The Neglect Scale - SAGE Journals - Sage Publications

15 downloads 0 Views 115KB Size Report
Harrington et al. / NEGLECT SCALE. CHILD MALTREATMENT / NOVEMBER 2002. The Neglect Scale: Confirmatory Factor Analyses in a Low-Income Sample.

10.1177/107755902237266 Harrington CHILD MALTREATMENT et al. / NEGLECT / NOVEMBER SCALE 2002

The Neglect Scale: Confirmatory Factor Analyses in a Low-Income Sample Donna Harrington Susan Zuravin Diane DePanfilis Laura Ting University of Maryland School of Social Work Howard Dubowitz University of Maryland School of Medicine

The Neglect Scale is an easy-to-administer, retrospective, selfreport measure of neglect. Research conducted by Straus and colleagues with college students indicates that this scale has a high level of internal consistency reliability and moderate construct validity. The purpose of this article is to examine the reliability and validity of the Neglect Scale when used with a low-income, inner-city sample. The sample included 151 women who were participating in a neglect prevention demonstration project. The Neglect Scale was completed as part of a computer-administered baseline interview before services were provided. To assess whether the 20-item, four-factor structure reported by Straus et al. fit the data from this sample, a confirmatory factor analysis was performed; the model did not fit the data well. Additional analyses identified a model that did fit the data well and suggest that the Neglect Scale is a promising self-report measure.

Child neglect is the most common of the four types

of child maltreatment, has increased over the years, and results in more harmed children than any other type of maltreatment. Data from the 1998 National Child Abuse and Neglect Data System (NCANDS) (Department of Health and Human Services [DHHS], 2000), based on official reports for child abuse and neglect, reveal that more than twice as many children reported to Child Protective Service (CPS) agencies were neglected as physically abused CHILD MALTREATMENT, Vol. 7, No. 4, November 2002 359-368 DOI: 10.1177/107755902237266 © 2002 Sage Publications

(55.9% to 22.7%) and almost five times as many children were reported for physical neglect as sexual abuse (55.9% to 11.5%). Moreover, these data show that from 1990 to 1998, the number of children reported for neglect per 1,000 children has increased (6.3/1,000 to 7.2/1,000), whereas the other three types have declined—physical abuse from 3.5/1,000 to 2.9/1,000, sexual abuse from 2.3/1,000 to 1.6/ 1,000, and emotional maltreatment from .8/1,000 to .6/1,000. In addition, the Third National Incidence Study of Child Abuse and Neglect (NIS-3) (DHHS, 1996) found that more neglected children suffered harmful consequences: 879,000 children met the harm standard for neglect whereas fewer than half met the same standard for physical abuse (381,700), and less than a quarter met it for sexual (217,700) and emotional abuse (204,500). On the basis of these findings, neglect might be expected to be the focus of considerably more research and attention than the other three types of maltreatment; however, it is not. Physical abuse and, particularly, sexual abuse have attracted more research attention than neglect even though child maltreatment researchers have been bringing “the neglect of neglect” to the field’s notice for several years (Dubowitz, 1994; Wolock & Horowitz, 1984). Authors’ Note: Development of this article was partially supported by Grant No. 90CA1580 from the Children’s Bureau, U.S. Department of Heath and Human Services to Diane DePanfilis, Principle Investigator. Portions of this article were presented at the 7th International Family Violence Research Conference, July 22-25, 2001, held in Portsmouth, NH. 359


Harrington et al. / NEGLECT SCALE

Why Has Neglect Been Neglected? Why is the body of empirical knowledge on neglect so small relative to that for physical and sexual abuse? Wolock and Horowitz (1984) attribute the “neglect of neglect” to two factors: (a) “the prestige of the definers of physical abuse (well-known medical doctors) and its reception and recognition within the wider medical community” (p. 535) and (b) “the link between neglect and poverty, reflecting in essence the low priority accorded to the alleviation of poverty” (p. 536). Dubowitz (1994), writing 10 years later in an article titled, “Neglecting the neglect of neglect,” suggested, “child neglect may not attract attention due to the inherent vagueness of what constitutes neglect” (p. 558). He noted, “that some of the confusion surrounding a definition of child neglect stems from the lack of a consensus at a conceptual level” (p. 558). Despite the probable truth of Wolock and Horowitz’s (1984) rationales for the neglect of neglect, it may be neglected primarily because it is more difficult to define than the other two types of maltreatment. Fifteen years ago, the National Center on Child Abuse and Neglect (1987) recognized the field’s failure to develop a standard definition of neglect as a critical problem for developing knowledge. They strongly recommended more effort be directed at definitional issues. Recently, Straus, Kinard, and Williams (1995) noted “the relative lack of research on neglect may be due to the absence of a brief yet valid measure that can be used in epidemiological research” (p. 1). Today, there is still no standard operational definition of neglect with known psychometric properties that is used by a majority of the researchers in the field. Most researchers use their own definitions, making it very difficult to integrate findings across studies and develop a body of knowledge. The aim of this article is to contribute to psychometric knowledge about one self-report operational definition of neglect, the Neglect Scale (Straus et al., 1995). More specifically, our article presents findings from internal consistency and confirmatory factor analyses of this scale in a low-income sample. The Developmental History of Neglect Definitions Why has the field failed to develop a standard, wellvalidated definition of neglect? A glance at the developmental history of the neglect definition is helpful for elucidating current problems. First, a CPS label most often operationally defines neglect. In other words, if the CPS case record or the central registry identified the child as having been substantiated by the CPS caseworker for neglect, the type of maltreat-

