Research That Matters (January 17 - 20, 2008)
|Sunday, January 20, 2008: 8:45 AM-10:15 AM|
|Empire Ballroom (Omni Shoreham)|
|[CW] Modeling Risk for Maternal Physical Abuse and Neglect Using the Fragile Families and Child Wellbeing Study|
|Symposium Organizer:||Shawna J. Lee, PhD, Wayne State University|
|Discussant (Optional):||Diane DePanfilis, PhD, University of Maryland at Baltimore|
|How Neighborhoods Influence Risk for Physical Child Maltreatment: Path Analyses from a National Study|
Neil B. Guterman, PhD
|Race and Ethnic Differences in the Pathways Linking Neighborhood Characteristics, Parenting Characteristics, and Risk for Maternal Physical Child Abuse|
Shawna Lee, PhD
|Structural Model Analysis Linking Intimate Partner Violence and Risk for Maternal Child Abuse and Neglect|
Catherine A. Taylor, PhD
Proposed are three interrelated studies that use structural equation modeling (SEM) to examine how social environmental factors, particularly community characteristics and presence of intimate partner violence (IPV), shape parenting behaviors and risk for physical child abuse and neglect (CAN). Past research links community variables such as poverty to higher rates of IPV and CAN, with evidence that parental support and other relationship qualities may mediate the influences of the surrounding community. Although community resources are believed to facilitate positive parenting behaviors, few studies have examined direct and mediating relationships linking perceptions of community social disorganization and collective efficacy to parenting variables such as parenting stress and mastery, and how those factors may relate to CAN risk. Additionally, despite the high rate of co-occurrence of IPV and CAN, the specific pathways that link these types of family violence are unknown.
Data is drawn from the Fragile Families and Child Wellbeing Study (FFCWS). FFCWS is a national birth cohort of 3,356 families across 20 large U. S. cities (populations > 200,000). FFCWS is uniquely positioned to investigate a range of factors related to CAN risk because it includes data assessing both self-reported parenting behaviors and perceptions of community social disorganization and collective efficacy. Furthermore, the FFCWS is a racially and ethnically diverse national population-based sample as opposed to a problem-based sample.
Analyses used data from in-home and phone interviews that were conducted when index children were 3 years old. Child maltreatment risk is assessed using proxy variables based on mothers' self-reports on the Conflict Tactics Scales Parent-to-Child (CTS-PC) physical aggression, psychological aggression, and neglect subscales. Observations of the mother-child relationship are assessed based on observed HOME scale outcome measures.
Paper #1 examines how self-reported measures of social disorganization and collective efficacy influence mothers' mastery and parenting stress, and how those variables directly and indirectly relate to CAN. Paper #2 examines maternal race/ ethnicity as a moderator of these pathways. Paper #3 examines potential mediating pathways that link mothers' reports of experiencing IPV, including both physical and psychological aggression, and self-reports of maternal CAN
We use SEM for a number of reasons. First, SEM procedures reduce the impact of measurement error that can occur with imperfectly measured characteristics such as parenting stress and mastery. SEM produces a macro-view of complex relationships because it is possible to enter a series of variables into the model to allow for concurrent tests of multiple hypotheses related to a problem behavior. Finally, SEM procedures are ideal for conducting subgroup analyses with a large data set to examine potential moderating effects.
All models were tested by a confirmatory latent-variable structural analysis using Mplus 4.0. Given the large sample size, we randomly divided the sample into developmental and confirmatory samples to provide a cross-check of the final models. Goodness of fit and misfit measures (TLI, CFI, SRMR, and RMSEA) are reported as well as the Satorra-Bentler chi-square test, which was used to account for non-normal distribution of the outcome variables. Standardized path coefficients are presented for individual pathways.