Pediatric Social Illness: Toward an Etiologic Classification

Pediatrics, 60:178-185, 1977.

Pediatric Social Illness: Toward an Etiologic Classification

Eli H. Newberger, M.D., Robert B. Reed, Ph.D., Jessica H. Daniel, Ph.D., James N. Hyde, Jr., M.S., and Milton Kotelchuck, Ph.D.

From the Department of Medicine, Children’s Hospital Medical Center, and the Department of Pediatrics, Harvard Medical School, Boston

ABSTRACT. The significance of ecological stress factors for understanding the etiology and interrelationships among the pediatric social illnesses was explored in a case-control study of 560 children under 4 years of age. Cases of child abuse and neglect, failure to thrive, accidents, and poisonings were matched on age, socioeconomic status, and ethnic group with children who had comparably acute medical conditions. Data were ascertained from the children’s medical records and from an extensive maternal interview which probed historical and contemporary familial, environmental, and child developmental realities.

The findings support the basic hypothesis that the occurrence of pediatric social illness is associated with increased family stress. Child abuse is associated with more extreme stresses in all categories studied; failure to thrive with maternal historical stresses, perceived sickness of the index child, and contemporary social isolation; and accidents with contemporary household crises. An additive mode of pathogenesis of the more severe symptom manifestations is suggested by these data.

Specific at-risk items were also noted. Although child abuse separated sharply from the other entities in a discriminant function regression analysis of the data, the insufficient predictive power of the principal discrimination features suggests that proposed programs to screen for risk of child abuse are of questionable accuracy and social utility. Pediatrics 60:178-185, 1977, SOCIAL ILLNESSES, STRESS, CHILD ABUSE, FAILURE TO THRIVE, ACCIDENTS, POISONING, EPIDEMIOLOGY.

The “social illnesses” of pediatrics include child abuse and neglect, failure to thrive, accidents, and poisonings. They account for a major share of the mortality of preschool children and often have significant physical and psychological sequelae.^1-4 They are classified partly according to their manifested symptoms and partly on supposed causal factors. But the logic underlying this taxonomy, as can be seen in Table I, provides the clinician with a conceptual framework inadequate to organize the complex data dealt with in practice. These simple formulations can misdirect the approach to the individual patient, and they contribute to the developmental impact of these illnesses on children, for whom clinical practice is of inconsistent organization and quality.⁵

There is, moreover, little reliable observational information to support the notions of cause and effect built into these diagnoses. For example, a child with scattered bruises on his body might be identified either as an accident victim or as a victim of child abuse.⁶ In the latter case, there is a presumption, but rather rarely in practice knowledge, of parental fault.⁷ Intervention, when it is made available, is often individual-directed psychiatric counseling of the parents while deliberations proceed on whether or not to place the child in foster home care. The criterion of successful management is protection from his parents, the proximal cause of the child’s disease, with little regard to the social, familial, environmental, or child developmental determinants of the child’s injury. Help in reducing urgent stress on the family is not acknowledged as a treatment vehicle when the diagnostic focus is toward defining the responsibility of the perpetrator for the injuries of the victim.

By contrast, if the child is classified as an accident victim, there may be no implications of familial cause and no treatment. If the presenting lesion is seen as resulting from an act of God, there is hardly any need for diagnosis or therapy from a social worker or psychiatrist. This process of selective classification, based on slim logical and empirical supports, becomes a matter with serious ramifications for clinical practice and social policy given the findings of previous work on the preferential susceptibility of poor and minority children to receive the diagnoses child abuse and neglect, while children of middle and upper class homes may be more often identified as victims of accidents.^8.9

Several small clinical studies have suggested common relationships among the various categories of pediatric social illnesses (for example, prior accidents in child abuse cases). ^10-14  This report explores underlying common origins among these conditions, with a view to defining a more etiologic (as opposed to manifestational) illness taxonomy.

