Evaluation of Belonging and Emotional Security Tool (BEST): A Confirmatory Factor Analysis (CFA) Approach

Background and Purpose: The BEST instrument was developed for use in the foster care permanency planning process, in which foster parents explore their feelings about foster youth, their level of commitment to them, and their expectations regarding the future. Due to limited standardized measures of belonging and emotional security, key outcomes for adoptive/guardianship families, the BEST was adapted for use with this population. This paper evaluates the properties of the adapted version of the BEST through confirmatory factor analyses (CFA).        

Methods: Data was collected on the BEST in five QIC-AG sites, with economically and racially diverse populations (n = 1,684). Our modified BEST survey questionnaire version consisted of twenty 5-point Likert scale items, intended to measure two domains: emotional security (13 items) and sense of belonging (7 items). To evaluate the structure of the scale, we conducted CFA with two models: 1) a single factor model and 2) a two-factor model with items loaded on those two domains of questionnaire items for the BEST. Because data was collected from multiple sites, we also conducted multiple group confirmatory analyses to examine the appropriateness of CFA models across sites. All models were estimated with the weighted least square mean and variance (WLSMV) estimator using Mplus software. Three model fit statistics were used to evaluate goodness of fit: Chi-square Test, Root Mean Square Error of Approximation (RMSEA), and Tucker Lewis Index (TLI).

Results: The single factor model showed that all 20 item factor loadings were statistically significant, with p<.001. The standardized factor loadings of 18 indicators were all above .5 which was set as the lowest cutoff. The two items with factor loadings below the cutoff were “How often do you include your child in family vacations?” and “How often do you stand behind your child even when he/she is wrong?”  Model fit statistics showed a moderately good fit for this first model (RMSEA=.08; TLI=.95). The second model with two factors also showed item factor loadings that were statistically significant, with p<.001. Those two items identified as indicators with low factor loading values in the first model were also low in the second model. The model fit statistics for this two-factor model were slightly improved, but similar to the single factor model. Multiple group analysis found a similar factor structure across sites.

Conclusions and Implications: CFA found moderately good fit with the data collected on the adapted BEST, but also suggested some items that, while informative for foster families, may not be as relevant for adoptive/guardianship families. Other items may be used to indicate risk for low belonging and emotional security in adoptive/guardianship families. Based on these initial results, we will discuss quantitative strategies to improve the BEST as a tool. Results relating to differences across racial groups and permanency types will also be discussed.