Are
Children
Born
to Young
Mothers
at Increased
Risk
of Maltreatment?
David M. Stier, MD*; John M. Leventhal, MD*; Anne T. Berg, PhD*; Lyla Johnson, RN*; and
JoAnne Mezger”
ABSTRACF. Previous case-control or cross-sectional
studies have provided conflicting results about whether
children of teenage mothers are at increased risk of mal-treatment compared with children of older mothers. This study is the first to examine this question using a longi-tudinal, cohort design and the first to address important
methodobogic issues such as detection bias. Subjects
were 219 consecutive index children born to inner-city
women who were 18 years or younger and 219
sociode-mographically similar comparison children born to
women 19 years or older. Data were collected by
review-ing the medical records of each child through the fifth birthday. Three outcomes were examined: maltreatment,
poor growth, and a change in the child’s primary
care-taker. Maltreatment was ascertained by having two
ex-perts, one of whom was blind to the group status, review
each injury documented in the records. Predefined
crite-na were used to distinguish unintentional injuries from
maltreatment (abuse, neglect, or sexual abuse).
Maltreat-ment occurred more frequently in the children of young
mothers (12.8%) than in the comparison group (6.4%)
(risk ratio IRRI = 200; 95% confidence interval [CI] =
1.17, 3.64). Poor growth, defined by growth criteria,
oc-curred in 6.9% of the index group and in 41% of
corn-parison children (RR = 1.67; 95% CI = 0.75, 3.73). A
change in the child’s primary caretaker, either because of
placement in foster care or because the mother left the
home, occurred in 12.8% of the index group and in 3.2%
in comparison children (RR = 4.00; 95% CI = 1.80, 8.87).
Children of teenage mothers, compared with children of
older mothers, are at increased risk of maltreatment and
of changes in their primary caretakers. Pediatrics 1993;
91:642-648; child maltreatment, child abuse, teenage
preg-nancy, adolescence, parenting.
ABBREVIATIONS. YNHH, Yale-New Haven Hospital; FT1 failure to thrive; RR, risk ratio; CI, confidence interval.
Teenage pregnancy continues to be a major social
and medical problem in this country.” In 1988, 13%
of births nationally were to women younger than 20
years of age.3 This rate is higher in urban centers; for
example, in New Haven, CT, in 1988, the comparable
figure was 18% of all births, and 24% of births to
From the *paetment of Pediatrics and Child Study Center, Yale
Univer-sity School of Medicine, New Haven, CT.
Received for publication May 22, 1992; accepted Sep 24, 1992.
Presented, in part, at the 25th Annual Meeting of Ambulatory Pediatric Association, Washington, DC, May 1988.
Dr Stier’s address is Dept of Pediatrics, University of California SF, San
Francisco, CA 94143.
Reprint requests to (J.M.L) Dept of Pediatrics, Yale University School of
Medicine, 333 Cedar St, New Haven, CT 06510.
PEDIATRICS (ISSN 0031 4005). Copyright lb 1993 by the American
Acad-emy of Pediatrics.
nonwhite mothers were to women younger than 20
years of age.4
A major concern about teenage pregnancy is the
adequacy of the young mother to care for her child. Over the last two decades, considerable attention has focused on the question of whether children of
teen-age mothers, compared with children of older
moth-ers, are more likely to be abused or neglected.
Al-though there is a solid theoretical foundation for
believing that teenagers have more difficulty being
adequate parents,5’6 as well as some empirical
evi-dence to suggest that such young mothers have more
deficits inparenting skills and interactions with their
infants,711 the data about the association between
young maternal age and child maltreatment are
con-ificting. Of 23 studies that assessed the relationship
between maternal age and child maltreatment, 14
found that young maternal age was a risk factor,1223
and 9 found that young maternal age was not
asso-ciated with an increased risk of child
maltreat-ment.26-M No study found that young maternal age
was associated with a decreased risk of
maltreat-ment.