Stress Theory of Common Etiology

It was posited that this common set of circumstances included elements of historical and contemporaneous stress. Historical stresses were defined as stresses occurring in the life of the maternal care-giver up to the time of the conception of the index child. Contemporaneous stresses refer to environmental, social, familial, and health problems occurring since the conception of the child as well as to stress imposed by unique attributes of the child.

SUBJECTS AND METHODS

All children under 4 years of age seen in either inpatient or outpatient departments of the Children’s Hospital Medical Center in Boston for pediatric social illness were eligible for selection into the study as “cases.” Children not bearing pediatric social illness diagnoses were eligible for selection into the control group; children suffering from chronic or terminal illnesses, however, were excluded from the control population. The sample was ascertained between December 1972 and May 1974.

Cases were matched to controls on the basis of age, race, and the most readily available index of socioeconomic status at the time of the family’s first contact with the hospital (whether or not the welfare department paid the medical bill).

Because interviews in the emergency room could not be performed after the visit with the physician, cases and controls in that area were sampled on the basis of their presenting symptom (injury or ingestion), not on the basis of a medical diagnostic formulation.

TABLE I CONCEPTUAL MODELS IMPLICIT IN PEDIATRIC SOCIAL DIAGNOSES
Diagnosis	Conceptual Model
Child abuse and neglect	Intentionally motivated parent or caretaker assaults a defenseless child or withholds care from him
Accidents	Isolated, random traumatic events
Failure to thrive	Idiopathic failure of a baby to gain weight

Five hundred sixty children and families were studied, including 303 inpatients and 257 outpatients. Table II summarizes the number of maternal interviews performed for each diagnostic group.

To assure comparability with previous research, child abuse was defined in terms of inflicted injury and a clinical impression of great risk by professionals on the hospital’s Trauma X team experienced with such “protective” problems. Child neglect is a rare clinical diagnosis at Children’s Hospital; the single case in the present study is included for analytic purposes with the cases of child abuse.

Interview

The principal instrument for the study was a structured interview of the subject’s mother, conducted at the hospital. The interview focused on realities of child development, family relationships, health, finances, employment, and housing, as well as on specific life experiences of the mother and her child. Interviews lasted about 45 minutes and were conducted by specially trained interviewers. Although it was not possible to blind the interviewers to the child’s clinical diagnosis, careful review of the interview process by a research supervisor and frequent meetings with the interviewers by a staff psychiatrist with no other tie to the project were performed continually to foster interobserver reliability and to minimize observer bias.

TABLE II NUMBER OF INTERVIEWS IN EACH PATIENT GROUP
		No.
Inpatient Cases Accidents Ingestions Failure to thrive Abuse Controls Total	73 34 42 16	165         138 303
Emergency room Cases Accidents Ingestions Controls Total	112 26	138     119 257

 

TABLE III CHARACTERISTICS OF CASE AND CONTROL GROUPS
Inpatient Case Control	53.9 62.3	66.7 73.2	38.2 31.2	57.6 58.0
Emergency room Case Control	33.3 48.7	53.6 45.4	57.2 54.6	55.1 53.8

Ethical Issues: Confidentiality, Informed Consent, and Advocacy

As information elicited in the course of the interview could serve as a basis for concern about risk to the child, the project developed formal guidelines for sharing access to the data with the hospital professional staff. This was not an easy matter to tackle, for the implications of sharing investigative data in research of this nature are great. On the one hand, we would have preferred to have absolute confidentiality as the operational imperative, because of the potentially deleterious effects of labeling a family as “at risk” for child abuse or neglect. On the other hand, we feared the consequences of not taking any action after obtaining information which suggested danger to the child.

The written consent form and method for obtaining consent and treating research data attempted to reconcile this ethical dilemma. We scrupulously adhered to the following, multistep procedure for obtaining consent and sharing interview data:

1. Prior to making contact with a mother to ask permission to interview, the physician responsible for the child’s hospital care was asked for his or her permission to interview the mother.

