School-Based Health Centers and School Connectedness

Abstract

Background: Improvements in health behaviors and academic outcomes have been associated with school-based health centers (SBHCs). However, underlying mechanisms for these associations have been largely unexamined, particularly among lower-income youth. The current study examines the relationship between SBHCs and school connectedness and whether this relationship differs by youths' socioeconomic status (SES).

Methods: Student-level cross-sectional data from 503 traditional high schools in California were analyzed using multilevel regression models. California Healthy Kids Survey 2013-2014 data included information on 3 dimensions of school connectedness and demographic characteristics including SES as measured by parental education. School-level demographic data was gathered from publicly available sources.

Results: Although no significant relationship between SBHCs and any of the school connectedness dimensions emerged, there were significant cross-level interactions between SBHCs and parent education. SBHCs were more positively associated with school connectedness (adult caring, adult expectations, and meaningful participation) among lower SES students compared to students with higher SES.

Conclusions: SBHCs may be particularly effective in affecting school connectedness among lower income youth populations. This has wide ranging implications with regards to planning (eg, careful selection of where SBHCs can be most effective), and future research (eg, examining the effectiveness of specific SBHC strategies that support connectedness).

Keywords: parental education; school connectedness; school-based health centers; socioeconomic status.

Figure 1. Predicted Standardized School Connectedness Scores for Students With and Without School-based Health Centers by Parent’s Highest Education Level (N=300,000): California Healthy Kids Survey, California, 2013–2014

Note. Predicted scores were estimated from multilevel linear regression models and averaged across level 1 and level 2 variables. P-values were calculated using Stata’s post estimation lincom command, which estimated p-values from linear combinations of coefficients based on the final models with the cross-level interactions. ^† p < .10; * p< .05; ** p < .01