Trends in the Utility of Structural Equation Modeling in Social Work: A Systematic Review

Background: Structural equation modeling (SEM) is a statistical technique that allows the testing of various models to depict a relationship between observed and latent variables Increasingly, social work scholars have been engaging in testing theories and finding relationships among observed and latent variables. Hence, this makes SEM an important statistical technique for social work scholars and researchers. To date, there only exist two comprehensive reviews detailing the extent to which SEM is being utilized in social work research (Guo et. al.,2009; Okech et. al., 2013). Both these reviews only reviewed n=32 and n=59 articles respectively and listed several limitations associated with the reviewed studies. The purpose of the present systematic review is to assess further developments over the past half-decade and to document the evolving use of SEM by social work researchers.  

Methods: To perform our systematic review and select articles, we utilized the PRISMA method. We began by referring to the list of journals indexed within ‘Web of Knowledge.’ The keywords for the search were, “SEM” OR "structural equation modeling" OR “CFA” OR "confirmatory factor analysis" OR “EFA” OR "exploratory factor analysis" OR "path analysis" OR "latent growth curve" OR "latent class analysis." We focused on articles that were published between March 2012 to December 2017 because this is a follow-up article to the previous review. The initial search yielded 525 articles. After a careful review using the established criteria, a total of n=466 articles were included in the review.

Results: From March 2012 to December 2017, 466 studies using SEM methods were published across 44 social work journals. This period witnessed a gradual increase in the number of studies that applied SEM across social work journals. Although we retrieved our samples from social work journals, only 188 studies (40.3%) reported a social work faculty member as the first or second author. About 30% of the studies used full structural equation modeling, 26.2 % used CFA, 17.2% used path analysis, and more interestingly, 5.6 % used latent growth curve modeling. Others (21.5%) applied more advanced SEM techniques such as latent class analysis or multi-level growth curve models. Two journals - Children & Youth Services Review (20.8%) and Child Abuse & Neglect (16.3%) published the most SEM articles.

Implications: The findings of our review indicate that between March 2012 to December 2017, there has been a substantial increase in social work studies that utilized SEM. SEM is becoming one of the most useful statistical modeling tools for social work researchers. Even so, our findings show quite a few studies did not fully explain their results- model fit, multivariate normality, handling of missing data, and factor loading/coefficients. Therefore, social work researchers need to pay more attention to going through the steps and checking the issues of the sample size, method of parameter estimation, and assessment of model fit. Going forward it is imperative for social work students to be provided more training and free courses on using SEM techniques in their research.