2. If a social worker was assigned to the case, he or she was also asked for permission.

3. After permissions 1 and 2 were obtained, contact was made with the mother.

4. After explaining the goals and nature of the study, but before beginning the interview, it was explained to the mother that the information elicited during the conversation was confidential, but that it was possible that information might be shared with the physician and/or social worker if it were felt that it might assist them in caring for the child.

5. In the instances where it was felt necessary to share the information with a professional person responsible for the management of a patient, the interviewer submitted a written abstract of the pertinent portions of the interview to the physician or social worker. The original interview was never released. Each of these abstracts was then stamped: “Not for insertion in the medical record.” In only ten of 560 interviews was information shared.

6. Interview schedules were kept in a locked file and referenced only through a coded system designed to prevent linking the names of respondents to the interview form without access to the code.

Because of the emphasis on environmental stress in the interview, we felt an ethical obligation to offer assistance to ameliorate the identified problems. To this end a family advocacy program was developed which was available to all participants. Designed initially to help families get such essential, and lacking, supports as adequate housing, child care, legal services, and adult health care, the program evolved into an organized service available to all hospital patients. Personnel with no formal professional training were taught and supervised to help families deal with contemporary life stresses and in gaining access to essential services.¹⁶

RESULTS

Demographic Characteristics

This study population reflects the differences in demographic composition of the hospital’s inpatient and emergency room services. Table III summarizes the demographic characteristics of the case and control groups. The inpatient study population comes from the greater Boston area and tends to be younger, predominantly white, and more middle class. The emergency room sample more nearly represents the predominantly black and low-income community directly around the hospital. There are slightly more male children in all groups.

TABLE IV CHARACTERISTICS OF SPECIFIC CASE GROUPS
Inpatient Accident Ingestion Failure to thrive Abuse	46.6 29.4 81.0 68.8	68.5 44.1 83.3 62.5	26.0 52.9 33.3 75.0	54.8 44.1 69.0 68.8
Emergency room Accident Ingestion	32.1 38.5	50.0 69.2	59.8 46.2	53.6 61.5

The matching of cases and controls on social class, race, and age is satisfactory.

As Table IV illustrates, however, there are marked demographic differences among the case categories. In the present sample, the patients suffering from failure to thrive and child abuse tend to be younger and male, those suffering from failure to thrive are more frequently white, and those suffering from child abuse are poor.

Medical and Family Data

Figure I summarizes the weight at admission for the children in the inpatient groups. Implicit in the definition of failure to thrive is the small size of the child.

It is of interest to note that children bearing the child abuse diagnosis in the study sample were also disproportionately small. Inpatient control subjects had acute medical conditions requiring hospitalization, accounting in part for their low weight. Children identified as having had “accidental” traumatic injuries tended to be significantly more robust, as indicated both by their weights and by their mothers’ reports of their health, than those in the other study categories.

The results of the maternal interviews were organized into a series of a priori scales developed to integrate and express data bearing on the central hypotheses of the study, the arithmetic means of which are expressed in Table V. (To develop these summative measures, an estimate of the discriminating power of each attribute was made, and a weighted score was devised. The study instruments and details of the analytic method are available on request from the senior author.)

FIG. 1. Proportion of children in inpatient groups
who were under tenth percentile for weight.

These scales are based on the sum of positive responses in a given category. Stress in the mother’s childhood included frequent family mobility, a broken home, and volunteered information about a personal history of violence and/or neglect. The scale “stress in the current household” was based on recent mobility and change in household composition. The scale “lack of social support” was designed to measure social isolation and included the absence of a telephone and a mother’s perception of her neighborhood as unfriendly.

As this table shows, stress was positively associated with all pediatric social illness categories. Accidents were characterized uniquely by a high level of contemporaneous stress. Cases of failure to thrive and of child abuse shared high levels of maternal historical stress and lack of social support. Subjects bearing the diagnosis of child abuse had higher scores in all three stress categories. Particularly of note is the very high level of current household stress in the child abuse cases, suggesting a greater role of ongoing crisis than is commonly acknowledged in the etiology and treatment of child abuse.