There are five major methodologic issues that may
have contributed to these conflicting results: (1)
These studies have all used either case-control or cross-sectional designs; of the designs used to test
causal inference, these two are most susceptible to
bias.38’- None of the previous studies has used a
longitudinal design. (2) In many of the studies, the comparison group was not appropriate, or there was
no comparison group. (3) The definition of young
maternal age has varied among studies. Age of the
mother has been defined either at the time of
mal-treatment, at the birth of the index child, or at the birth of the mother’s first child. Also, the
demarca-tion for “young” has been variously defined as 20
years, 18 years, or some other age. (4) There is
marked variability in the definition of maltreatment and in the inclusion of specific types of maltreatment such as physical abuse, neglect, sexual abuse, failure-to-thrive, and “high-risk” children. (5) Previous stud-ies have failed to address the problem of detection bias, which can occur when the likeithood of detect-ing maltreatment is inadvertently enhanced for one
group. Thus, the risk of maltreatment may be
Therefore, to provide a new approach to address the question of whether children of young mothers
are at increased risk of child maltreatment, we
con-ducted a longitudinal, cohort study in which compa-rable control subjects were obtained from birth logs
and careful attention was paid to the problem of
detection bias by using predefined criteria to define
child maltreatment and a rater who was “blind” to
the subjects’ group status.
METHODS
We used a retrospective, longitudinal, cohort study. Children
born to young mothers (18 years of age or younger at the time of
the child’s birth) comprised the index group. These were matched with sociodemographically comparable children of mothers who were older than 18 years of age. Both groups were then followed longitudinally to their fifth birthday, and three kinds of adverse outcomes were ascertained: child maltreatment, poor growth, and changes in the child’s primary caretaker.
Selection of Subjects
All subjects met three eligibility criteria: (1) born at Yale-New Haven Hospital (YNHH) from October 1, 1979, through December
31, 1981; (2) a singleton birth, because twins may have an
in-creased risk of subsequent maltreatment-7; and (3) at least two
visits to the YNHH’s primary care center before 6 months of age and at least one visit to any clinical service at YNHH after 10 months of age. This last criterion selected for subjects who lived in the inner city of New Haven and used the hospital for their pri-mary care. It also ensured a minimum of approximately I year of follow-up for all children.
To assemble the subjects, we reviewed a computerized log of all births at YNHH beginning with October 1, 1979. Based on the independent variable of the mother’s age at the time of her deliv-ery, we selected two groups to be followed. For the index group, we identified infants born to mothers who were 18 years of age or younger. We chose this age because for most adolescents it coin-cides with the final year of high school. Children in the
compar-ison group were born to mothers aged 19 to 34 years. To obtain a
demographically comparable comparison group, index children
were matched with comparison children according to date of birth (within 9 months), ethnicity, gender, birth order, and method of payment for the hospitalization (Medicaid, private insurance, or self-payment) as a marker for socioeconomic status. For expedi-ency pairing was not done on a case-by-case basis, but rather by “category matching.” This was accomplished by identifying the number of index children born over a 6-month period with a specific set of characteristics (eg, black, female, first-born,
receiv-ing Medicaid). We then selected an equal number of comparison
children with the same identifying characteristics. For a small number of subjects, we could not find comparison subjects who matched on all four variables. In these instances, either gender or ethnicity was dropped as a matching variable, and the search continued until an appropriate match was made.
Identification of Outcomes
The adverse outcomes that were assessed for each child are similar to those of a previous study’9: (1) maltreatment, induding physical abuse, neglect, or sexual abuse; (2) growth failure defined by specific growth criteria, or hospitalization for nonorganic fail-ure-to-thrive; and (3) changes in the child’s primary caretaker.
Maltreatment. Our approach to the assessment of maltreatment
was designed to distinguish unintentional injuries from abuse or neglect and to minimize detection bias, which may occur if the assessment of maltreatment is made with knowledge of the moth-er’s age. To accomplish these goals, we used a “blinded” review of events by one of the investigators.
One investigator (either D.MS., L.J., or J.M.) abstracted the
medical records at YNHH and at three major local health care
facilities (the only other hospital in New Haven and two
neigh-borhood health centers). For each subject, we collected data from
birth to the child’s fifth birthday. The investigator recorded details of all medical visits where there were injuries or suspected injuries
to the child.
The second investigator (J.M.L.) then reviewed a summary of each injury episode, but was “blind” to all other elements of the child’s history including details such as the mother’s age, the child’s race, and who accompanied the child on the visit.
In the final step, an “unblinded” investigator (D.M.S.) and the “blinded” investigator (J.M.L.) independently dassified each epi-sode of injury by using predefined criteria for physical abuse,
sexual abuse, neglect, and unintentional injuries. Overall
agree-ment between the raters was 92%. When disagreements occurred, the case was jointly reviewed by these two investigators until a consensus could be reached.