ment was identified as neglect. With caseworkers interpreting neglect in their own idiosyncratic ways, the “neglect” label includes varied circumstances that differ greatly within and among agencies. This makes it difficult to interpret and integrate findings from research using this label. Second, as some researchers began to get more specific regarding the definition of neglect, definers divided their neglect definitions into dimensions (neglect, unlike physical abuse and sexual abuse is multidimensional—i.e., supervisory neglect, physical neglect, educational neglect, etc.) and the dimensions were sometimes further divided into subtypes (Zuravin, 1999). CPS records or caseworkers were used as sources of data for these types. Unfortunately, however, each researcher conceptualized and operationalized their dimensions and subtypes differently. In other words, the number and type of dimensions, the number and identity of each subtype, whether frequency and severity of each subtype was included in the definition, and the operational translations proposed by the various definers differed across models. For example, the NIS-3 (DHHS, 1996) addressed three types of neglect: physical, educational, and emotional. Each of these three dimensions was then subdivided into subtypes; for example, physical neglect included seven different subtypes. Trocme (1996) divided neglect into six dimensions— supervision, nutrition, clothing and hygiene, physical health care, mental health care, and developmental educational care. No subtypes were included under the dimensions, but each was rated for severity and age of child. Zuravin and DePanfilis (1996) divided neglect into 14 dimensions with subtypes under most. Such a diversity of definitions does little to help the field to build a body of knowledge. Third, a few researchers (McGee, Wolfe, Yuen, Wilson, & Carnochan, 1995; Straus et al., 1995) have initiated, as recommended by Straus and colleagues (1995), the development of brief self-report measures of neglect. The 20-item version of the Straus et al. (1995) Neglect Scale divided neglect into four dimensions: supervisory, physical, cognitive, and emotional needs. Each dimension was operationally defined by a set of five items. The McGee and colleagues’ (1995) Neglect Scale includes 14 items, 7 for omissions in care and 7 for positive parenting practices. Research on Neglect Definitions Psychometric and other types of research on definitions are extremely important for validating measures, but unfortunately not much is extant in the published literature. To identify measurement research on neglect we hand-searched six journals— CHILD MALTREATMENT / NOVEMBER 2002

Harrington et al. / NEGLECT SCALE

The International Journal of Child Abuse and Neglect, Child Maltreatment, Development and Psychopathology, Journal of Family Violence, Child Welfare, and Journal of Interpersonal Violence—for 5 years (1995 through 1999), or from origin if less than 5 years, and then searched the reference list of each article. We found only six articles (Gaudin, Polansky, & Kilpatrick, 1992; Kaufman, Jones, Stieglitz, Vitulano, & Mannarino, 1994; Manly, Cicchetti, & Barnett, 1994; McGee et al., 1995; Trocme, 1996; Wolfe & McGee, 1994) that presented findings from one or more types of measurementoriented research. A seventh article was identified through a conference presentation (Straus et al., 1995). Five of the articles were authored by different sets of researchers, two were authored by the same team of researchers (McGee et al., 1995; Wolfe & McGee, 1994), and all but one (Gaudin et al., 1992) presented findings on a definition developed by the authors. With the exception of Trocme (1996), authors did not present findings from a comprehensive psychometric analysis of the measure. The most often reported findings were on construct validity (Kaufman et al., 1994; Manly et al., 1994; McGee et al., 1995; Wolfe & McGee, 1994). Two of the most interesting and important sets of findings were reported by McGee et al. (1995) and Kaufman et al. (1994). Findings from both studies have negative implications for the two most popular sources of data for maltreatment definitions: CPS case records and caseworker ratings. Both teams focused on determining whether different sources of data would provide the same or different findings regarding the child’s maltreatment experience(s). McGee et al. (1995) compared three data sources—CPS case records, CPS caseworker ratings, and adolescent selfratings. Findings showed fairly large disagreements regarding the occurrence of neglect with the adolescents. The CPS records and CPS caseworkers agreed 82% of the time that the adolescents had been neglected, whereas the CPS records and CPS caseworkers agreed with the adolescents only 59% and 65% of the time, respectively. With regard to neglect severity, there was a moderate correlation between CPS case record and CPS caseworker data sources (.52) whereas the severity correlations of case record and caseworker with adolescent self-ratings were very low (.32 and .02, respectively). The construct validity findings support the above results. They showed that adolescent self-ratings of neglect had significant predictive power for self-rated behavioral problems on the Achenbach Youth Self-Report (Achenbach, 1991) after controlling for case record and caseworker ratings, although case records and caseworker ratings CHILD MALTREATMENT / NOVEMBER 2002


had no significant predictive power for overall, externalizing, or internalizing behavior problems after controlling for adolescent self-ratings. Our decision to perform psychometric analyses on the Neglect Scale (Straus et al., 1995), a retrospective, self-report instrument, was predicated on two factors. First and foremost are the findings from the McGee et al. (1995) and Kaufman et al. (1994) studies, both of which suggest limitations in CPS case records and CPS caseworkers as sources of information for neglect definitions. Second, we agree with Straus et al. (1995) that the “relative lack of research on neglect may be the absence of a brief yet valid measure that can be used in epidemiological research. Such measures exist for sexual abuse (Finkelhor, 1986) and physical abuse (Straus, 1979; Straus & Hamby, 1995) and their use has made important contributions” (pp. 1-2) to knowledge on both. METHOD

