Early Life Conditions of Overall and Cause-Specific Mortality Among Inner-City African Americans

Abstract

Objectives. We examined how early life conditions influence midlife overall and cause-specific mortality in a community cohort of disadvantaged African Americans.

Methods. Using a prospective design, we assessed first-grade children and their teachers and families when children were 6 years old, with follow-up at ages 16, 32, and 42 years. We obtained information on death from family members, neighbors, and the National Death Index (NDI). We conducted a survival analysis and competing risk analysis to examine early life predictors of mortality.

Results. Of 1242 participants, 87 (7%) had died by 2004. In multivariate Cox proportional hazards regression, males who lived in foster care and females with lower math grades in first grade were more likely to die by age 42 years. In multivariate competing risks analysis, hospitalization by the time of first grade was related to mortality from acute and chronic illness. Male gender, being in foster care, and aggressive behavior in first grade were related to mortality from drug use, violence, or suicide.

Conclusions. Early classroom, environmental, and family-level interventions are potentially beneficial in reducing later overall and cause-specific mortality.

African Americans historically have had the highest mortality rates among American racial and ethnic groups.¹ This gap has remained fairly stable over the last 40 years and may have even widened between Whites and African Americans.^1–3 Satcher et al. estimated that, as of 2002, approximately 83 570 excess deaths could be avoided by eliminating this disparity between African Americans and Whites.³ African Americans also have much higher years of potential life lost, a measure that takes into account the years of life lost due to early deaths.^4,5 Years of potential life lost is an especially important comparison measure because it documents early mortality. African Americans have higher mortality for most leading causes of death in the United States, including heart disease, stroke, and cancers, as well as homicide and HIV/AIDS.¹ Many of these racial differences in mortality are attributed to inequalities in economic status, education, and occupation.^1,5–7 However, few studies have examined the effect of conditions and behaviors as early as age 6 on later overall and cause-specific mortality among African Americans.

Research on early factors as related to mortality has primarily focused on infant and childhood mortality rather than mortality that extends into early and mid-adulthood. However, the possible impact of early childhood circumstances on longevity into early and mid-adulthood has been understudied. Early life circumstances initiate trajectories that progressively may increase the risks for later morbidity and mortality, making the identification of early life risk factors important for understanding later mortality. Early life factors such as low socioeconomic status (SES),^1,6–8 being in foster care,^5,9 childhood cognitive ability (i.e., childhood IQ),^10–12 and poor childhood health^13,14 have been shown to increase the risk of later mortality. However, much of the work thus far has retrospectively assessed early life circumstances using reports from older adults, making these findings subject to survivor and recall bias.

We extended and updated previous research by prospectively examining the impact of childhood socioeconomic and family disadvantage, early cognitive and classroom behavior, and physical and mental health on overall and cause-specific time to midlife mortality among a community of urban African Americans younger than 42 years. We used data from a longitudinal cohort study, the Woodlawn Project.

METHODS

This prospective, longitudinal study has followed a cohort of 1242 first-grade children in Woodlawn, an African American, disadvantaged community on the south side of Chicago, Illinois. At the time of initial assessment in 1966, the Woodlawn community was the fifth poorest of the 76 Chicago communities and more than 99% of the first graders were African American. In 1966 and 1967, there were no agreed-upon procedures for gaining the consent of human participants, nor were study protocols evaluated by institutional review boards for the protection of human participants; however, there was a community board that oversaw all the study procedures.¹⁵ In the initial contact, participation was voluntary, with 13 of the children’s families choosing not to participate. First-grade teachers were asked about each child’s classroom behavior, and mothers (or mother surrogates) were interviewed about their child and their family. There have been 3 further assessments of this cohort: In 1975 and 1976, 939 of the mothers or mother surrogates and 705 of the adolescents were reinterviewed. In 1992 and 1993 (at ages 32–33 years), 952 cohort members were interviewed, and in 2003 (at ages 42–43 years), 833 were reinterviewed.¹⁶ At these times, full informed consent was obtained from the cohort members and from the mothers during the adolescent assessment. In these assessments, the study procedures were approved by an institutional review board. During the adolescent assessment, the institutional review board was at the University of Chicago, and for later assessments the Johns Hopkins School of Hygiene and Public Health (later named Bloomberg School of Public Health) reviewed and approved the study procedures.

