Computational Thinking (CT) is regarded as a crucial competency for all children in the 21st century. There are misconceptions about CT which confounds CT skills with programming skills. However, CT skills and programming skills are not the same. CT is a broader skillset of cognitive thinking that is integral to complex problem solving. Teaching and assessing CT should not only focus on computer programming. In this project, a new design-based learning (DBL) approach was proposed to improve elementary school students' CT. This approach not only included programming skills, but also integrated CT practices with authentic real-life contexts. Corresponding to recent conceptions of CT, we also assessed CT in a more comprehensive manner. A quasi-experiment study was carried out to assess how the intervention was associated with students' self-perceived CT skills using both quantitative and qualitative methods. In the treatment group, twenty-three fourth graders engaged in three curriculum units to create artifacts that solve specific real-world problems using LabPlus electronic kits and Scratch programming software. The control group took a traditional computer science class. Results revealed that students' self-perceived CT skills increased to a greater extent in the treatment group compared to the control group. In addition, we observed and interviewed three student cases in the treatment group to understand the potential learning effects of our new DBL approach. This study contributes to research in CT instruction, its application in STEM education in particular, and how it can be used in elementary level education in general.