TABLE V A PRIORI STRESS SCALES: MEANS FOR INPATIENT GROUPS STANDARDIZED TO MEAN AND STANDARD DEVIATION OF CONTROLS
	Stress in Mother’s Childhood	Stress in Current Household	Lack of Social Support
Accident Ingestion Failure to thrive Abuse	.04 .46º .47º 1.15º	.59º .34 .27 1.58º	.19 .15 .52º .83º

ºP < .01 by one-tailed t test.  

TABLE VI SIGNIFICANT DESCRIPTORS (P < .05) FOR INPATIENT GROUPS IN ORDER OF PREDICTIVE IMPORTANCE
Accident	Ingestion	Failure to thrive	Abuse	Control
Good health of child	Child-rearing problems	Poor health of child	Recent moves No telephone	Regular health care
Low household density	Mother-child separations	Younger child	Mother-child separations	Few recent moves
Not welfare dependent	Older child	Male child	Serious childhood troubles for mother	Few child-rearing problems
Older child	Regular health care	Mother less education than father	Few children	No broken family in mother’s childhood
Baby-sitting help			Father older	Child-initiated separations (e.g., for health  reasons)
Recent moves	Female child	Neighborhood unfriendlyFamily physician	Low job status for father	Nobody to care for child when mother goes out

Classification Discriminants

Subsequent discriminant function regression analyses were conducted to determine which specific interview variables were predictive of a given category in the conventional taxonomy. (These discriminant functions were determined by defining the category of interest as “1” and defining all other categories, both cases and controls, as “0.” A step-wise regression was then calculated on this [1,0] variable.) The results are similar to the stress scale expressions of the findings.

Table VI shows those items, in order of importance, which were significantly predictive of a given inpatient classification. The “control” column summarizes distinctions between cases and controls in the aggregate. Families of children with pediatric social illness contrast sharply with the comparison group. These families have less regular health care, many recent moves, many child-rearing problems, and a history of a broken family in the mother’s childhood; they have also experienced mother-initiated separations from the child. These factors suggest several, and somewhat different, patterns of stress on the families of children in the case group. No clear-cut similarities across groups are noted.

The predictors of the specific conditions lead, however, to tentative formulations of etiology which may begin to be translated into a more logical classification scheme. For example, those attributes which are highly predictive of the “child abuse” entity include early and continuing family instability, expressed in mobility, isolation, and early separations of the child from its mother. The familial origins seem prominent, as compared to “failure to thrive,” where

attributes of the child himself sort out as the more significant descriptors. Although the present data do not define pathogenesis, they describe associations which may help inform practice and guide further research.

Implications of Classification of Social Illness in Pediatric Practice

In present clinical practice, whether or not a child’s injuries are characterized as having been “abusively” or “neglectfully” obtained depends on the clinician’s ability-or willingness-to attribute the cause of the symptoms to the child’s parents. The names “battered child syndrome” and “maltreatment syndrome” have formalized the concept of parental fault in the medical literature.^17.18

Making such diagnoses and filing legally mandated case reports have immense value implications which may contradict the traditional ethical posture of medical and behavioral professionals: to help individuals in distress.” As it is rare in practice to know with certainty the exact timing, instrument, and circumstances of children’s injuries, it is not surprising that many are misclassified as “accident victims,” meaning isolated, random events, because of the clinician’s understandable reluctance to implicitly condemn the parents of his patients.

Misclassification and Child Abuse Screening

The matter of misclassification is particularly important when one considers current interest in screening for risk of child abuse. Using those items from this study which are most highly discriminating for child abuse (Table VI), it is possible to construct an equation which would allow one to see the extent to which subjects in the pediatric social illness categories and the control group might be identi6ed or misidenti6ed as being at risk for child abuse at different levels of a scale.

