The opportunity recently to present a talk at the Inaugural Conference of the Wisconsin Institute for Healthcare Systems Engineering gave me a chance to think about the intersection of data science and health systems engineering.
Health systems engineering (HSE) is that branch of industrial engineering concerned with improving the systems that surround care—making clinics run effectively, applying concepts from supply chain to the challenges of delivering chemotherapy, establishing best practice standards for surgical procedures—all with a goal of ensuring safe, effective care.
Health systems engineers (HSEs) draw from many research traditions to guide their work, including economics and finance, mathematics, operations research, human factors, manufacturing processes, and systems theory. I think they have much to offer those for whom discovery comes from data, which are voluminous, unlikely to follow normal distributions, and sometimes just plain messy.
Like other branches of industrial engineering, the work of health systems engineers is to analyze complex systems and look for strategies to improve processes or outputs. Most health systems engineers work in clinical systems and their work has led to improved patient throughput, reduced infection rates, and better utilization of workers.
Now, some health systems engineers already think they do data science—as proof, they point to those enormous databases, like the Medicare claims data, used to characterize the complex financial structure of health care.
Health systems engineers employ a wide variety of analytical approaches to systems analysis and improvement, from classic regression to economic modeling to decision sciences. Far from being methodological purists, health systems engineers display skill in matching characteristics of the analytical method to features of the available data. HSE relies on capturing data collected in the work process, so they have intimate knowledge at the closest point of the intersection of health care phenomena and the data used to represent it. HSEs could help enhance the quality of data at the point of capture. And, because of their intimate knowledge of health care delivery and mathematical modeling, HSEs could be quite helpful in creating best practices for data capture and storage.
HSEs often are well-versed in industry regulations. This knowledge could be of great value for data science pipelines that originate in data collected in the clinical environment, stored in the electronic health record, and subsequently used in large scale data science investigations.
There’s another way that data science could benefit from better engagement with HSEs. HSEs bring a methodological portfolio that includes econometric modeling as well as methods from optimization and stochastic simulation that are better suited to data science investigations than are statistical approaches. Statistical approaches rely on well-behaved data sets that are collected with a degree of precision and experimental control and are representative of a randomization process. Much of the data likely to be of use in data science approaches lacks the robustness required for the application of well-known statistical approaches and would be better suited to the analytical strategies from other traditions.
So, all you data scientists and modelers—get out there! Find some health systems engineering partners. You may discover that, as in other industries, health systems engineers hold the keys to making good, better!