second, the REP records-linkage system was unique in providing access to historical medical data across multiple healthcare institutions and over more than half a century. We were able to link events that occurred many decades apart and that were documented in records spread across pediatric, psychiatric, reproductive, preventive medicine, or gynecological specialties. The median length of medical records in the REP at the index year was ~20 years. In addition, for 336 women in the study (32.7%), the REP included medical records dating back to age 15 years or earlier; therefore, there were at least 4 years of concurrent capture of childhood experiences (age 15 through 18 years). For 185 women in the study (18.0%), the medical records dated back to birth. On the other hand, the collection of ACE cannot be considered entirely prospective. The ACE were documented in the records at variable ages and across multiple medical encounters; however, there was always a retrospective component in the acquisition of information. This study has some limitations. First, our detection of ACE was imperfect, and we may have missed some ACE in the abstraction assessment. We likely have underestimated the frequency of adverse experiences similar to what has been observed in other studies 1 , 3 . Whether data are collected through an interview, a mail survey, or through medical record abstraction, the risk of underestimation is inherent to the sensitivity of the information. Second, the abstraction of ACE from medical records is a time-consuming process and involves some level of judgment. We developed specific guidelines to obtain and interpret the data, as detailed elsewhere 19 . In addition, both intra-rater and inter-rater reliability were excellent in two previous studies 19 , 20 . Third, consistent with the original use of the ACE score, each one of the 10 items was given the same weight when we used cut-offs 1 , 15 , 16 . However, some ACE may have been more emotionally traumatic than others. Finally, the same experience may have had a greater or lesser impact depending on other factors such as the entire family environment (e.g., the presence of grandparents or other surrogate caring figures). Fourth, the population of Olmsted County is primarily white of European ancestry, and our findings may not be generalizable to other populations with a different racial or ethnic composition 36 . Finally, some of the stratified analyses or the analyses for individual chronic conditions were based on small numbers (limited statistical power). We conducted a set of sensitivity analyses using one-tailed statistical testing and were able to detect significant differences in risk for two additional chronic conditions, hypertension and hyperlipidemia.

The impact of early-life traumatic experiences on late-life morbidity remains partly unknown. A 2017 systematic review and meta-analysis provides a good summary of the sizeable and rapidly expanding body of evidence that has accumulated since the pioneering work of Felitti and coworkers in 1998 1 . This body of evidence has linked adverse childhood experiences (ACE) with a variety of diseases and health conditions later in life 1 , 2 . However, all 35 studies included in the systematic review were based on a retrospective self-report of ACE at a single point in time. Unfortunately, the presence of a late-life adverse outcome may motivate a selective recall and the willingness to report ACE thus causing a recall bias in case-control studies 3 – 6 . In addition, there was a wide variability in the number and list of ACE considered and in the number and list of outcomes considered across different studies, and men and women were often considered together. Differences between men and women may be important because there is some evidence that ACE are more common in women and may have a stronger impact on women compared to men 7 , 8 . We tested the hypothesis that adverse childhood experiences (ACE) and specifically abuse in childhood or early adulthood are associated with a higher rate of accumulation of 18 chronic conditions considered individually and combined into a morbidity score. A higher rate of accumulation of chronic conditions has been considered a clinical marker of accelerated aging 9 – 11 . We used the original 10-item ACE score, we assessed ACE using medical records abstraction in a records-linkage system, we considered the 18 most common and most important chronic conditions as outcomes, we focused on women, and we used a population-based sample to reduce possible selection biases. In this sample of women living in a US Midwestern County, the experience of ACE or of abuse in childhood or in early adulthood is associated with an accelerated accumulation of morbidity later in life. We show that the association is not explained by differences in socioeconomic status and is only partly mediated by a series of intermediate events or characteristics.

Transparent Peer Review file Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Reporting Summary Transparent Peer Review file Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Reporting Summary