 

Figure 2 expresses as a cumulative percent graph the discriminant function scores for all cases and controls. It is clear that a few characteristics distinguish the child abuse cases from those in the other diagnostic categories. High scores mean that families are similar in these

discriminating attributes to families where child abuse occurred. The difference in the distribution between child abuse and other cases notwithstanding, it may be noted that were one to develop a “quick and dirty” screening instrument on the basis of these features, one would correctly screen in only 75% of the child abuse cases at the level which would include over 30% of the other categories as well.

 

TABLE VII PERCENT MISCLASSIFICATION DERIVED FROM DISCRIMINANT FUNCTIONS FOR INDIVIDUAL DIAGNOSTIC CATEGORIES
Classifying Patients in Group	Using Discriminant Function for Patient Category||
	Inpatient (%)|					Outpatient (%)|
	Abuse	Failure to Thrive	Accident	Ingestion	Control	Accident	Ingestion	Control
Inpatient
Abuse	25º	25	62	25	50	62	68	87
Failure to Thrive	31	19º	24	31	74	64	64	56
Accident	23	8	12º	55	71	66	68	71
Ingestion	21	17	83	21º	39	81	54	62
Control	16	26	56	39	16º	63	28	78
Outpatient
Outpatient Accident	31	5	85	67	61	15º	44	62
Ingestion	42	12	69	75	50	61	19º	61
Control	37	9	74	67	51	71	44	13º

º Percentages marked with asterisks represent “false negative” misclassification; all others are “false positive” misclassification.

Similar equations can be constructed for each diagnostic category. The classification capacity of the set of discriminant functions described in the previous section is summarized in Table VII. (In constructing this table, the cutoff point was set at the mean minus one standard deviation for all discriminant scores [roughly comparable to but generally to the left of the dotted lines discussed in Fig. 2]. A column in Table VII refers to the category for which the discriminant function was calculated; a row refers to the group of cases being classified; entries marked with an asterisk show the impressive percentages of false negatives, i.e., the proportion of each category that was not identified by its own discriminant function, using the [���x-1] cutoff. All other entries are percentages of false positives, i.e., the proportion of a given patient group that would be misclassified by some other discriminant function. Scanning this table indicates that the FTT discriminant function [second column] performs better than the others except on inpatient controls, and that the misclassification is generally large when using the outpatient discriminant functions [last three columns].)

It is well to point out that in the face of rapidly rising numbers of child abuse case reports, protective service institutions across the United States, which even in better economic times were poorly funded and staffed, have had increasingly to resort to rapid clinical screening methods and radical management alternatives to protect victims of child abuse.

Florida’s three-year-old central reporting system for cases of suspected child abuse and neglect is still bogged down in an overload of complaints, currently running at 1.500 to 2,000 per month (87,000 complaints have been made since the state-wide hotline started October 1, 1971). . . . In metropolitan areas they are so swamped, workers limit investigations to complaints which “sound the worst,” says hotline supervisor Mary Ann Price.²⁰

Especially because of known selection bias favoring minority and poor children for the child abuse diagnosis, a phenomenon partly attributable to the public clinical settings in which most of these diagnoses are made and partly to the reluctance of physicians in private practice to make damning value judgments about parents, caution is urged in interpreting these findings to support the value of predictive screening for child abuse. The social policy implications for poor and minority families particularly might be ominous. Other writers have underlined pertinent issues in regard to child abuse screening.^21.22

Further study, focusing more specifically and directly on the major discriminating characteristics, is necessary to disentangle the seemingly causal strands associated with symptoms of pediatric social illness. Before more is known about the process of pathogenesis, the extent and nature of what we already know about misclassification should incline us away from child abuse screening.

In the search for a more etiologic taxonomy of pediatric social illness, we shall have to be vigilant neither to blame the victim by focusing on the parent assumed to be responsible for a child’s injury nor to fulfill the prophecy of risk by a reflexive application of statistical findings.^23.24 A focus on the stresses-and the strengths-associated with the victim, his family, and his life setting may enable us more accurately and humanely to identify, to treat, and to prevent these illnesses.

REFERENCES

1. Newberger EH, Newberger CM, Richmond JB: Child health in America: Toward a rational public policy. Health Society 54:249, 1976.