Mortality

We obtained information on mortality from several sources. First, in the process of locating the population for the adolescent, young adult, and mid-adult follow-ups, we obtained reports of deaths from family members and neighbors. As far as was possible and appropriate, we recorded date, place, and cause of death. Second, we submitted these names and the names of all those we could not locate to the National Death Index (NDI) and searched the records from 1979 to 2002. The NDI records begin with the year 1979; deaths occurring before this time were not recorded in the NDI. We then determined whether the matches received from the NDI were or were not cohort members. Third, for positive matches, we obtained the cause of death from the death certificate from the state where death occurred. Most reports of deaths that were made by family members or neighbors were confirmed by actual death records obtained from the NDI. This was not the case for those few deaths (n = 5) that occurred prior to the establishment of the NDI in 1979. Here, we relied on the reports of family members at the time of the adolescent assessment. In 1 case of death occurring after 1979, we counted the individual as dead even though he was not found in the NDI because (1) we had multiple and independent reports of the death from family members and friends, (2) it was reported by them as a recent death and might not have yet been recorded in the NDI, and (3) the individual had changed names several times and may have been listed by a name unknown to us. For those 4 individuals without any information of age at death, we imputed values for age at death using the midpoint between the interviews (i.e., between the ages of 6 and 16 years).¹⁷

We attempted to categorize the causes of death using information from the death certificate and, when appropriate, with from the information provided by relatives. Among the 87 deaths, more than one third were from acute and chronic illness such as bronchopneumonia, heart disease, cancer, sickle cell disease, liver disease, and diabetes mellitus (35.6%). Other causes of death included homicide (13.8%), HIV/AIDS (12.6%), drugs or alcohol (10.4%), and trauma (drowning, automobile fatalities; 10.4%; Table 1).

TABLE 1—

Distribution of Deaths by Deceased Status as of 2004: The Woodlawn Project, Chicago, IL

Status	Sample, No. (%)	% of Deaths
Alive in 2004	1155 (93)
Dead in 2004	87 (7)	100
Date of death known (NDI)	67	77
Death before 1975 (NDI not available)	5a	5.7
Reported death no NDIb	15c	17.3
Cause of death (n = 87)
Acute and chronic illness	31 (35.6)
Homicide	12 (13.8)
HIV/AIDS	11 (12.6)
Drugs or alcohol	9 (10.4)
Trauma	9 (10.4)
Suicide	5 (5.7)
Unknown	10 (11.5)

Note. NDI = National Death Index. The total sample was n = 1242.

^aAge of death was known for 3 participants and imputed for 2 participants.

^bNDI file is available beginning in 1979.

^cAge of death was known for 13 participants and imputed for 2 participants.

Measures of Early Conditions and Behaviors

For the present analyses, we chose several individual and family background characteristics measured at the time of the first assessment of the cohort (aged 6–7 years) on the basis of previous literature examining early childhood influences on mortality and previous research of the Woodlawn cohort. Other studies have found poverty or SES, cognitive ability, early health, and family characteristics to be important for later mortality or health outcomes, and those are the indicators that we examine here.^10–13,18 In addition, we included early behavior ratings in first grade, as our past work has shown their importance for drug use, delinquency, and graduation from high school, all of which may affect later mortality and health.^16,19–22

Early family characteristics.

We measured family structure by mother’s report of adults who were living in the child’s household. For our analyses, we categorized these into 1 of 3 family types—mother as the sole caregiver (“mother alone”), mother not present as a caregiver (“mother absent”), or mother along with another caregiver such as father, grandmother, or aunt (“mother other”). In earlier analyses, we found that children from mother–father families performed better in first grade than children from mother-alone families. Because children from mother-other families did almost as well,²³ however, we combined mother–father and mother–other family categories. On the basis of univariate analyses of family structure and mortality, we further subdivided the mother-absent category into foster families with a nonrelative caregiver (1.9%) or families in which a family member other than the mother was the caregiver—for example, grandmother, father, or aunt (6.8%).

Early socioeconomic conditions.

Early socioeconomic disadvantage was indicated by living in a family with an income below 100% of the 1965 Federal Poverty Guideline. This indicator takes into account family income, size, and the inflation rate.²⁴ When income was missing, poverty was assumed if the mothers reported that their income was mainly from receiving welfare.

Cognitive performance.