Sample A total of 151 maternal caregivers were selected from 154 (female and male) caregivers who were originally recruited for participation in a federally funded demonstration program to help families prevent child neglect. Participants were self-referred or referred through schools, health centers, vocational training facilities, and other community agencies because they were identified as a family “at risk for neglect.” Families met inclusion criteria if (a) they had at least one child between the ages of 5 and 11 at home who lived or went to school in a specific geographic location within the city, (b) the participant and family did not have any current involvement with CPS, (c) the family was considered at risk for at least 1 of 19 defined subtypes of neglect (the Appendix provides the 19 subtypes and references for each), (d) at least two risk factors related to the child, parent, or family were identified (e.g., caregiver unemployment, caregiver mental health problem, caregiver alcohol/drug problem, homelessness, more than three children in the family, domestic violence, child behavior/mental health problem, and child physical, learning, or developmental disability), (e) and the primary caregiver was willing to accept services. A total of 212 referrals met these criteria; 154 caregivers were willing to participate in services. For the purposes of this study, 3 male caregivers were excluded leaving a final sample of 151 respondents. Caregivers in this study (N = 151) represent characteristics that are reflective of the high-risk neighborhoods that were the target of this intervention. The


Harrington et al. / NEGLECT SCALE

mean age of caregivers was 39.13 (s = 12.3, range 21-78 years) with educational levels of respondents ranging from third grade to college degrees (mean = 10.79 years; s = 2.22). Approximately two thirds of the respondents were never married (64.9%, n = 98), 5.3% were married (n = 8), 9.9% were divorced (n = 15), 12.6% were separated (n = 19), and 7.3% were widowed (n = 11). The majority of respondents were African American (86.8%, n = 131), whereas 10.6% were Caucasian (n = 16) and the remainder identified themselves as mixed race, Hispanic, or other (2.7%, n = 4). At the time they entered the program (1997-2000), the majority were unemployed, and not receiving any training (57%, n = 86). Some worked full-time (19.2%, n = 29), some worked part-time (8.6%, n = 13), some were in school or receiving training (10.6%, n = 16), and a few were retired (4.6%, n = 7). Although respondents could report more than one source of income, the majority received Temporary Assistance to Needy Families (TANF) (58.3%, n = 88), food stamps (70.9%, n = 107), and Medicaid (55.6%, n = 84). The mean annual income for families was $9,586.87 (s = $5,824). The mean number of children in these families was 3.1 (s = 1.58, range of 1-9) with a total of 461 children in these 151 families. Procedure The Neglect Scale (Straus et al., 1995) was administered during the baseline interview for a neglectprevention demonstration project. Respondents selfadministered a computer-assisted interview that asked them questions about their growing up experiences, stresses and strains in their lives today, behavior and emotional functioning of their children, their use of alcohol or other drugs, their emotional functioning, their parenting knowledge and attitudes, and their neighborhoods and housing situations. Measure The Neglect Scale (Straus et al., 1995). The Neglect Scale was developed to provide a retrospective, easily administered, standardized measure of neglect. A pool of 63 items was generated based on expertise in the area of child abuse and neglect and existing measures of child maltreatment; 40 items were selected for the test development version (see Table 1 for the items). Respondents rate the degree to which they agree with each item on a 4-point Likert-type scale (1 = strongly agree to 4 = strongly disagree). The 40-item version was administered to a sample of 377 undergraduates at a state university in New England. In the Straus

et al. (1995) sample, “96% were white, 54% of the fathers and 39% of the mothers of the respondents were college educated, and 72% were from households with both biological parents” (p. 4). Positive items (e.g., “kept the house clean”) were reverse coded prior to conducting psychometric analyses, and item analyses were conducted to select a subset of 20 items by eliminating “the half of the items least highly correlated with the total score of the respective scales” (Straus et al., 1995, p. 5). Internal consistency reliability was very good, with alphas of .80 to .89 for the four subscales and .93 for the full 20-item scale. Straus et al. (1995) conducted two principal components analyses on the 20-item scale, one with orthogonal (Varimax) rotation and one with oblique rotation. The orthogonal rotation yielded two components. Twelve items had loadings of .41 or greater on the first component, with the five emotional needs items loading most highly, followed by the five cognitive needs items; two supervisory needs items loaded at .41 and .45, although one of these items loaded slightly higher (.46) on component two. Eight items had loadings of .40 or greater on component two, with four supervisory needs and four physical needs items loading most highly; one supervisory needs and one physical needs item loaded below .40 on the second component, both were positive items. All emotional and cognitive needs items loaded below .40 on the second component. According to Straus et al. (1995), “the oblique analysis produced two factors almost identical in nature to those described above . . . [the two factors] correlated .57 with each other” (p. 6). Data Analyses Cronbach’s alpha was used to examine internal consistency reliability of the full 40-item scale and the four 5-item subscales. Alphas if item deleted were examined to identify items that may not fit well with each subscale. A confirmatory factor analysis of Straus et al.’s (1995) 20-item four-factor structure was performed. Raykov, Tomer, and Nesselroade (1991) recommend the following fit indices be used to assess a model’s goodness-of-fit: chi-square, its degrees of freedom (df) and p value, goodness-of-fit index (GFI), normed fit index (NFI), non-normed fit index (NNFI), and comparative fit index (CFI). In addition, because the significance of chi-square is dependent on sample size, they recommend that the ratio of chi-square to df also be reported. Ideally, the chi-square will be


