Assessing the potential for child maltreatment: Measuring the reliability and validity of the AAPI

1
Assessing the potential for child maltreatment
Measuring the reliability and validity of the AAPI

• FCAP Conference
• Fordham University
• June 26-27, 2008
• Nina Esaki, MBA, MSW, PhD
• University at Albany School of Social Welfare
• Funding provided by The Foundling/Fontana Center
  and
• U at Albany Initiatives for Women

2
Overview

• Dissertation study investigating
  intergenerational transmission of child
  maltreatment
• Is parenting attitude a mediator between
  cumulative maternal childhood abuse and frequency
  of perpetration of child maltreatment in
  adulthood?
• Longitudinal Studies of Child Abuse and Neglect
  (LONGSCAN) database and Adult-Adolescent
  Parenting Inventory (AAPI Bavolek, 1994)
• Reliability and factor analysis conducted as part
  of larger study

3
AAPI

• Purpose of current instrument to assess parenting
  and child-rearing attitudes of adolescents
  (12-19) and adults (20)
• Common instrument in child welfare research
• Instrument consists of 32 items grouped into four
  scales Appropriate parental expectations of the
  child, Appropriate empathy towards childrens
  needs, Parental value of non-corporal discipline,
  and Appropriate parent-child roles
• Responses scored on a 5-point Likert scale that
  ranges from Strongly Agree to Strongly Disagree
• Higher scores reflect more appropriate attitudes

4
LONGSCAN AAPI

• LONGSCAN version includes 38 items
• Study researchers added 6 additional items to
  Appropriate Expectations subscale to improve
  reliability and validity
• Sample of 477 mothers at risk of child
  maltreatment from 3 sites in LONGSCAN study
• Data collected at child age 4

5
Reliability analysis

• For full 38 item AAPI instrument, a coefficient
  was .94
• Alpha coefficients for all 4 subscales exceeded
  .70 recommended by Nunnally (1978)

6
Details of reliability analysis
Means, Standard Deviations, Intercorrelations,
and Coefficient Alpha Reliability Estimates for
the AAPI
Note. N477. Reliability estimates appear on the
diagonal.
7
Prior factor analysis

• Relatively recent study by Lutenbacher (2001)
  testing psychometric properties of AAPI in sample
  of low income single mothers
• Demonstrated poor internal consistencies for 2
  subscales Appropriate Expectations and Values
  Non-corporal discipline
• Additionally, instruments 4 factor structure was
  not supported
• 19 item modified version of AAPI with 3
  dimensions identified
• Other than research by developer, limited
  psychometric data available (Conners et al.,
  2006 Lutenbacher, 2001)

8
Approach

• Followed factor analysis approach used by Conners
  et al. (2006) in their assessment of AAPI
  following paradigm by Anderson and Gerbling
  (1988)
• First, confirmatory factor analysis (CFA) was
  conducted to assess uni-dimensionality of
  subscales
• Structural equation modeling (SEM) used for all
  CFA modeling

9
CFA results
10
Exploratory factor analysis (EFA)

• Due to failure of CFA to support factor structure
  suggested by developer, exploratory procedures
  performed using SPSS
• Examined all 38 items
• Since earlier Connors et al. (2006) study showed
  little intercorrelation between factors, items
  examined using Principal Components Analysis
  (PCA) with Varimax rotation
• PCA resulted in 7 factors with eigenvalues equal
  to or greater than 1 and which explained 54.2 of
  variance

11
PCA results
12
Conclusion I

• Results of CFA supported uni-dimensionality of
  one AAPI subscale Appropriate Expectations
• EFA provided some support for factor structure
  proposed by AAPI developer
• Strongest support for Appropriate Roles
• Some support for Expectations and Empathy
• Alternative factor structure would have fewer
  factors
• Importance of psychometric testing on common
  instruments used in field

13
Conclusion II

• Lack of predictive ability of three AAPI
  subscales and post-4 CPS report
• Association ? causality
• Precision of outcome variable
• Cross-cultural issues with instrument
• Need additional research to further explore
  issues