2. Morse CW, Sahler OJ, Friedman SB: A three-year follow-up study of abused and neglected children. Am J Dis Child 120:439, 1970.

3. Elmer E, Gregg GS: Developmental characteristics of abused children. Pediatrics 40:596, 1967.

4. Martin HA, Beezeley P, Conway EF, Kempe CH: The development of abused children, in Schulman F (ed): Advances in Pediatrics. Chicago, Year Book Medical Publishers Inc, 1974, pp 25-73.

5. Nagi SZ: Child Maltreatment in the United States: A Cry for Help and Organizational Response. Columbus, Ohio, Ohio State University, 1976.

6, Gregg GS. Elmer E: Infant injuries: Accidents or abuse? Pediatrics 44:434, 1969.

7. Newberger EH, Hyde JH: Child abuse: Principles and implications of current pediatric practice. Pediatr Clin North Am 22:695, 1975.

8. Gil DC: Violence Against Children. Cambridge, Mass, Harvard University Press, 1970.

9. Newberger EH (reviewer), Gil DC: Violence Against Children, book review. Pediatrics 48:688. 1971.

10. Koel BS: Failure to thrive and fatal injury as a continuum. Am J Dis Child 118:51, 1969.

11. Holter JC, Friedman SB: Child abuse: Early casefinding in the emergency department. Pediatrics 42: 128, 1969.

12. Martin HL: Antecedents of burns and scalds in children. Br J Med Psychol 43:39, 1970.

13. Bullard PM, Glaser HH, Heagarty MC, Pivchik EC: Failure to thrive in the neglected child. Am J Orthopsychiatry 37:680, 1967.

14. Sobel R: Psychiatric implications of accidental poisonings in childhood. Pediatr Clin North Am 17:653, 1971.

15. Newberger EH, Hagenbuch JJ, Ebeling NB, et al: Reducing the literal and human cost of child abuse: Impact of a new hospital management system. Pediatrics 51:840, 1973.

16. Morse AE, Hyde JN, Newberger EH, Reed RB: Environmental correlates of pediatric social illness: Preventive implications of an advocacy approach. Am J Public Health, to be published.

17. Kempe CH, Silverman FN, Steel BF, et al: The battered child syndrome. JAMA 181:1, 1962.

18. Fontana VJ: The Maltreated Child: The Maltreatment Syndrome in Children, ed 2. Springfield, III, Charles C Thomas Publisher, 1971.

19. Newberger EH, Daniel JH: Knowledge and epidemiology of child abuse: A critical review of concepts. Pediatr Ann 5:140, 1976.

20. Child Protection Report. Washington, DC, March 13, 1975.

21. Light R: Abused and neglected children in America: A study of alternative polices. Harvard Educ Rev 43:556, 1973.

22. Foltz A-M: The development of ambiguous federal policy: Early and periodic screening, diagnosis, and treatment (EPSDT). Health Society 53:35, 1975.

23. Ryan W: Blaming the Victim. New York, Pantheon, 1971.

24. Hobbs N: The Futures of Children. San Francisco, Jossey-Bass, 1975.

ACKNOWLEDGMENT

We thank the following people for their contributions to this work: S. Block, N. Bloom, G. Farrell, G. Gardner, A. Gordon, T. Holtzman, J. Jameson, E. McAnulty, A. Marshall, P. Moriarty, N. Morse, C. M. Newberger, G. Phillips, H. Reynolds, L. Stein, S. Weiser, N. Williams, and M. C. Winokur. The consistent enthusiasm and interest of Drs. C. A. Janeway, chairman emeritus, and J. B. Richmond, present chairman, of the Departments of Medicine and Psychiatry at the Children’s Hospital Medical Center, are also gratefully acknowledged.

Received June 25; revision accepted for publication December 9, 1976.

Presented in part before the Society for Research in Child Development, Denver, April 11, 1975.

Supported by a grant from the Office of Child Development, Department of Health, Education and Welfare (Project OCD-CB-141).