We measured cognitive performance using first-grade math grade (0 = A or B; 1 = C or D) as graded by teachers. The Kuhlmann-Anderson IQ test was administered in the public schools but not the parochial schools, and so was not included in the analyses. The IQ scores were highly correlated with math grades (r = 0.54; P < .001).

Child behavior and health.

The Teachers’ Observations of Classroom Adaptation scales are teacher-rated measures of students’ level of social adaptation in the classroom.¹⁵ These scales were developed through use of teachers’ responses to questions about which tasks were most important for students’ success in school. There were 5 items—social contact (shy behavior), authority acceptance (aggressive behavior), maturation, concentration, and achievement. In past studies, we have found that shy and aggressive behaviors have been important for later behavioral and health outcomes.^16,19–22 We included the combination of shy and aggressive behavior as rated by teachers as an additional measure (0 = neither; 1 = shy only; 2 = aggressive only; 3 = both shy and aggressive). We adapted the Mother’s Symptom Inventory, which includes 38 items, from an instrument developed by Connors²⁵ to assess children’s psychological symptoms. The mothers were asked to rate their children on a 4-point scale from “not at all” (= 0) to “very much” (= 3) on items reflecting emotional difficulties such as clinging too much, afraid to go to school, and having temper tantrums. The Mother’s Symptom Inventory has been shown to be a reliable instrument (α = 0.83) in assessing children’s mental health symptoms.^9,21 We categorized scores into 3 groups based on tertiles. For child hospitalization, we asked each mother whether her child, by first grade, had ever been hospitalized overnight (1 = yes, 0 = no).

Statistical Analysis

In all analyses of mortality, members of the cohort whose statuses were unknown (n = 185)¹⁷ were treated as “alive,” thereby ensuring that any association between early characteristics and subsequent mortality would be conservative. An initial descriptive analysis examined the distribution of early conditions and overall mortality. We used survival analysis methods to investigate the cumulative probability of mortality and to assess how early childhood factors were related to time to mortality. We used the log-rank procedure to compare cumulative mortality curves.

We used Cox proportional hazard regression models to assess how the influence of early conditions were related to time to overall mortality. Additionally, we fit models that included interaction terms for gender by other covariates to examine whether there were gender differences in early conditions for the outcome.

We also estimated the model for all competing risks. The fundamental concept in competing risk models is the cause-specific hazard function, the hazard of death from a given cause in the presence of the competing events. Those with a specific cause of death are compared with all those who died, but specific causes of death are not compared. Specifically, we examined mortality from acute or chronic diseases and mortality from HIV, drug or alcohol use, homicide, suicide, or trauma. We combined HIV, drug or alcohol use, homicide, suicide, and trauma because of the association of these deaths with risk behaviors such as substance use, unprotected sexual behavior, and aggressive behavior. This is not to imply that other acute and chronic diseases are unrelated to risk behaviors; for example, risk factors for these diseases include tobacco use and diet and activity patterns.²⁶ We recognize that most causes of death are multifactorial in nature; we have tried to sort out a primary cause as indicated on the death certificate.

Analysis using list-wise deletion of cases with missing data would have limited our sample to 1146. We used the Stata ICE (imputation by chained equations) module, version 11 (StataCorp LP, College Station, TX),²⁷ creating 10 imputed data sets for the 1242 cases who had mortality data at age 42 years.

RESULTS

Of 1242 participants, 87 deaths (7%) were reported by 2004. Males had higher mortality than females (10.2% vs 3.9%; P < .001). We computed the total analysis time at risk (54 656 person-years). The distribution of child and family characteristics is shown in Table 2.

TABLE 2—

Sample Characteristics of African Americans in a Prospective Study of Childhood and Family: The Woodlawn Project, Chicago, IL, 1967

Characteristics	No. (%)
Gender
Female	636 (51.2)
Male	606 (48.8)
Poverty or welfare receipt (n = 1189)
No	583 (49.0)
Yes	606 (51.0)
Family type
Mother with other adults	699 (56.3)
Mother alone	459 (37.0)
Mother absent (family member as caregiver)	61 (4.9)
Foster care	23 (1.9)
First-grade math (n = 1153)
A or B	528 (45.8)
C or D	625 (54.2)
Shy and aggressive behavior
Neither	646 (52.2)
Shy only	202 (16.3)
Aggressive only	211 (17.0)
Both shy and aggressive	179 (14.5)
Mother’s Symptom Inventory score (n = 1232)
Low	470 (38.1)
Medium	425 (34.5)
High	337 (27.4)
Hospitalization by first grade
No	991 (79.8)
Yes	251 (20.2)

Note. The sample size was n = 1242. Frequencies do not always sum to 1242 because of missing responses. Totals may not add to 100% because of rounding.