Specific predefined criteria were revised from a previous stud?9 and used to define the following: (1) physical abuse (def-mite, probable, or questionable); (2) sexual abuse (definite, prob-able, or questionable); (3) physical neglect; (4) supervisional ne-glect; (5) medical neglect; (6) unintentional injury-neglect; (7)
unintentional injury; and (8) household violence. When insuffi-cient information was available in the medical record to classify an episode, no rating was provided by the reviewers. Unintentional injury-neglect indudes incidents that were considered accidental but were likely to have been preventable by reasonable parental
supervision. Examples induded falls from the bed by children younger than 9 months of age or the first episode of aburn from a household object, such as heater or iron. For purposes of anal-yses, maltreatment was defined as physical abuse (definite or probable), sexual abuse (definite or probable), or neglect (physical neglect, supervisional neglect, or medical neglect). Episodes clas-sified as “questionable” physical abuse or “questionable” sexual abuse were included in the unintentional injury category, so that our final definitions of abuse and sexual abuse were conservative
ones.
Poor Growth. The second outcome was poor growth or
hospi-talization for nonorganic failure-to-thrive (lTD. To examine the child’s growth, the subject’s height and weight at each visit were plotted (corrected for gestational age in children born before 37 weeks of gestation) on the appropriate growth curves from the National Center for Health Statistics.#{176} Each child’s growth curve was then reviewed for evidence of adecreased velocity of growth or an abnormality of attained growth. The velocity of growth was considered abnormal if a child’s weight decreased and crossed two major weight percentiles.4’ Attained growth was considered abnormal if the child’s weight dropped below the fifth percentile and if the weight was less than 90% of the ideal weight for age and height.42 addition, hospitalizations for FTF were reviewed and, based on the dinidans’ notes, classified as due either to organic or nonorganic causes or to both.
Change in Child’s Primary Caretaker. The third outcome was a
change in the child’s primary caretaker due to “parenting failure”; such conditions included a mother’s leaving her child to the care of arelative or maltreatment of the sibling of one of the subjects. Changes were classified as those in which the child was placed by the state’s child protective service agency in foster care or those in which less formalized changes in the guardian occurred, often because the mother left the child for a period of time in the care of a relative. Also, the reason for a change in the caretaker and the duration were recorded.
Ascertainment of Other Data
Information about the following perinatal variables was
ex-traded from each chart the mother’s age at birth, marital status,
history of substance abuse, history of psychiatric illness, and type of delivery; and the child’s gestations! age, birth weight, Apgar
scores, and neonatal complications. In addition, we recorded the
date of the child’s last visit to either YNHH or one of the three other major health care facilities so that we could determine the duration of follow-up data on each subject. We also recorded the yearly number of visits for each subject and information about special services (such as involvement with the pediatric social
worker or attendance at a special school for teenage mothers).
Statistical Analysis
To compare the frequency of the outcomes in the children born to young mothers with those in the comparison group, we
calcu-lated unmatched risk ratios (RRs) and their associated 95%
controlling for statistically significant variables that may have been potential confounders.” In these analyses, the odds ratio was
used as an approximation of the RR.’
RESULTS
During the 3-year period, we identified 219
chil-dren born to young mothers (aged 18 years or
young-er) and 219 matched comparison children born to
older women. Of the young mothers, at the time of
delivery, 5% were 13 or 14 years old, 33% were 15 or
16 years old, and 62% were 17 or 18 years old. Of the comparison mothers, 52% were 19 or 20, 22% were 21 to 23, and 26% were 23 to 34. Characteristics of the sample are shown in Table 1. As expected, 52.5% of the subjects were boys, and the majority were
first-born children of black or Hispanic women for whom
medical care was paid by Medicaid. The index and
comparison groups were similar on six of the
non-matching characteristics shown in the bottom of
Table I. Not surprisingly, the young mothers less
freq uently were married, more often had been seen by a hospital social worker on the postpartum ward,
and more often had attended the school for teenage
mothers in the New Haven community.
Children in the index and comparison groups
re-ceived medical care at YNHH or at the three other
health care facilities for similar durations and at
sim-ilar frequencies during the period of the study. The
median age of the children at the last recorded visit
in the medical record was 62 months for the index
group and 59 months for the comparison group.
When the children were 30 months of age,
informa-tion was available on 92% of the index group and
86% of the comparison group; comparable figures at
60 months were 62% and 50%, respectively. The
fre-quency of visits to a health care facility was similar for both groups. For instance, during the first year of life, 44% of both the index and comparison children had made more than five visits to a health care facil-ity; during the third year, 11% of the index children
had made more than three visits compared with 9%
of comparison children.
Table 2 shows the number of episodes of
maltreat-ment, household violence, unintentional injuries,
and unintentional injury-neglect; the percentage of subjects in each group; and the RRs and their
asso-ciated 95% CIs. Maltreatment occurred two times
more often in index children than in those in the
comparison group (RR = 2.00; 95% CI = 1.17, 3.64).