Harrington et al. / NEGLECT SCALE



Neglect Scale Items With Subscale Identification, Means, and Standard Deviations

Item 1. Took me places where I could learn things like a zoo or a library. 2. Paid no attention to me. 3. Made sure I saw a dentist when I needed one. 4. Did not care if what I did was right or wrong. 5. Left me alone without an adult when I was too young for that. 6. Did not hug me. 7. Locked me out of the house on purpose. 8. Ignored my feelings about things. 9. Did things with me just for fun. 10. Made sure I got enough sleep. 11. Were not interested in my activities or hobbies. 12. Did not help me with homework. 13. Did not give me clean clothes. 14. Wanted to know what I was doing when I was not home. 15. Were proud when I succeeded in something. 16. Did not comfort me when I was upset. 17. Did not keep me clean. 18. Did not help me to do my best. 19. Did not make sure I went to school. 20. Did not care if I got into trouble in school. 21. Did not make sure I saw a doctor when I needed one. 22. Were not interested in the kind of friends I had. 23. Did not make sure I did my homework. 24. Did not give me enough to eat. 25. Helped me when I had trouble understanding something. 26. Forgot about things they were supposed to do for me. 27. Did not read books to me. 28. Did not help me when I had problems. 29. Did not praise me. 30. Did not care if I did things like shoplifting. 31. Did not show interest in my grades at school. 32. Did not tell me they loved me. 33. Let me know when I did something right. 34. Put clear limits on what I was allowed to do. 35. Did not give reasons for wanting me to do something. 36. Worried about my getting into trouble after school. 37. Kept the house clean. 38. Did not give me enough clothes to keep warm. 39. Did not talk about things in the news with me. 40. Did not give me presents for my birthday or holidays.

Straus Scale b

A40 A40 A40 A40 A40 A40 A40 A40 e A20-E A40 f A20-C A20-C A40 g A20-S A40 A20-E h A20-P A20-C A20-S A20-S A20-P A20-S A40 A20-P A20-C A40 A20-C A20-E A20-E A20-S A40 A20-E A40 A40 A40 A40 A20-P A20-P A40 A40

Conceptual Scale c

Cognitive c Emotional c Physical d Super/Emot c Supervisory c Emotional d Super/Phys/Emot c Emotional d Emot/Cog d Phys/Super d Emot/Cog c Cognitive c Physical c Supervisory d Emot/Cog c Emotional c Physical d Cog/Emot d Super/Cog d Super/Cog c Physical d Super/Emot d Cog/Super c Physical c Cognitive c Emotional d Cog/Super d Emot/Super c Emotional c Supervisory d Cog/Super/Emot c Emotional d Emot/Cog c Supervisory d Cog/Emot/Super c Supervisory c Physical c Physical c Cognitive c Emotional

Mean 3.01 3.17 3.09 3.19 3.38 3.09 3.48 3.16 2.94 3.32 2.99 2.96 3.45 3.19 3.41 3.02 3.50 3.16 3.40 3.40 3.32 3.19 3.24 3.48 2.97 3.05 2.74 2.99 2.86 3.50 3.28 3.07 3.07 3.27 2.62 2.76 3.46 3.32 2.73 3.24


Standard Deviation .93 .85 .89 .95 .84 .96 .69 .86 .90 .72 .89 .86 .70 .87 .75 .88 .68 .87 .75 .75 .80 .83 .80 .69 .94 .84 .92 .84 .91 .66 .84 .99 .94 .86 .95 1.02 .75 .85 .92 .84

a. Means and standard deviations calculated after reverse coding of positive items. b. Item from Straus, Kinard, and Williams (1995) 40-item scale. c. All seven raters agreed on placement. d. Raters disagreed—primary placement listed first, secondary placement listed second. e. Item from Straus et al. 20-item scale, Emotional subscale. f. Item from Straus et al. 20-item scale, Cognitive subscale. g. Item from Straus et al. 20-item scale, Supervisory subscale. h. Item from Straus et al. 20-item scale, Physical subscale.

