The development of data science curricula has gained attention in academia and industry. Yet, less is known about the pedagogical practices and tools employed in data science education. Through a systematic literature review, we summarize prior pedagogical practices and tools used in data science initiatives at the higher education level. Following the Technological Pedagogical Content Knowledge (TPACK) framework, we aim to characterize the technological and pedagogical knowledge quality of reviewed studies, as we find the content presented to be diverse and incomparable. TPACK is a universally established method for teaching considering information and communication technology. Yet it is seldom used for the analysis of data science pedagogy. To make this framework more structured, we list the tools employed in each reviewed study to summarize technological knowledge quality. We further examine whether each study follows the needs of the Cognitive Apprenticeship theory to summarize the pedagogical knowledge quality in each reviewed study. Of the 23 reviewed studies, 14 met the needs of Cognitive Apprenticeship theory and include hands-on experiences, promote students' active learning, seeking guidance from the instructor as a coach, introduce students to the real-world industry demands of data and data scientists, and provide meaningful learning resources and feedback across various stages of their data science initiatives. While each study presents at least one tool to teach data science, we found the assessment of the technological knowledge of data science initiatives to be difficult. This is because the studies fall short of explaining how students come to learn the operation of tools and become proficient in using them throughout a course or program. Our review aims to highlight implications for practices and tools used in data science pedagogy for future research.