Although all three types of maltreatment occurred
more commonly in the index children, this difference was statistically significant only for neglect (RR = 2.83; 95% CI = I .14, 7.05). In contrast, the outcomes classified as unintentional injuries (RR = 1.11; 95% CI
=
0.94, 1.31) or unintentional injury-neglect (RR =1.13; 95% CI = 0.65, 1.52) occurred in approximately
the same proportion of children in the two groups.
As expected, the majority of the cases of abuse or neglect occurred within the first 2 years of life. This
was especially the case in the index group, where 26
(87%)
of 30 episodes of maltreatment occurred before 2 years of age compared with 6 (50%) of 12 episodesin the comparison group (P = .02 by Fisher’s Exact
Test). In contrast, 7 of the 9 episodes of sexual abuse occurred after 2 years of age.
When each child’s growth was evaluated, 6.9% of
index children and 4.1% of comparison children met
the criteria for poor growth (RR = 1.67; 95% CI =
0.75, 3.73). The majority of cases of poor growth oc-curred during the first year of life (82% in the index
group and 90% in the comparison groups). Of the
five episodes of hospitalization for nonorgamc FIT,
four were in children in the index group. All of these hospitalized children also were identified by growth criteria.
To determine whether there were differences in the
occurrence of maltreatment or poor growth within
the two broad categories of maternal age (18 years
of age or >18), we examined the rates of either of
TABLE 1. Perinatal Characteristics of Index and Co mparison Subjects
%of %of P
Index Comparison
Children Children
(n = 219) (n = 219)
Matched variables
Male gender 52 53 NS
Birth order: first-born 80 81 NS
Race
Black 73 75
Hispanic 19 16 NS
White 8 9
Method of payment: Medicaid 90 91 NS
Nonmatched variables
Married 9 15 <.05
Cesareansection 13 14 NS
Seen by hospital social worker 55 21 <.001
Attended school for teenage mothers 44 4 <.001
Mother with history of psychiatric illness and/or 8 8 NS
substance abuse
Gestational age < 37 wk 5 6 NS
Birth weight < 2500 g 12 9 NS
5-Minute Apgar score < 8 6 4 NS
Neonatal complication 9 12 NS
TABLE 4. Changes in Caretaker in Index and Comparison Children
*CPS, Child Protective Services. TABLE 2. Assessments of Injuries in Index and Comparison Children
Outcome No. of Events % of Subjects Risk
Ratio
95% Confidence
Interval
Index Comparison Index Comparison
Maltreatment 37 14 12.8 6.4 200 1.17,3.64
(Abuse) (11) (6) (4.1) (2.7) (1.50) (0.54, 4.14)
(Sexual abuse) (7) (2) (2.7) (0.9) (3.00) (0.61, 14.70)
(Neglect) (19) (6) (7.8) (2.7) (2.83) (1.14, 705)
Household violence I I 0.5 0.5 1.00 ...
Unintentional injury 239 203 58.9 53.0 1.11 0.94, 1.31
Unintentional injury-neglect 86 80 31.1 27.4 1.13 0.65, 1.52
Insufficient information 10 ii 4.1 5.0 0.91 ...
Total 373 309
these outcomes at 2-year intervals beginning with a
maternal age of 13. As seen in Table 3, the rates of
adverse outcomes varied from 17% to 21% in the
three age clusters in the index group and from 7% to 14% in the age clusters in the comparison group. Similar results occurred when the rates of the out-come were examined for each yearly interval of
ma-ternal age or when the outcome was maltreatment
alone. Thus, within either the index or comparison
groups, there is no grouping by maternal age that
results in rates substantially different from the over-all rates.
We also attempted to determine from the medical
records who were the suspected perpetrators in cases of physical abuse or sexual abuse. Of the nine
phys-ically abused children in the index group, two were
abused by the father, one by the mother’s boyfriend,
one by an uncle, one by a maternal great aunt, two by
either the mother or her boyfriend, and two by a
perpetrator who could not be clearly identified from the record. In contrast, of the six physically abused
children in the comparison group, five were abused
by the father, and in one case it was not clear who
abused the child. Thus of the 15 cases of physical abuse, in 10 cases, the mother was definitely not the abuser, while in two she was considered as the pos-sible abuser. In the eight cases of sexual abuse, the
perpetrator was a neighbor or relative in six cases
and unknown in the other two. None of the
identi-fled perpetrators of sexual abuse was the natural
mother or father.