nonsignificant; the ratio of chi-square to df will be less than two; the GFI, NFI, NNFI, and CFI will all be greater than .90; and the critical N will be greater than 200. Because the confirmatory factor analysis indicated that the 20-item four-factor model did not fit the data well (see Figure 1 and Table 2), item-total correlations were examined to determine if the same items would be identified for the current samCHILD MALTREATMENT / NOVEMBER 2002

ple as were identified by Straus et al. (1995). They were not. Therefore, seven child neglect researchers were asked to independently place each of the 40 items on the Neglect Scale test development version (Straus et al., 1995) on one of the four subscales identified by Straus and Colleagues—Emotional, Cognitive, Supervisory, or Physical needs. See Table 1 for factor placement of all 40 items according to Straus et al. and the


Harrington et al. / NEGLECT SCALE

Did not help me when I had problems (28)

Did not help me to do my best (18)

.88 .75

Did not comfort me when I was upset (16)


Did not praise me (29)


Emotional Needs


Helped me … trouble understanding (25)

.30 .97

Did not read books to me (27) .72


Did not tell me they loved me (32)

…not interested in my activities/hobbies (11)


Did not help me with homework (12)

Cognitive Needs


Did things with me just for fun (9)



.79 .77

Did not make sure I went to school (19) .81

Did not care … trouble in school (20)


Did not keep me clean (17)

Supervisory Needs

Did not make sure I saw a doctor… (21)

.83 .80 .73

Did not care … things like shoplifting (30)

Did not give me enough clothes… (38)

.81 .97 .76

Were not interested in … friends I had (22)


… know… doing when not home (14)

Figure 1:

Did not give me enough to eat (24)

Physical Needs

.92 .29

Kept the house clean (37)

Confirmatory Factor Analysis of Straus, Kinard, and Williams (1995) 20-Item Neglect Scale With Four Factors

researchers. The researchers unanimously agreed on placement of 24 of the 40 items; these 24 items were used in a second confirmatory factor analysis to test whether the factor structure identified by the researchers fit the data. The 16 items that the researchers did not agree on were not included in further analyses. Modification indices were assessed and reasonable changes were made to the model.

9 (reverse coded) would result in an alpha of .88; for the Physical subscale, removal of item 37 (reverse coded) would result in an alpha of .87; for the Cognitive subscale, removal of item 25 (reverse coded) would result in an alpha of .83; and for the Supervisory subscale, removal of item 14 (reversed coded) would result in an alpha of .87. Deletion of any of the negatively phrased items from any of the four subscales would lower the alpha.


Confirmatory Factor Analyses

Internal Consistency Reliability Internal consistency reliability for the 40-item version was very high (alpha = .96) and was moderate for the four 5-item subscales identified by Straus et al. (1995), with alpha = .85 for Emotional, .82 for Physical, .78 for Cognitive, and .81 for Supervisory. Each of the 5-item subscales consists of 1 positive and 4 negative items. Examination of the alphas if item-deleted indices indicated that for all four subscales, alpha would be improved if the one positive item was deleted. For the Emotional subscale, removal of item

Straus et al.’s (1995) 20-item four-factor model. A confirmatory factor analysis of Straus et al.’s 20-item fourfactor structure was performed (see Figure 1). All items loaded significantly (p < .05) on the factors identified by Straus et al. and all four factors were significantly correlated with each other. Consistent with the principal components analyses reported by Straus et al., the emotional and cognitive needs factors were highly correlated with each other (r = .97) and the supervisory and physical needs factors were highly correlated with each other (r = .97). Other correlations CHILD MALTREATMENT / NOVEMBER 2002

Harrington et al. / NEGLECT SCALE



Fit Indices for Confirmatory Factor Analyses

Fit Index Chi-squared df Significance Chi-square/df Goodness-of-fit index Normed fit index Non-normed fit index Comparative fit index Critical N

Straus et al. 4-Factor 394.39 164 < .005 2.40 .79 .82 .86 .88 80.51

among factors ranged from .60 for physical and emotional needs to .79 for supervisory and cognitive needs. Overall, the fit indices (see Table 2) suggest that the data from this sample do not fit the four-factor model well. Examination of the modification indices for this model indicated that three of the supervisory needs items (14, 19, and 22) were also indicators of physical needs. In addition, one of the supervisory needs items (22) and one of the cognitive needs items (27) were also indicators of emotional needs. Elimination of the three supervisory needs items that were indicators of other constructs would have resulted in only having two indicators of supervisory needs. Modification indices also suggested that model fit could be improved by allowing several of the indicators to have correlated errors, especially for the positive items. Revised four-factor model. A revised model based on the 24 items unanimously identified by the panel as indicative of the four factors was tested. The 24-item four-factor model did not fit the data better than the original Straus et al. (1995) four-factor model (see Table 2). Modification indices suggested that five items (Items 3, 26, 29, 34, and 40) were related to more than one construct. For example, Item 29, “Did not praise me,” which was identified by the panel as an indicator of emotional needs, was also related to supervisory and physical needs. Removal of these five items resulted in a better fitting model (see Table 2 for Rev ised 4-Factor Mod ification 1 fit in d ices ) . Modification indices also suggested that allowing the error variances of some of the positive items to correlate would improve the fit of the model. Therefore, in the Revised 4-Factor Modification 2 model, three of the positive items (Items 14, 25, and 37) were allowed to have correlated error variances, and one item (Item 36) was removed because it did not load significantly. The data fit this model well (see Table 2 and Figure 2). CHILD MALTREATMENT / NOVEMBER 2002

