Maya Petersen is associate professor, biostatistics and epidemiology in the school of public health at UC Berkeley. Her research focuses on the development and application of novel causal inference methods to problems in health, with a focus on the treatment and prevention of HIV. She has a strong interest in and has published on the interface between biostatistics, epidemiology and clinical medicine.
Modern big data resources detect variability in provider-level performance in near real time. Quantifying, visualizing, and reporting this variability can improve outcomes by identifying the best and worst performers and creating peer-based incentives for improvement. But the impact of simply describing existing variability is inherently limited without a deeper dive to understand its causes.
That’s where, as Stefano Bertozzi, dean of the UC Berkeley School of Public health, and his colleague Maya Patersen will explain, the emerging field of targeted learning analytics comes in.
In a hospital or other healthcare provider organization, targeted learning analytics can identify modifiable sources contributing to variability of performance and efficiency as well as suggest optimal courses of action that change over time as performance and other factors change. This enables a smarter, more data-driven healthcare system that learns, implements, and evaluates in rapid cycles, to continuously improve performance.
Many questions in healthcare are causal in nature – analytic tools should directly seek to answer those questions.
Current machine learning analytics can predict who is at risk, but do not tell what its causes are, nor what to do about it.
Targeted learning analytics fill this gap, and combine machine learning, causal inference and statistical inference.
Smarter healthcare systems can apply these tools to continuously learn from themselves and map the best course of action, assured of the soundness of their decision-making process.