Analytics and Data Science

Translating Data Into Action

The DCRI pushes beyond what’s most apparent. By tapping the experience of Duke faculty statisticians and the real-world insights of practicing physicians, we find the question that most effectively probes the challenge in your research and the fullest potential in data it generates. The result: meaningful decisions reached more quickly and efficiently, evidence that has more immediate impact on patient care and datasets with value beyond the immediate analyses.

Using examples from completed and ongoing studies, Michael Pencina, PhD, DCRI faculty associate director, describes DCRI’s approach to data science – collaboration between faculty clinicians and statisticians and novel methodologies employed to uncover new insights to improve health.

group photo

Where Do We Go From Here?

Utilizing rigorous, best-in-class tools and mathematical algorithms, our renowned biostatisticians and data scientists can tailor methods to the question. We find patterns within datasets that lead to stronger and more actionable insights. This approach of framing the right question to provide context and focus and applying optimal methods marks our work in:

The Knowledge to Predict

The Chinese philosopher Lao Tzu is said to have observed that those with knowledge don’t predict and those who predict lack knowledge. DCRI’s Center for Predictive Medicine is turning that ancient wisdom on its head using pools of data that have never been wider or deeper.

And we’re experts at linking unlimited patient-generated health data from electronic health records (EHR), clinical trial data, claims data, and other shared data to newly available giant data sets to answer what’s next.

Can Randomized Pragmatic Trials Fulfill Their Promise?

Michael Pencina, PhD and Eric Peterson, MD, and co-authors from McKinsey & Company, describe the steps that researchers, the private sector and regulators can take to design and implement clinical trials that are better, faster, and cheaper using real world evidence.

Illustrating Informed Presence Bias in EHRs

EHR data have a number of appealing characteristics, but they also pose a number of analytic challenges. How a patient interacts with a health system can influence which data are recorded in the EHR. Benjamin Goldstein, PhD, and his colleagues explore the implications.

Three Risks to Avoid in the Rush for Health Care Big Data

How can we ensure that big data in health care will be the rich resource we imagine? In this commentary for STAT, Michael Pencina, PhD, likens the clamor to collect more and larger datasets to the 19th century Gold Rush and describes three risks that data scientists and data consumers must avoid.