Revised 4-Factor 603.75 246 < .005 2.45 .72 .72 .79 .81 75.66

Revised 4-Factor Modification 1

Revised 4-Factor Modification 2

279.34 146 < .005 1.91 .82 .82 .89 .90 102.31

178.18 126 .002 1.41 .88 .88 .95 .96 140.61


Demographically, the Straus et al. (1995) sample and the current sample are very different, and it is likely that the sample differences account for at least some of the discrepancies in the findings. In the current sample, interviewers reported that several respondents had some difficulty understanding the items and response scale, especially the double-negative items (e.g., “Did not tell me they loved me,” which requires a response of strongly disagree or disagree to indicate that parents did tell the respondent they loved them). Straus et al. did not report any difficulty understanding the items in their sample, but it seems likely that the women in the current study, with a mean educational level of less than 11 years and some reporting as little as a third-grade education, may have had more difficulty comprehending the items than did the college students in Straus and colleagues’ sample. Given that research on neglect is often conducted with samples that may have limited educational levels, it is very important that any selfreport scale be comprehensible for respondents with potentially limited reading levels. Perhaps the strongest difference in the findings from the two samples is in the use and interpretation of the positive items. This may in part relate to the comprehension issue raised above. Because only one quarter of the items were positive, the respondents in the current study may have become accustomed to responding to the more frequently occurring negative items and then had difficulty responding in the opposite direction as was required by the positive items. In the principal components analysis reported by Straus et al. (1995), the positive items had loadings that were similar to the negative items for the emotional and cognitive subscales and were lower than the negative items for the supervisory and physical subscales. In the current study, the positive items con-


Harrington et al. / NEGLECT SCALE

Paid no attention to me (2)

Did not talk about … news with me (39)

.50 .77

Did not comfort me when I was upset (16)

Emotional Needs


Helped me … trouble understanding (25)

.74 .27

Ignored my feelings about things (8)



Took me places … could learn (1) .40


Did not help me with homework (12)

Did not hug me (6) Cognitive Needs


Did not tell me they loved me (32)



.65 .67 .67

Left me alone… too young (5)

Did not keep me clean (17)

Supervisory Needs

Did not make sure I saw a doctor… (21)

.81 .82 .75

Did not care … things like shoplifting (30)

Did not give me enough clothes… (38)

.31 .95 .75

Wanted to know… when not home (14)

Did not give me enough to eat (24)

Physical Needs

.94 .34

Kept the house clean (37)


Did not give me clean clothes (13)


Confirmatory Factor Analysis of Neglect Scale Revised Four-Factor Model

sistently had loadings much lower than the negative items in the confirmatory factor analysis (see Figures 1 and 2). Alternatively, the positive items may not have fit well with the negative items in the current study because they do not represent the same construct. Neglect involves the omission or failure to provide for basic needs; therefore, it is not clear that the absence of positive experiences (e.g., “did things with me just for fun” and “helped me when I had trouble understanding something”) would necessarily be indicative of neglect. Strengths and Limitations This study is based on a sample of 151 low-income, predominantly African American women, who live in a large mid-Atlantic city; therefore, the generalizability of these findings may be limited to similar populations. Given that this sample is so different from that reported on by Straus et al. (1995), when the two studies yielded similar findings, there may be reason to think that the findings are reasonably generaliz-

able. For example, the 11 items (Items 12, 14, 16, 17, 21, 24, 25, 30, 32, 37, and 38) included in both the 20item Straus et al. version and the final model for the current study may provide a good, brief neglect scale (alpha = .86 for these 11 items in this sample). The four-factor model depicted in Figure 2 is the result of a combined conceptual and empirical approach to identifying the factor structure of the Neglect Scale (Straus et al., 1995). The conceptual approach began with Straus and colleagues’ (1995) assertion that the Neglect Scale consists of items representing emotional, cognitive, supervisory and physical needs. When the four-factor model presented by Straus et al. did not fit the data from the current sample, seven neglect researchers independently placed items on one of the four subscales. The items unanimously agreed on by the researchers were then used in a confirmatory factor analysis, with modification indices used to guide changes to the model. This revised model has several strengths: (a) The data fit the model well, (b) the items identified for each subscale all have high content (or face) validity, (c) CHILD MALTREATMENT / NOVEMBER 2002

