How trustworthy AI enhances workforce management and efficiency

Photo: Ariel Skelley/Getty Images

Today’s healthcare organizations are facing myriad challenges as they work to maintain adequate staffing to meet growing patient demand. According to the American Hospital Association, hospitals and health systems are looking toward emerging technology solutions, including the newer forms of artificial intelligence (AI), to assist in workforce and capacity planning.1

Applying analytics and AI to optimize these types of operational efficiencies, however, requires models that can provide trustworthy results, according to Christian Hardahl, Advisory Industry Consultant in Health Care and Life Sciences Industry Solutions at SAS.

“We know that we are facing nursing and physician shortages. We know we have a growing elderly population that will put more pressure on healthcare systems. And we know that many clinicians feel overworked,” he said. “We can use models to forecast and plan to optimize staffing and capacity, but it is important to have reliability and transparency as part of those analyses.”

The right data for optimization

When trying to match staffing to capacity and demand, flexibility is key. Hardahl said that organizations trying to make these important determinations benefit when they can include a variety of data points not just from human resources records and overtime logs, but a wide range of other sources including salary, education and behavior-based data points across demographic groups to project workforce supply and full-time equivalent (FTE) capacity over time. Additional information, including emergency department (ED) flow, operating room (OR) times, staff schedules and more can also help optimize an organization’s workforce management.

“You want to include the data that will help you conduct both short-term and long-term forecasting,” he explained. “In the short term, you want to make sure you have the right talent to take care of the patients that you have or expect to have. Taking a longer-term perspective, you might want to look at churn models as well as long-term planning, behavior and education data to understand which doctors or nurses might be at risk of leaving and come up with a plan to address it.” For example, tracking workforce entry, re-entry and exit patterns across different demographic groups and applying behavior-based methods can help project workforce supply and full-time equivalent capacity over time.

By placing these different variables into AI models, healthcare stakeholders can gain deeper insights into workforce management, provided they can defend and trace the data used to inform them. This is where approaches like retrieval-augmented generation (RAG), a large language model (LLM) that leverages relevant, verified data for analysis, become highly relevant. By grounding LLMs in curated, domain-specific knowledge bases, RAG ensures outputs are anchored in verified and contextually appropriate sources. In practice, this allows RAG-enabled agents to continuously retrieve and apply relevant policy and regulatory context to workforce analytics, making those dashboards not only more interactive but more trustworthy and defensible and leading to dynamic, policy-aware decision support, said Hardahl.

Workforce management and staff effectiveness

The use of RAG models for workforce optimization goes beyond scheduling. Service-line managers can use AI to gain deeper insights into staff performance and efficacy. They can use the available data to analyze and visualize resource allocation, employee retention and workforce investments. Taken together, these analytics can ultimately help enhance workforce productivity, patient satisfaction and employee engagement while keeping costs in check.

“This kind of technology can give you insights and efficiencies that we have not thought of before,” Hardahl noted. “We have only scratched the surface of what is possible with RAG. And, as we put strong guardrails in place, make sure we are acting in an ethical manner and gain access to strong data, we are going to see AI’s value in making improvements to our workforce decisions and overall operations.”