In bivariate Cox proportional hazard regression analyses, those who were male, were in a foster home at the time of first grade, had a low first-grade math grade, had more reports of psychiatric symptoms as made by mothers, and had been hospitalized between birth and first grade were more likely to have died (P < .05).

The risks of mortality as indicated by the multivariate Cox proportional hazard regression analyses are also presented in Table 3. Because the interaction between gender and math grades was statistically significant, we conducted separate analyses by gender. For males, being in a foster family (hazard ratio [HR] = 4.17; 95% confidence interval [CI] = 1.45, 11.93) was an important predictor of overall premature mortality, whereas having a low math grade (HR = 3.51; 95% CI = 1.16, 10.58) was related to later overall mortality among females.

TABLE 3—

Cox Proportional Hazard Ratios for Early Life Conditions of Mortality: The Woodlawn Project, Chicago, IL, 2004

				Adjusted HRa (95% CI)
Predictors	No. of Participants	No. of Deaths	Crude HR (95% CI)	Males (n = 606)	Females (n = 636)
Gender
Female (Ref)	636	25	1.00	NA	NA
Male	606	62	2.70* (1.70, 4.30)	NA	NA
Early SES: poverty or welfare receipt
No (Ref)	583	38	1.00	1.00	1.00
Yes	606	49	0.94 (0.61, 1.45)	1.43 (0.85, 2.42)	0.50 (0.21, 1.15)
Family type
Mother with other adults (Ref)	699	42	1.00	NA	NA
Mother alone	459	36	1.33 (0.85, 2.07)	NA	NA
Mother absent	84	9	1.83 (0.89, 3.75)	NA	NA
Foster care				NA	NA
No (Ref)	1219	82	1.00	1.00	1.00
Yes	23	5	3.61* (1.46, 8.90)	4.17** (1.45, 11.93)	2.22 (0.29, 17.30)
Cognitive ability: first-grade math
A or B (Ref)	528	27	1.00	1.00	1.00
C or D	625	53	1.68* (1.06, 2.68)	1.09 (0.63, 1.89)	3.51* (1.16, 10.58)
Shy and aggressive behavior
Neither (Ref)	646	42	1.00	1.00	1.00
Shy only	202	15	1.15 (0.64, 2.07)	1.08 (0.53, 2.20)	0.87 (0.28, 2.71)
Aggressive only	211	19	1.41 (0.82, 2.42)	1.13 (0.59, 2.16)	1.23 (0.43, 3.53)
Both shy and aggressive	179	11	0.94 (0.49, 1.83)	0.77 (0.35, 1.68)	0.68 (0.15, 3.13)
Mother’s Symptom Inventory score
Low (Ref)	470	26	1.00	1.00	1.00
Medium	425	30	1.29 (0.76, 2.17)	1.10 (0.59, 2.04)	1.64 (0.57, 4.77)
High	337	31	1.69* (1.01, 2.85)	1.58 (0.84, 2.96)	1.86 (0.66, 5.24)
Hospitalization by first grade
No (Ref)	991	62	1.00	1.00	1.00
Yes	251	25	1.61* (1.01, 2.56)	1.24 (0.70, 2.20)	2.02 (0.87, 4.73)
No. of deaths			87	62	25

Note. CI = confidence interval; HR = hazard ratio; NA = not applicable; SES = socioeconomic status. The sample size was n = 1242 participants and n = 87 deaths.

^aUsed multiple imputations, and adjusted for poverty, foster care, math grade, shy and aggressive behavior, mother’s symptom inventory, and hospitalization.

*P < .05; **P < .01.

Table 4 shows the multivariate competing risks analysis using the Cox proportional hazard model. Hospitalization by first grade was related to acute and chronic mortality (HR = 2.63; 95% CI = 1.27, 5.45). Mortality from HIV, drug or alcohol use, homicide, suicide, or trauma was related to male gender (HR = 4.21; 95% CI = 2.02, 8.79), being in foster care (HR = 5.47; 95% CI = 1.90, 15.80), and aggressive behavior in first grade (HR = 2.01; 95% CI = 1.03, 3.99).

