“Most people spend more time and energy going around problems than trying to solve them.” — Henry Ford
How do we apply process improvement and data science to pinpoint root causes?
Previously, we discussed one of the fundamental tools in prescriptive analytics: simulation. We looked at the inpatient pharmacy as an example and demonstrated the massive amount of value that prescriptive analytics can provide towards mitigating drug waste and reducing cost. However, we made a point to bring up how much work must take place beforehand to ensure such an endeavor is successful. This included three key points:
- A thorough understanding of the underlying business problems
- A solid descriptive profile of the population in which one wishes to drive change
- Being able to robustly link dispense events to administration events.
Another massive undertaking in the pipeline is to be able to identify the root cause of each waste event. Let’s dig into this step more closely – because while it requires the disciplines of both process improvement and data science to be done effectively, it should not be understated how much value can be added by this step alone.
Five Whys to Find the True Cause
Root cause analysis is undoubtedly a multi-disciplinary problem, as there are many tools and a variety of disciplines aimed at solving it. The Lean Six Sigma methodology for process improvement lays out a series of tools for attacking a root cause analysis problem. A classic method is the ‘five whys’ – where generally, asking up to five “why” questions allows someone to drill down to a root cause. Using this approach, descriptive analysis can be wisely informed. For example, we may find that a large number of waste events are associated with low cost drugs. Why? It could be these drugs have different workflow. Why does the workflow cause these drugs to be more susceptible to waste than medium or high cost drugs? Because there are such a larger number of them, and it is enormously costly to sort them. This method is incredibly simple, yet very powerful for identifying ways to target meaningful change.
Statistically Identify the Highest Priority Causes
Once the key features and events which drive waste events are identified, enter data science. Paradoxically, the application of data science to these problems can sometimes be more of a subjective “art” than a strict science. A key component of a multivariate system like an inpatient pharmacy, is the controlling for and explaining of variation in the system. For example, the events that drive waste for PO drugs are highly unlikely to be the same as the events that drive waste for IV drugs, simply due to the difference in workflow for the two drug types. Another critical requirement is the ability to rank multiple causes that influence one single waste event. Several machine learning algorithms are equipped to describe “variable importance”. Through some clever application, each waste event can be assigned a cause. After this, the leading causes of waste in the entire facility can be reported and continuously monitored by pharmacy leadership.
To the outsider it can be easy to think of process improvement and data science as very disparate fields. However, nothing could be further from the truth. Process improvement informs meaningful data analysis, providing a well-defined description of the population of study. Once this is understood, process improvement specialists, data analysts/scientists, and pharmacists can brainstorm what constitutes intrinsic features of the system, what constitutes unpreventable causes, and what constitutes preventable causes. Data science comes in to quantify and rank the impact of preventable causes and ultimately define which areas would benefit most from implementing change. Indeed, on their own, process improvement and data science are powerful tools. Together, they make an unbelievably powerful and robust force capable of improving hospital operations and drastically mitigating unnecessary costs.
About the Author
Richard Bryant is a data scientist with experience using data analysis, predictive modeling, and discrete event simulations to drive innovative technology solutions. Utilizing an M.S. in Applied Statistics from the University of Michigan and a Lean Six Sigma Green Belt certification, Richard has provided analytical support to healthcare process improvement projects in various hospital departments, such as radiology, obstetrics, and pharmacy. Prior to his role at Improvement Path Systems, he provided statistical consulting services to the Michigan Surgical Quality Collaborative and to the educational technology provider Echo360.