Software plays an important role in scientific research. Often, scientists write their own software without the help of software engineers. This results in questionable software quality, low software development productivity, lack of reproducibility of results, and other issues that impact the quality of the scientific research. Our research aims to improve the way that scientists develop software. Current research studies are using surveys and interviews to determine how scientists create software and make decisions about their software development approaches. The long-term goal is to develop a Scientific Software Process Improvement Framework (SciSPIF), which is a decision support system that provides proven software engineering practices to address the specific concerns of scientific software development projects.