TABLE 4—

Regression Models of Multivariable Competing Risks for Prediction of Specific-Cause Mortality: The Woodlawn Project, Chicago, IL, 2004

Predictors	Mortality From Acute and Chronic Illness, HR (95% CI)	Mortality From Risky Behaviors,a HR (95% CI)
Gender
Female (Ref)	1.00	1.00
Male	1.53 (0.74, 3.15)	4.21* (2.02, 8.79)
Poverty or welfare receipt
No (Ref)	1.00	1.00
Yes	0.99 (0.48, 2.03)	1.38 (0.75, 2.52)
Foster care
No (Ref)	1.00	1.00
Yes	2.07 (0.28, 15.48)	5.47* (1.90, 15.80)
First-grade math
A or B (Ref)	1.00	1.00
C or D	1.70 (0.75, 3.83)	1.33 (0.68, 2.61)
Shy and aggressive behavior
Neither (Ref)	1.00	1.00
Shy behavior only	1.30 (0.54, 3.12)	0.95 (0.37, 2.43)
Aggressive behavior only	0.48 (0.14, 1.68)	2.01* (1.01, 3.99)
Both shy and aggressive	0.64 (0.21, 1.99)	0.87 (0.34, 2.26)
Mother’s Symptom Inventory score
Low (Ref)	1.00	1.00
Medium	1.30 (0.52, 3.20)	1.34 (0.66, 2.70)
High	1.59 (0.66, 3.85)	1.49 (0.70, 3.16)
Hospitalization by first grade
No (Ref)	1.00	1.00
Yes	2.63* (1.27, 5.45)	0.97 (0.48, 1.99)
No. of deaths	31	36

Note. CI = confidence interval; HR = hazard ratio. We used multiple imputations, and adjusted for gender, poverty, foster care, math grade, shy and aggressive behavior, mother’s symptom inventory, and hospitalization. The total sample was n = 1242.

^aRisk factors were HIV/AIDS, homicide, suicide, drugs, alcohol, and trauma.

*P < .05.

DISCUSSION

Our goal was to examine how factors early in the life course contributed to premature mortality among a cohort of African Americans. Of the original 1242 Woodlawn cohort members, 87 (7%) died by age 42 years, with about one third of deaths attributed to acute and chronic illnesses and many others related to violence, substance use, and risky sexual behavior. The mortality rates for this cohort were higher than the national rates; 10.2% of the Woodlawn males died between the ages of 6 and 42 years, compared with a rate of 6.7% in the comparable life table for African American males. For females, the Woodlawn mortality rate was 3.9%, compared with 3.2% in the comparable life table for African American females. The comparable life table death rates between the ages of 6 and 42 years were 3.7% for White males and 1.8% for White females.²⁸

We found that several family and behavioral factors early in life affect both overall and cause-specific mortality later in adulthood. Our findings emphasize that early family environment, such as being in foster care, and behavioral and physical health indicators at a young age have a significant impact on mortality.

These analyses highlight the considerable risk of early death to which children in foster care are exposed. Previous analysis of this cohort found a relationship between being in foster care and early mortality,⁹ which was supported in this updated analysis. Although the number of this cohort’s children in foster care at age 6 to 7 years was small in absolute terms (23, or 1.9%), the odds of dying relative to the others in the cohort were elevated by a factor of 4. Over 20% of the children in foster care died before the age of 42 years. Of the 5 foster children who died, 1 died from motor vehicle accident, 2 from HIV-related illnesses, 1 from drug overdose or possible suicide, and 1 from homicide or possible suicide. Our findings indicated that children in foster care were at high risk for mortality from violence, substance use, and suicide. Previous research suggests that mental and physical health may be potential mediating factors in this relationship.²⁹ In an examination of the impact of foster care on adult health, Zlotnick et al.²⁹ found that adults with histories of foster care had twice the odds of receiving Social Security Disability Insurance compared with those without such histories. It is also possible that foster parents are not as protective of foster children in their interactions with others as other parents are. In our earlier study, we raised the concern of whether foster care plays a causal role in the premature death of foster children⁹ or whether placement in foster care reflects problems already occurring in the lives of those children (e.g., death of parents or abusive parents). Further research is clearly warranted to examine the broader health and well-being outcomes for children in foster care, and the pathways that may contribute to this increased morbidity and mortality.

