Inside the Quantitative Sciences Unit

The Quantitative Sciences Unit (QSU) is a collaborative, interdisciplinary team of 30 data scientists that includes faculty members and PhD- and Master’s-level scientific staff. Its primary role is to collaborate with Stanford investigators on study design; data management and database creation; data analysis using state-of-the-art biostatistics and informatics tools; and software development. It also serves as a data coordinating center for randomized clinical trials. The group currently has about 100 projects underway in various stages.

Another component of QSU’s mission is the education, both formally and informally, of the next generation of data scientists and data-science-savvy investigators. For divisions and departments, QSU members often design and deliver short courses on data science. On a weekly basis, QSU holds two-hour meetings where in-house mentoring takes place in the form of jointly tackling complex data-related issues that arise in practice and discussing the underlying principles of team science.

A typical example of a QSU project is the NIH-sponsored Stanford GOALS trial, led by Thomas Robinson, MD, Professor of Pediatrics and of Medicine. This randomized clinical trial designed and evaluated a novel, multi-faceted intervention to reduce obesity in children. The study presented numerous data science challenges and yielded a rich data resource. One of the QSU’s largest and most technically challenging projects was the Apple Heart Study (AHS), led by Mintu Turakhia, MD; Marco Perez, MD; Manisha Desai, PhD; and Kenneth Mahaffey, MD.

Manisha Desai, director of the Quantitative Sciences Unit, (on the right) leads a group discussion on study design issues arising in one of their projects. Other data scientists shown, from left to right: Justin Lee, Natasha Purington, Vibhu Agarwal, and Vidhya Balasubramanian.

Photo:  Kris Newby

QSU’s key role in the Apple Heart Study

For 15 months, recorded heart-beats from 400,000-plus Apple Watch wearers streamed into the data center of the Apple Heart Study (AHS), a project initiated through a joint collaboration between Stanford and Apple. The project aimed to evaluate a new app on the Apple Watch designed to detect atrial fibrillation (aFib). This oft unnoticed condition significantly increases the risk of stroke and heart failure, and it is experienced by one-in-four individuals during their lifetime. If an irregular rhythm consistent with aFib was identified by the app, the participant received an alert and instructions for a physician follow up. The goal of the study was to evaluate how many participants would be notified of an aFib event and the performance of the app in detecting aFib among those notified.

At the core of this enormous real-world trial was a seven-person interdisciplinary team of QSU data scientists who became integral members of the larger team of experts from Apple and four School of Medicine entities — the Division of Cardiovascular Medicine, the Information Research & Technology group, the Center for Digital Health, and the Stanford Center for Clinical Research. From the beginning, QSU’s AHS squad advised the larger team on study design, data wrangling of diverse types of data, the methodologies for evaluating the Apple Watch algorithm, and the interpretation of findings. In addition, QSU members methodically tackled the unanticipated and challenging problems that arose with the processing of large volumes of data during the study.

“Apple relied on us to figure out how to manage and analyze the voluminous, disparate data streams flowing into the servers run by the School of Medicine’s Information Resources and Technology (IRT) group,” said Manisha Desai, PhD, Professor of Medicine and of Biomedical Data Science. “The exciting thing about this project was that it demonstrated how Stanford could partner with innovative Silicon Valley companies in assessing novel biomedical devices designed to improve the health of the general public. Given the diverse disciplines involved in making this project successful, this was an example of team science at its best.”

How to work with QSU

Investigators who want to collaborate with the QSU can start by filling out the simple project initiation form posted on the QSU website. These efforts are funded through an annual subscription fee, which encourages early, deep collaborations with partnering entities and a commitment to a team-science approach.

QSU is also involved in a pilot initiative, launched by the Office of the Dean of Research in the School of Medicine, where members can provide high-level data science advice to investigators across the School of Medicine. This engagement can be initiated through the “Request a Consultation” button on the Stanford Data Science Resources (DASHER) website.

Regardless of how researchers work with QSU, Desai advocates for team science, where practicing data scientists become integrated and invested members of a scientific team, an approach that she believes will lead to more discoveries and better outcomes. She is also dedicated to providing her data science staff and faculty members with clear career trajectories and the ability to conduct their own research on novel data solutions and methodologies that arise in practice.

“We’re really trying to change the culture at Stanford, where biomedical researchers understand that data science in practice is really just that — a science — and that we are partners in solving complex biomedical questions,” Desai said.