The third outcome, changes in the child’s primary
caretaker, occurred four times more frequently
among children in the index group than among those
in the comparison group (Table 4). Of the 28 index
children who had a change in caretaker, 14 were
TABLE 3. Occurren
Mother’s Age at Birth
ce of Maltrea tment or Poor Growth by the
Age Group (y)
Group Size
No. (%) of Subjects With Outcome
13-14 15-16 17-18
12 72 135
2(17) 15(21) 26(19)
Total 219 43 (20)
19-20 21-22 23+
113 49 57
11 (10) 7(14)
4(7)
Total 219 22(10)
% of Subjects
Index Comparison
Risk Ratio
95% Confidence
Interval
Placement by CPS* 6.4 1.4 4.67 1.38, 15.8
Change in guardian 6.4 1.8 3.50 1.16, 105
Either of above 12.8 3.2 4.00 1.80, 8.87
placed by the child protective service system and 14
were left in a less formalized arrangement in the care
of a primary guardian other than their mother;
com-parable figures for the 7 comparison children were 4
and 3.
The majority of the changes in the caretakers were
not linked to visits that children made to the health care facilities, such as a visits for injuries due to phys-ical abuse or hospitalizations for FTT. In the index group, 9 (32%) were linked, but 19 (68%) were not. Of
these 19, there were a variety of reasons why the
child was no longer living with the mother induding abuse of a sibling (1 case), history of multiple epi-sodes of abuse (1), a chaotic family that was unable to
provide a safe home for the child (2), the mother
moved out and left the child with a relative or sent
the child away (7), the mother had a serious psychi-atric disorder and could not care for the child (2), the
grandmother became seriously ifi and could not help
care for the child (1), and reasons unclear (5). In the comparison group, three cases were linked to visits
for maltreatment or FiT, while four cases were not
(two mothers left the child with a relative, and in two cases the reasons were unclear).
When the three outcomes are combined
(maltreat-ment, poor growth, or a change in the primary
care-taker), 26.9% of the index children had at least one of
these adverse events compared with 11.4% of the
children in the comparison group (RR = 2.36; 95% CI
=
1.54, 3.62).Finally, logistic regression was used to examine the outcomes after controlling for potentially
confound-ing perinatal variables noted in Table 2. Two
peri-natal variables were associated independently with
maltreatment: a maternal history of psychiatric
ifi-ness and/or substance abuse (odds ratio = 3.21; 95%
CI = 1.36, 7.60) and prematurity of the infant (odds ratio = 3.67; 95% CI = 1.37, 9.82). After adjusting for
these two variables, the association between young
statisti-cally significant (odds ratio = 2.13; 95% CI = 1.06, 4.26) and did not differ substantially from the unad-justed RR of 2.00 (or, in fact, from the unadjusted odds ratio of 2.15). Similarly, when the other major outcomes of the study were examined, the adjusted odds ratios were not substantially different from the unadjusted RRs presented above.
DISCUSSION
In this cohort study of children living in an inner city, we found that by the child’s fifth birthday, the
adverse outcomes of maltreatment, poor growth, or a
change in the child’s primary caretaker occurred in
27% of children born to mothers who were aged 18 or
younger at the time of birth. This rate was 2.4 times higher than for a demographically similar compari-son group born to older mothers. Moreover, physical abuse, neglect, or sexual abuse occurred in 13% of the
children of young mothers in our sample, twice the
rate of maltreatment in the comparison group.
Al-though each type of maltreatment occurred more
commonly in children of young mothers, only
ne-glect, by itself, was statistically significantly different with an RR of 2.83. For physical abuse, the RR was 1.50; since this outcome occurred infrequently, a
sam-ple of approximately 2600 subjects per group would
have been necessary for this RR to be statistically significant (a = .05 and 3 = .20). Poor growth also
occurred more commonly in children of young
moth-ers, but again this difference was not statistically
sig-nificant.
Changes in the child’s caretaker did occur statisti-cally significantly more often in children born to
young mothers (RR of 4.0). Some of these changes
were in part due to the maltreatment or serious FTT
that occurred among the children; many of the
sep-arations, however, were not directly related to
mal-treatment, but rather were a consequence of the
dis-organized social environments that seemed to occur
more commonly in the families of the young
moth-ers.
Our categorization of “young age” of the mother
differs from that used in other studies in the
litera-ture, which have typically chosen an age of 17 or 19
as the upper boundary of adolescence. We chose age
18 because it is the usual age of graduation from high
school and therefore reflects a “natural” cutoff
be-tween adolescence and young adulthood. In fact, an
analysis based on the exact age of the mother
sup-ports this demarcation, since children of 17- or
18-year-olds had a rate of maltreatment or poor growth twice that of children of 19- or 20-year-old mothers.
