At HIMSS 2022, one of the big topics of conversation was how to drive efficiency in healthcare without compromising the patient experience. Healthcare is human-centric, and the literal life-or-death situations faced by clinicians and patients every day means there’s little opportunity to experiment with new processes and technologies that could drive efficiency. Any change needs to be simple to implement, work quickly and reliably, and produce results in order to be deemed a worthy investment.
To modernize within these limitations, many industries have turned to data solutions. However, this has proven difficult for healthcare systems. While organizations in other industries can dedicate resources to building out a data science team, hospitals are severely short-staffed in the IT department with resources focused on delivering top quality patient care. For healthcare executives the immediate demands of patient care leave little bandwidth or budget to test out data initiatives without a concrete outcome.
For quick results and actionable insights, machine learning (ML) provides a path forward for healthcare organizations. ML is the foundational step for artificial intelligence applications. It identifies patterns in data sets to predict future outcomes by building and training models. Once a model is developed, it can be used repeatedly to develop a comprehensive data set incorporating both historical and projected data. In this way, ML can help healthcare systems not only predict future results, but understand the drivers behind those results.
ML models are particularly useful for exploring “what-if” scenarios. Once a model is established, various aspects of a hospital’s patient population, supply chain, and other many other critical operations can be adjusted to predict outcomes for an entire patient population, a department, or even an individual. This kind of predictive experimentation can be used to develop plans for emergency situations, find workarounds for supply chain shortages, or test proposed policy changes without having to disrupt current operations or affect patient care.
To consider ML a successful investment, it needs to generate more than just knowledge. The insights ML models deliver are only valuable if they are actionable and easy to implement. Recent Gartner reports have warned that “85 percent of AI and machine learning projects will fail to deliver,” but a large percentage of those failures are not on the technological side.
Often, the failure is in the data strategy, commonly in the first step of the process: the quality of the data fed into the model. When data sets are small, homogeneous, or irrelevant for their intended purpose, the resultant predictions will be equally unhelpful.
Though this seems like an insurmountable barrier for hospitals and other healthcare organizations that do not have dedicated analytics departments, there are ways to engage predictive analytics without diverting resources. A robust solution, Qlik’s AutoML, addresses the issue of data quality by incorporating disparate data sources into an “ML-ready data set.” From there, the user can select the appropriate ML model based on target fields and the goals of the analysis. With the support of automated ML, healthcare organizations can use the data they already have to support repeated predictive analyses and “what-if” testing scenarios.
HIMSS anticipates more healthcare leaders will look to AI and ML to help triage demand, improve patient care, ease provider burden and reduce clinical variation. For their investment to succeed, healthcare teams must first identify their use cases and to make the best use of ML and predictive analytics. With a long view of both the technology and its applications hospital systems and other healthcare organizations will reap the benefits of these data-driven insights for both operations and patients for years to come.