Harrington et al. / NEGLECT SCALE

the model includes both positive and negative items, which should decrease response set bias, and (d) although physical and supervisory needs are still highly correlated, the correlation between emotional and cognitive needs is much lower in the revised model than in the original Straus et al. model. Implications The Neglect Scale (Straus et al., 1995) is a promising self-report measure of neglect that preliminary analyses suggest is worth further research and development. It is noteworthy that the measure is premised on assumptions of what constitutes neglect; however, the impact or risks of some of the experiences included remains uncertain. Another issue pertains to the extent to which those who have not had their needs met are able and/or willing to attribute these problems to their parents or caregivers. Developmentally, it may be important that someone “makes sure [a child] went to school”; however, it may not necessarily have to be the parents who do so if this need is met by other family or friends. Additional research on this measure is needed to establish both criterion and construct validity; it would be especially useful to examine how the Neglect Scale relates to more commonly used definitions of neglect, such as the CPS definitions discussed earlier. In addition, the measure needs to be used with different samples to establish its external validity. Finally, a retrospective measure of neglect administered to adults cannot fully address many of the operational definition issues raised in the introduction. Future research utilizing this measure with adolescents is needed; a pictorial version for children is being developed (Kantor et al., 2001) that may also address many of these issues. Scale development would benefit from continued use of the full 40-item test development version until a subset of items can be identified that is appropriate for a variety of populations. If the full 40 items cannot be used, the 11 items that fit the data in both the Straus et al. (1995) and current studies may be a promising short version to consider.

APPENDIX Subtypes of Neglect 1. Inadequate/delayed health care—failure of a child to receive needed care for physical injury, acute illnesses, physical disabilities, or chronic condition or impairment that if left untreated could result in negative consequences for the child (Magura & Moses, 1986; U.S. Department of Health and Human Services [USDHHS], 1988; Zuravin & DePanfilis, 1996). CHILD MALTREATMENT / NOVEMBER 2002


2. Inadequate nutrition—failure to provide a child with regular and ample meals that meet basic nutritional requirements, or when a caregiver fails to provide the necessary rehabilitative diet to a child with particular types of physical health problems (Zuravin & DePanfilis, 1996). 3. Poor personal hygiene—failure to attend to cleanliness of the child’s hair, skin, teeth, and clothes (Magura & Moses, 1986; Zuravin & DePanfilis, 1996). 4. Inadequate clothing—chronic inappropriate clothing for the weather or conditions (Magura & Moses, 1986). 5. Unsafe household conditions—presence of obvious hazardous physical conditions in the home that could result in negative consequences for the child (Magura & Moses, 1986; Zuravin & DePanfilis, 1996). 6. Unsanitary household conditions—presence of obvious hazardous unsanitary conditions in the home (Magura & Moses, 1986; Zuravin & DePanfilis, 1996). 7. Unstable living conditions—moves of residence due to eviction or lack of planning at least three times within a 6month period or homelessness due to the lack of available, affordable housing or the caregiver’s inability to manage finances (Zuravin & DePanfilis, 1996). 8. Shuttling—the child is repeatedly left at one household or another due to apparent unwillingness to maintain custody, or chronically and repeatedly leaving a child with others for days/weeks at a time (USDHHS, 1988; Zuravin & DePanfilis, 1996). 9. Inadequate supervision—child left unsupervised or inadequately supervised for extended periods of time or allowed to remain away from home overnight without the caregiver knowing the child’s whereabouts (USDHHS, 1988). 10. Inappropriate substitute caregiver—failure to arrange for safe and appropriate substitute child care when the caregiver leaves child with an inappropriate caregiver (Magura & Moses, 1986; Zuravin & DePanfilis, 1996). 11. Drug-exposed newborn—a newborn infant has been exposed to drugs because the mother has used one or more illegal substances during her pregnancy (National Council of Juvenile and Family Court Judges, 1992). 12. Inadequate nurturance or affection—marked inattention to the child’s needs for affection, emotional support, attention, or competence; being detached or uninvolved, interacting only when absolutely necessary, failing to express affection, caring, and love for the child (American Professional Society on the Abuse of Children, 1995; USDHHS, 1988). 13. Isolating—the child is consistently denied opportunities to meet needs for interacting/communicating with peers or adults inside or outside the home; markedly overprotective restrictions that foster immaturity or emotional overdependency; chronically applying expectations clearly inappropriate in relation to the child’s age or level of development; inattention to the child’s developmental/emotional needs (American Professional Society on the Abuse of Children, 1995; USDHHS, 1988). 14. Witnessing violence—a child witnesses violence in the home, such as partner abuse or violence between other per-


Harrington et al. / NEGLECT SCALE

sons who visit the home on a regular basis (USDHHS, 1988). 15. Permitting alcohol or drug use—encouraging or permitting of drug or alcohol use by a child (USDHHS, 1988). 16. Permitting other maladaptive behavior—encouraging or permitting of other maladaptive behavior (e.g., severe assaultiveness, chronic delinquency) under circumstances where the caregiver had reason to be aware of the existence and seriousness of the problem but did not attempt to intervene (USDHHS, 1988). 17. Delay in obtaining needed mental health care—a child is not provided needed treatment for an emotional or behavioral impairment (USDHHS, 1988; Zuravin & DePanfilis, 1996). 18. Chronic truancy—a child (age 6) is not enrolled in school or habitual truancy (minimum of 20 days) without a legitimate reason (USDHHS, 1988; Zuravin & DePanfilis, 1996). 19. Unmet special education needs—a child fails to receive recommended remedial educational services, or treatment for a child’s diagnosed learning disorder or other special educational needs or problems of the child (American Professional Society on the Abuse of Children, 1995; USDHHS, 1988).