We need more examination of those characteristics of foster care placement that are supposed to protect these children, who may be quite vulnerable to poor health or social maladaptation. Such studies are difficult given the need for a longitudinal perspective to answer this question and the relatively small number of foster children in any one community. Studies such as the Woodlawn Project are too small to definitively answer this question, and larger-scale studies at state and national levels are needed. The US Department of Health and Human Services has estimated that 547 000 children in the United States were in foster care in 2000, of whom 41% were African American, 40% were White, and 15% were Hispanic.³⁰ This represents a significant population at risk for physical, mental, and social problems, which may lead to excess mortality. Further research is critical given that foster children are easily identifiable and can be targeted for appropriate treatment and preventive interventions. Chamberlain et al.³¹ have developed a preventive intervention for children in foster care that has been shown to be effective in reducing problem behaviors, which may further affect health and premature mortality.

Although African American females are at lower risk for mortality than African American males, they are still an important and understudied group. Females exhibited different risk factors, with cognitive performance being the most salient predictor of premature mortality. In this study, we used math grade in first grade as the measure of cognitive performance. It is highly correlated with both reading grades and IQ test scores. Large cohort studies have similarly found cognitive ability to be a significant predictor of mortality,^10–12 although it remains unclear what may mediate this relationship. Some have suggested it might be mediated by persistent disadvantage over the life course, including more dangerous occupations and living conditions.¹¹ This cumulative disadvantage may directly expose one to more dangerous situations that put one at risk for death.

Additionally, we found that children rated by their teachers as aggressive in first grade were more likely to die from homicide, suicide, trauma, or HIV. Previous research of this cohort has shown that aggressive behavior in first grade is related to later substance use,¹⁹ low educational attainment,²⁰ and crime.²² Aggressiveness in first grade may be connected to early mortality through these risk factors of increased antisocial behavior and lowered achievement. This suggests the need for early screening and intervention for young children, particularly those from high-risk environments. Several early prevention and intervention programs have targeted early aggressive behavior and have been successful in reducing aggression, with lower risks of later problem behaviors such as smoking, criminal justice involvement, and alcohol use.³²

As expected, males had significantly lower survival rates than females. It is established that African American urban males are at high risk for early mortality, and this study further emphasizes this fact.⁵ Sadly, the predominance of violence and drug-related deaths found in this cohort is fairly representative of the major causes of death among young and middle-aged African American males in the United States.² This again stresses the vast need for understanding the root causes of this disparity.

We did not find that early socioeconomic disadvantage, as indicated by poverty or welfare receipt, significantly differentiated those who died by age 42 years from those who did not. Most past studies of childhood predictors of mortality have found that children from low-SES families had a higher risk of early mortality. In this study, although the odds of early death were higher among those who were in poor families than among those who were not, these odds were not statistically significant. We have 2 possible explanations. First, the cohort members were all from a disadvantaged community initially. There were variations in SES among the cohort members, but the range was not as great as it might have been in nationally based populations. Second, given the relatively small size of the cohort, the power to detect these differences might have been lacking.

Limitations

This study has many benefits from using longitudinal data on an important and understudied population, but there are some limitations. First, as study participants were all African Americans from a single urban community in Chicago, our findings may not be generalizable to other populations. However, our study is valuable because there are very few prospective cohort studies across the life course of African Americans (who have high rates of premature mortality) whose design allows the study of childhood origins of later mortality. Another limitation is the small size of the cohort for analyzing mortality. However, we did find important early predictors of mortality. Despite these limitations, this study highlights the importance of early life circumstances in understanding overall and cause-specific mortality among African Americans. Future research is crucial to understanding the mechanisms of these relationships to prevent excess mortality in this population.

Conclusions

We found that there were high rates of death before the age of 42 years in a cohort of African Americans followed longitudinally from age 6 to 42 years. Children in foster care were especially vulnerable to early mortality, highlighting the need for increased social service and public health vigilance for foster care children. We found that for males, aggressive behavior in first grade was related to later mortality from homicide, drug use, suicide, and trauma. This finding highlights the risk that early aggressive behavior entails for children, especially males; several prevention programs aimed at early aggressive behavior led to lower rates of later drug use, crime, and school dropout. These prevention programs may also be protective for premature mortality. Finally, these findings highlight the importance of early life conditions in setting a trajectory to later health and well-being.