Measuring Maternal-Fetal Attachment: Model Fit and Measurement Invariance of a New Assessment Tool

The importance of studying Maternal-Fetal Attachment (MFA) as part of the system of maternal-child perinatal functioning is well established in the literature, as MFA relates to health-related and psychosocial variables both during pregnancy and after the infant is born (e.g., positive health behaviors in pregnancy; levels of maternal stress, anxiety, and depressive symptoms; positive parenting behaviors after the child is born; and secure infant attachment at age one year). Despite the apparent value of studying MFA, there is no acceptable measurement tool for the construct. Existing assessment tools often do not meet minimum psychometric standards, may not be practical for broad research application, and have demonstrated inconsistent results in correlational research. Therefore, a clear need exists for a measurement scale with strong psychometric properties that is based on empirically supported development strategies. The present line of research sought to fill this need. The process of developing a new measurement tool began in a previous study, Hedrick (2015), which used Exploratory Factor Analyses to identify options for a potential measurement scale from subsets of a large pool of items intended to measure MFA. In Hedrick (2015), three options were identified as acceptable, then compared based on ranges of communalities, percentage of variance explained, internal consistency, and performance on two measures of concurrent and discriminant validity. However, these assessments failed to differentiate a single option as most preferable. Therefore, the first aim of present study was to complete the task of differentiating among the working measures to choose a measurement scale. Confirmatory Factor Analysis (CFA) was used to attempt to determine which working measure from the EFA study had the factor structure that fit best in a second sample of participants. The CFA process was not able to differentiate a superior working measure from the options. Therefore, the most theoretically sound of the three was chosen to use as a measurement tool, with some small adjustments made based on the CFA results. To continue to establish the validity of the new questionnaire measure, the second aim of the study was to assess its measurement invariance among different administration methods and populations. The measurement tool was found to be non-invariant in both areas. Specifically, the results of the measurement invariance analyses indicated that the measure should be administered on paper rather than online, would be best applied in research in African American/Black populations than in research with other racial/ethnic groups, and should not be used to compare MFA scores among racial/ethnic groups. Measurement invariance analyses also revealed that the scale was non-invariant at the metric level regarding parity groups. The implications of these findings for further research is discussed