Raykov, T., Tomer, A., & Nesselroade, J. R. (1991). Reporting structural equation modeling results in Psychology in Aging: Some proposed guidelines. Psychology and Aging, 6, 499-503. Straus, M., & Hamby, S. (1995). Measuring psychological and physical abuse of children with the conflict tactics scale. Unpublished manuscript, University of New Hampshire. Straus, M. A., Kinard, E. M., & Williams, L. M. (1995, July 23). The Neglect Scale. Paper presented at the Fourth International Conference on Family Violence Research, Durham, NH. Trocme, N. (1996). Development and preliminary evaluation of the Ontario Neglect Index. Child Maltreatment, 1, 145-155. U.S. Department of Health and Human Services, National Center on Child Abuse and Neglect. (1988). Study findings: Study of national incidence and prevalence of child abuse and neglect,1988. Washington, DC: Government Printing Office. Wolfe, D., & McGee, R. (1994). Dimensions of maltreatment and their relationship to adolescent maltreatment. Development and Psychopathology, 6, 166-181. Wolock, I., & Horowitz, B. (1984). Child maltreatment as a social problem: The neglect of neglect. American Journal of Orthopsychiatry, 59, 377-389. Zuravin, S. (1999). Child neglect: A review of definitions and measurement research. In H. Dubowitz (Ed.), Neglected children: Research, practice, and policy (pp. 24-46). Thousand Oaks, CA: Sage. Zuravin, S., & DePanfilis, D. (1996). Child maltreatment recurrences among families served by Child Protective Services. Final report to the National Center on Child Abuse and Neglect; Grant #90CA1497.


Donna Harrington, Ph.D., is an associate professor and director of the Ruth H. Young Child Welfare Center at the University of Maryland School of Social Work.

Achenbach, T. (1991). Manual for the child behavior checklist/4-18 and 1991 profile. Burlington: University of Vermont. American Professional Society on the Abuse of Children. (1995). Psychosocial evaluation of suspected psychological maltreatment in children and adolescents. Chicago, IL: Author. Department of Health and Human Services. (1996). National study of the incidence and severity of child abuse and neglect. Washington, DC: Author. Department of Health and Human Services. (2000). National child abuse and neglect data system. Washington, DC: Author. Dubowitz, H. (1994). Neglecting the neglect of neglect. Journal of Interpersonal Violence, 9, 556-560. Gaudin, J., Polansky, N., & Kilpatrick, A. (1992). The child well being scales: A field trial. Child Welfare, 71, 319-328. Kaufman, G. K., Straus, M. A., Mebert, C., Brown, W., Macallum, C. A., & Flannery, T. (2001, July 22-25). Assessment of child neglect in community and clinical samples: Development of the multidimensional neglect scale for child self-report. Paper presented at the 7th International Family Violence Research Conference, Portsmouth, NH. Kaufman, J., Jones, B., Stiglitz, E., Vitulano, I., & Mannarino, A. (1994). The use of multiple informants to assess children’s maltreatment experiences. Journal of Family Violence, 9, 227-248. Magura, S., & Moses, B. S. (1986). Outcome measures for child welfare services. Washington, DC: Child Welfare League of America. Manly, J., Cicchetti, D., & Barnett, D. (1994). The impact of subtype, frequency, chronicity, and severity of child maltreatment on social competence and behavior problems. Development and Psychopathology, 6, 121-143. McGee, R., Wolfe, D., Yuen, S., Wilson, S., & Carnochan, J. (1995). The measurement of maltreatment: A comparison of approaches. Child Abuse & Neglect, 19, 233-249. National Center on Child Abuse and Neglect. (1987). Proceedings of the neglect grantees meeting. Washington, DC: Author. National Council of Juvenile and Family Court Judges. (1992). Protocol for making reasonable efforts to preserve families in drug-related dependency cases. Reno, NV: Author.

Susan Zuravin is a full professor at the University of Maryland School of Social Work (UMSSW). She specializes in research methodology, which she has taught in the doctoral and master’s program for 10 years. Prior to coming to the UMSSW she worked at the Baltimore City Department of Social Services (BCDSS) for 22 years. Currently, she is conducting research on the relationship between maternal depression and child maltreatment. Diane DePanfilis, Ph.D., an associate professor at the University of Maryland School of Social Work and co-director of the University of Maryland Center for Families, has 30 years of experience in the child maltreatment field as a practitioner, administrator, and researcher. Research supported by federal, state and local grants support varied explorations about interventions designed to enhance the safety and well-being of children and families. She is a past president of the National Board of Directors of the American Professional Society on the Abuse of Children. Laura Ting, L.C.S.W., is a doctoral student at the University of Maryland School of Social Work. She is a research assistant with the Family Connections project and has been a Council on Social Work Education minority fellow. Howard Dubowitz, M.D., M.S., is a professor of pediatrics and co-director of the Center for Families at the University of Maryland, Baltimore. He is chair of the Child Maltreatment Committee of the American Academy of Pediatrics, Maryland Chapter, and is on the State Council of Child Abuse and Neglect; he served two terms on the APSAC Board. He is a clinician, researcher, and educator, and he is active in the policy arena at the state and national levels, representing APSAC on the National Coalition on Child Abuse and Neglect.


Suggest Documents