Isn’t This Fun? Structured Learning for Data Science and Data Management Services

August 12th, 2024

Introduction

Research data services is a rapidly growing area of academic and research librarianship. In addition to research data management (RDM), which focuses on policies and procedures for managing data, libraries also offer high-demand services supporting data analysis and exploration. In health sciences libraries, research data services is becoming an essential offering, yet there are scant opportunities for professionals to learn and practice these skills. The demand for research data services is increasing due to funder, publisher, and institutional policies and mandates. As libraries grow their staff and services, people are looking to gain knowledge and experience in this area. Accordingly, medical librarians need opportunities to learn the skills required to navigate a changing data ecosystem centered on data science, findability and interoperability, policy compliance, sharing, and transparency. There are now two free courses offered annually which teach core skills for providing research data services in health sciences contexts: Fundamentals of Health Sciences Research Data Management (FunRDM) and Fundamentals of Health Data Science (FunDS). We developed these courses in our capacity as part of the team of data services librarians comprising the NNLM National Center for Data Services (NCDS), which is based at NYU Langone Health. Established in July 2021, NCDS develops courses, workshops, and webinars in RDM and data science to help address the educational needs of librarians working in data services.

Course Development and Overview

Our team of data services librarians developed the two 8-week courses, focusing on two domains of research data services: (1) RDM and curation, including preparing and describing data for reuse, safe and effective data storage, and data preservation; and (2) data literacy and data science, including understanding how to ethically and effectively make use of research data. As such, these two courses provide entry-level skill building in each area. We determined which topics to cover based on our team’s experiences and expertise from working in these areas and created or reused materials in alignment with the learning objectives for each class. Since the target audience is health sciences librarians who are typically working in full-time academic positions, we needed to keep materials short so as not to over-burden them with long time commitments, so we aimed for three hours of content per week. We utilized the Quality Matters approach, creating alignment tables that showed the relevance of learning materials and assignments to the module learning objectives. Module learning objectives built upon each other to accomplish demonstrable course objectives. Another aim was to develop a learning environment that translates to on-the-job expertise, which meant creating assignments that emulate common professional tasks. Discussion prompts and innovative assignments provided learners with the opportunity to work through typical data services requests including reviewing an existing data management plan, critiquing a clinical survey instrument’s approach to data collection, and writing a script to clean a real-life messy dataset.

Course Components

FunRDM is focused on data management, storage, and description. FunDS teaches algorithmic thinking, introductory-level coding, data cleaning, exploratory analysis, and data visualization. Each 8-week course is available by application only, consists of 6 modules over 7 weeks, and concludes with a project during the final week. Both follow an asynchronous cohort model, where a group of 60 learners go through weekly materials on their own but post responses to group discussion boards. Learners also attend a handful of required meetings with community experts along with a final presentation session. Throughout, learners are assisted by course facilitators as well as “community experts”, established data services librarians who advise learners on their final projects.

Fundamentals of Health Sciences Research Data Management

FunRDM’s main focus is encouraging learners to connect with their local communities; for instance, one assignment asks them to interview a researcher about their data needs. This class covers topics ranging from the research data lifecycle through reproducibility, open science, data documentation, data sharing, and more.  Community experts guide learners as they engage in conducting interviews with researchers and answer questions pertaining to their final presentation on a research data management topic of their choice. Upon completing the course, learners will be able to:

• Explain different types and levels of research data management services
• Describe tools, resources, and workflows to facilitate reproducibility and replicability
• Advise researchers in best practices for data management planning
• Explain how to effectively document and describe research data

Advise how to safely store, protect, and share medical research data

Fundamentals of Health Data Science

FunDS focuses on data structures, data cleaning, and data visualization, with Python as the coding language. It includes a general look at data science and algorithmic concepts as well as specific topics in coding. Learners practice with software tools to develop the understanding needed to clean data, create data visualizations, and share reproducible results. For the final project, learners demonstrate their proficiency by analyzing variables from the National Cancer Institute’s Health Information National Trends Survey, a real-world public health dataset. Since people begin with a variety of backgrounds and familiarity with programming, a pre-module on understanding what a programming language is that provides walkthroughs for how to create a Jupyter notebook and run Python on Google Colaboratory. Upon completing the course, learners will be able to:

• Employ analytical thinking to solve data science problems with step-by-step procedures
• Prepare data using ethical practices to maintain research integrity and avoid bias
• Use Python programming techniques to clean and analyze datasets
• Communicate results to stakeholders using best practices for visualizing and reporting data

Reception

As of the writing of this editorial, FunDS has been offered once, and FunRDM has been offered twice. We accepted a larger number of participants for FunDS in anticipation of higher drop-off rates. Participant data can be found in Table 1.

Course	Applications Recevied	Applicants Selected	Active participants	Completed the Final Project
FunRDM 2022	105	60	52	42
FunDS 2023	146	75	54	36
FunRDM 2024	89	60	45	32

For both courses, learners showed a high level of engagement, frequently interacting within the discussion boards and in Slack. Evaluation included surveys with written feedback from learners, facilitators, and community experts. Learner feedback has been positive, with some past learners reaching out to let us know they are using materials from the course in their work. For the first offering of both courses, the majority of participant responses reported that the course was “engaging” (FunDS: 96%, FunRDM: 97%) and “informative” (FunDS: 100%, FunRDM: 97%). For both courses, respondents stated they will be better able to support end users’ needs (FunDS: 95%, FunRDM: 100%). Data collection for the second offering of FunRDM is ongoing.

One FunRDM survey respondent said, “It’s the first data management training… that was actually practical and gave real-life examples of how to implement data services in realistic ways.” Another stated, “I am getting ready to be the data management lead for my library and the readings, the very relevant homework assignments, the discussions, and prepping for the final project presentation all were very beneficial to me.” A respondent for FunDS stated, “No other data science related course that I’ve taken has lowered the barriers to understanding how to work with data using a computer language,” and they “have a better understanding of what is required of [my patrons] and what information services they might need that my library can provide. This is transformational for me and possibly my library.”

We also conducted a long-term survey of FunRDM participants from 2022 (see Figure 1). A majority of respondents reported that the course increased their confidence in providing RDM services (87.5%). They were better able to assist with data management needs (75%), create and enhance reference materials (75%) and classes (75%), and promote resources (75%). However, over 81% of respondents did not find the course helped them pursue job opportunities. This may be due to selection of participants already in data positions or tasked by their institutions to provide research data services. Future offerings of this class with a cohort of librarians not yet in data-related roles may show a different outcome.

We also received encouraging emails from participants. One stated that the course was “challenging” but “very valuable” as they are enrolled in train-the-trainer for research on the All of Us Researcher Workbench.  According to this individual, the instructing institution “demonstrated how Python is used in the workbench and demonstrated some of the accessible code snippets for researchers to utilize.  Because of your course, I felt very well prepared and confident in understanding how Python is essential to access and use this data for biomedical research. …There are currently about 8,000 researchers doing about 8,000 research projects with the All of Us data.  This signifies to me the growth in opportunity for librarians to learn new skills like Python in support of researchers at their institutions.”

Feedback will be utilized to improve future offerings of the classes.