We were surprised that the rates of these two
out-comes were not even higher among children of the
youngest mothers (ie, those aged 13 to 16 years).
However, grandmothers provide a substantial
amount of care to children born to very young moth-ers, and we speculate that without this care, the
oc-currence of adverse events among such children
might have been even higher.
Although this study focused on the extremes of
parenting dysfunction in families with young
moth-ers, it rarely was the mother who physically abused her child. Of the 15 cases of physical abuse that
oc-curred, 10 of the abusers were identified in the med-ical records, and none of these was the mother; there were seven fathers, two relatives, and one boyfriend of a mother. Of the 5 cases in which it was not clear
who hurt the child, the mother was considered the
possible abuser in 2 cases. Therefore, although we are
examining young maternal age as a risk factor, the
increased risk of maltreatment is not solely because of the mother’s behavior, but rather frequently be-cause of the behaviors of others in the family or in the home.
Our results are consistent with the previous case-control and cross-sectional studies that found an
as-sociation between young maternal age and child
maltreatment.12 When such associations have been
found, the strength of the association has been in the range of a twofold difference. Our own previous case-control study of physically abused children over
a 6-year period found an odds ratio of I .87 for the
association between abuse and mothers who were
younger than 20 years of age at the time of the child’s birth and 1.89 for mothers less than 18 years of age21; these associations are similar to the present result of 1.50 for physical abuse.
Our study is unique in that we used a cohort
do-sign as well as specific criteria to define whether
maltreatment had occurred or whether the child had
a period of poor growth. We also examined the
child’s separation from his or her primary caretaker;
although this outcome has not been examined in
pre-vious studies of parenting by teenage mothers, we believe that it is important because of the potential psychosocial consequences to the child and because of the burden that it places on the foster care system. In designing this study, we attempted to minimize three potential biases that may have falsely elevated the rates of the outcomes in the index children. First, selection bias in the choice of control subjects was
minimized by selecting demographically similar
con-trol subjects from a log of births and by using logistic regression to adjust the odds ratios for potential con-founders.
Second, the possibility of detection bias was
as-sessed by comparing both the frequency of visits to health care providers and the duration of follow-up
in the two groups. Detection bias would occur if
children of young mothers were seen more
fre-quently or had a longer period of follow-up than the
control children, and these differences resulted in a
falsely elevated rate of maltreatment in the index group. Since the number of visits to health care
pro-viders was similar in the two groups and since the
median ages of the last recorded visits also were
similar, it is unlikely that detection bias could ac-count for the differences found in our study.
Third, a type of detection bias-namely, reporting
bias-would occur if children of teenagers are more
likely, because of the mother’s age, to be reported for
maltreatment. This might occur if the same type of
injury was considered as intentional in the index
chil-dren but unintentional in the comparison cMldren.
To minimize this type of bias, we used a reviewer
who was “blind” to the child’s group status. We also
clini-cians’ assessments to determine the occurrence of poor growth.
Despite our efforts to minimize these important
biases, relying on medical records to assess these
outcomes has resulted in certain limitations. First, we may have underestimated the occurrences of the out-comes. It is likely that all episodes of maltreatment did not result in a visit to a health care site or that all changes in caretakers were not noted in the medical records. On the other hand, be reviewing each injury that did result in a visit, it is unlikely that we missed the serious injuries due to maltreatment.
Second, the problem of reporting bias is not com-pletely eliminated by a “blind” review of injuries. The potential for bias still exists if clinicians obtained
a more detailed history about what happened to an
index child compared with the history obtained
about a similar injury to a child of an older mother, and this more detailed history resulted in a classifi-cation of the injury as maltreatment.
A third limitation is in the ascertainment of
changes in the child’s primary caretaker. Our RR of
4.0 may be falsely elevated for two reasons: (1) sim-ilar levels of parental dysfunction may more readily result in the out-of-home placement of a child living
with a teenage mother compared to an older mother
and (2) changes in caretakers may be recorded more
readily for index children. We collected no data to
help determine the magnitude of this bias.
Finally, this study focused on children living in the
early and mid-1980s in urban environments. Our
re-sults, therefore, should be interpreted cautiously in reference to children of adolescent mothers who live in suburban or rural environments.
