Python Tools for Stem Cell Transplantation

Abstract

Background Transplant programs, unlike other clinical programs, have a need for coordination of care before, after and during the transplant process. Milestones, outcomes, and metrics need to be tracked for patient care, quality improvement, regulatory and financial purposes. Transplant specific software platforms can provide the functionality needed for a transplant program however they tend to require considerable financial commitments that is often beyond the reach of small transplant programs. Python is a very popular programming language that is easy to use and highly extensible. Open source tools such as Jupyter Notebook, Python data analysis library (PANDAS) along with visualization tools such Matplotlib as well as seaborn provide easy to use tools for automating day to day tasks and generate reports without a lot of technical know-how or financial commitments. Methods Transplant-related data in our program is maintained on a Microsoft Excel spreadsheet that is shared over a secured network drive. As our transplant volumes increased it became increasingly difficult to use and the ability to coordinate care became more and more challenging with the increasing number and complexity of the transplants. To overcome these difficulties we turned to the python data analysis library (PANDAS). This library is ideal for data manipulation and analysis. Reports generated using the PANDAS library are sent out in an automated fashion to the members of the transplant team via email through custom python scripts. Quality data such as survival metrics were calculated using the python survival analysis package Lifelines and this along with other metrics such as engraftment, infections, referral patterns etc were visualized using visualization tools such as Matplotlib and seaborn. All of these tools were designed to be run in an automated fashion without much need for coding or maintenance. We have also been able to use python tools to access external API's such as Google Maps, Geocoding and Distance Matrix API's to better understand the geographical location and needs of our patients. Results Python tools and packages has improved our efficiency and has helped us provide high quality coordinated care. Quality improvement and reporting requirements have been greatly simplified and team members can get survival and other quality metrics in an automated fashion with little to no effort. Discussion The python tools that we have developed has helped us in different aspects of the program. We intend to continue to build on this using open source tools that can be customized to the needs of our program. We plan to eventually transition our data to a modern database that can be accessed via a web framework. Python with its Django framework would be an ideal open source platform for such an endeavor and could be used with the AGNIS system to exchange data with CIBMTR.