In conclusion, we found that maltreatment, in
par-ticular neglect, occurred twice as frequently and
changes in caretaker four times as frequently to chil-dren of younger mothers compared with children of older mothers. These findings highlight the need to
provide adequate services to young mothers and
their children. Although access to good prenatal and pediatric care and appropriate educational services is important, for many young parents and their chil-dren, these services may not be sufficient to ensure good outcomes. In a recent evaluation of services for
teenage parents, Daro found that supportive
ser-vices, such as parenting classes, often were not ade-quate to change dramatically the teenager’s parent-ing abilities. In many cases, more intensive services,
such as those provided by home visits, may be
nec-essary. Recent research findings indicate that a home
visiting program can be helpful to some young
moth-ers in preventing maltreatment of their children.47 Based on our results, at least part of these intensive
services should focus on the prevention of neglect
and on the behaviors of others in the home in an
effort to prevent physical and sexual abuse of the
child.
ACKNOWLEDGMENTS
This work was supported by grant 90CA1374 from the National Center on Child Abuse and Neglect, Administration for Children, Youth, and Families and a student research grant from the Amer-ican Heart Association.
We thank Julia Robertson for secretarial assistance and Eugene Shapiro, MD, for his helpful advice.
REFERENCES
1. Trussell J. Teenage pregnancy in the United States. Fam Plann Perspect.
1988;20:262-272
2. McAnarney ER, Hendee WR. Adolescent pregnancy and its conse-quences. JAMA. 1989a62:74-82
3. National Center for Health Statistics. Vital Statistics of the United States,
1988, Vol 1, Natality. Public Health Service. Washington, DC: US
Gov-ernment Printing Office; 1990. Table 1-58
4. New Haven Health Department, Birth Statistics, 1988
5. Sadler IS, Catrone C. The adolescent parent: a dual developmental crisis. JAdolesc Health Care. 1983;4:100-105
6. Bbs P. The second individuation. Psychoanal Study Child. 196722:162-182
7. Philliber SC, Graham EH. The impact of age of mother on mother-child
interaction patterns. IMarriage Fam. 1981;43:109-115
8. Parks PL, Smeriglio VL. Parenting knowledge among adolescent
moth-era. JAdolesc Health Care. 1983;4:163-167
9. Garcia Coil CT, Hoffman J, Oh W. The social ecology and early parent-ing of caucasian adolescent mothers. Child Dev. 198758:955-963
10. Levine L, Garcia Coil CT, Oh W. Determinants of mother-infant
inter-action in adolescent mothers. Pediatrics. 1985;75:23-29
11. McAnarney ER, Lawrence RA, Ricciuti HN, Polley J, Szilagy M.
Inter-actions of adolescent mothers and their 1-year-old children. Pediatrics. 1986;78:585-590
12. Smith SM, Hanson R, Noble S. Parents of battered babies: a controlled
study. Br Med J.1973;4:388-391
13. Lauer B, Broecke ET, Grossman M. Battered child syndrome: review of 130 patients with controls. Pediatrics. 197454:67-70
14. Lynch MA. ill-health and child abuse. Lancet. 19752:317-319
15. Lynch MA, RobertsJ. Predicting child abuse: signs of banding failure in the maternity hospital. Br Med 1.1977;1.:624-626
16. Sills JA,Thomas U, Rosenbloom L Non-accidental injury: a two-year study in central Liverpool. Dev Med Child Neurol. 1977;19:26-33 17. McCarthy J, Radish ES. Education and childbearing among teenagers.
Fam Plann Perspect. 1982;18:12-18
18. Egeland B, Brunnquell D. An at-risk approach to the study of child
abuse: some preliminary findings. J Am Aced Child Psychiatry. 1979;18:219-235
19. Herrenkohl EC, Herrenkohi RC. A comparison of abused children and
their nonabused siblings. I Am Acad Child Psychiatry. 1979;18:260-269
20. JasonJ, Andereck ND. Fatal child abuse in Georgia: the epidemiology of
severe physical child abuse. Child Abuse Negl. 1983;7:1-9
21. LeventhalJM, Egerter SA, Murphy JM. Reassessment of the relationship of perinatal risk factors and child abuse. AJDC. 1984;138:1034-1039 22. Miller SH. The relationship between adolescent childbearing and child
maltreatment. Child Welfare. 1984;63:553-557
23. Oates RK. Child abuse and non-organic failure to thrive: similarities and differences in the parents. Aust Paediatr J.198-t20:177-180
24. Benedict MI, White RB, Comely DA. Maternal perinatal risk factors and
child abuse. Child Abuse NegI. 19859:217-224
25. Creighton SJ. An epidemiological study of abused children and their families in the United Kingdom between 1977 and 1982. Child Abuse Negl. 19859:441-448
26. Simons B, Downs EF, Hurster MM, Archer M. Child abuse:
epidemio-logic study of medicaily reported cases. NY State JMed. 1966;66:2783-2788
27. Elmer E. Children injeopardy: A Study ofAbused Minors and Their Families. Pittsburgh, PA:University of Pittsburgh Press; 1967
28. Gil DG. ViolenceAgainst Children: PhysicalChildAbuse in the United States. Cambridge, MA: Harvard University Press; 1970
29. Goldson E, Cadol RV, Fitch MJ, Umlauf HJ. Nonaccidental trauma and
failure to thrive: a sociomedical profile in Denver. AJDC.
1976;130:490-492
30. Kinard EM, Klerman LV. Teenage parenting and child abuse: are they
related? Am IOrthopsychiatry. 198050:481-488
31. Cater JI, Easton PM.Separation and other stresses in child abuse. Lancet.
1980;1:972-974
32. Earp JA, Ory MG. The influence of early parenting on child
maltreat-ment. Child Abuse NegI. 1980;4:237-245
33. Altemeier WA, O’Connor 5,Vietze PM, Sandier HM, Sherrod KB.
An-tecedents of child abuse. JPediatr. 1982;100:823-829
34. Dubowitz H, Hampton RL, Bithoney WG, Newberger EH. Inflicted and noninflected injuries: differences in child and familial characteristics.
Am JOrthopsychiatry. 198757:525-535
Philadelphia, PA: WB Saunders Co; 1985
36. LeventhalJM. Research strategies and methodologic standards in stud-ies of risk factors for child abuse. Child Abuse Negi. 1982;6:113-123 37. Groothius JR. Altemeier WA, Robage JE et al. Increased child abuse in
families with twins. Pediatrics. 1982;70:769-773
38. Kleinbaum DG, Kupper LL, Morgenstern H. Epidemiologic Research:
Principles and Quantitative Methods. Belmont, CA: Lifetime Learning
Publications; 1982:337-402
39. Leventhal JM, Garber RB, Brady CA. Identification during the
post-partum period of infants who are at high risk of child maltreatment. J Pediatr. 1989;114:481-487
40. National Center for Health Statistics. NCHS Growth Charts, 1976.
Monthly Vital Statistics Report 25 Suppl. Rockville, MD US Health
Re-sources Administration; 1976:76-1120
41. Frank DA, Zeisel SH. Failure to thrive. Pediatr Clin North Am. 1988;35:1187-1206
42. Kempe RS, Goldbloom RB. Malnutrition and growth retardation (‘fail-ure to thrive’) in the context of child abuse and neglect. In: Heifer RE,
Kempe RS, eds. The Battered Child. 4th ed. Chicago, IL: University of Chicago Press; 1987:312-315
43. Fleiss fl. Statistical Methods for Rates and Proportions. 2nd ed. New York,
NY: Wiley & Sons; 1981:112-119
44. SAS institute Inc. SAS User’s Guide: Statistics 1989 ed. Carey, NC: SAS
Institute mc; 1989
45. Schlesselman JJ.Case-Control Studies: Design, Conduct, Analysis. New York, NY: Oxford University Press; 1982:33-34
46. Center on Child Abuse Prevention Research. Adolescent Parent Services Evaluation: Final Report. Chicago, IL: National Committee for Prevention
of Child Abuse; 1990
47. Olds DL, Henderson CR, Chamberlin R,Tatelbaum R. Preventing child
abuse and neglect: a randomized trial of nurse home visitation. Pediat-rics. 1986;78:65-78
UNPROVEN STANDARD THERAPY
Proof of the effectiveness of “standard therapy” is often lacking in clinical med-icine. Some have speculated that “some sensible patients. . .might refuse standard therapy if they knew how little data there were to justify it.”1 Indeed, the provision of “standard” medical care (with the patient’s best interest in mind) and the
conduct of scientific research on promising (but not proved) experimental
thera-pies, is not as disparate as Dr. Levine suggests. “There need be no dichotomy
between clinical and scientific medicine if physicians are willing to admit their
ignorance with regard to therapeutic measures and to test their hypotheses prop-erly. In the long run, many more lives will be saved. . #{149}“2 especially true for resuscitation medicine.
Abramson NS, Safar P. Response to commentary. Controlled Clin Trials. 1991;12:551-552.
REFERENCES
Noted by J.F.L., MD
1. Chalmers TC. Ethical aspects of clinical trials. Am JOphthalmol. 1975;79:753.