Complex care is meant to be a radically different mode of care delivery than what occurs in traditional healthcare. Instead of 15-minute doctor visits at a clinic punctuated by longer hospital stays, in complex care, programs focus on both health and social needs, typically through an interdisciplinary team. For example, the participants in the Camden Coalition’s care management programs  enter into longer-term,  person-centered engagements with a team that includes community health workers, social workers, nurses, and even lawyers. The “high-touch” nature of our care management philosophy means that participants get much more access to and interaction with their care team than patients get in most care settings. Because of their different needs and trajectories, participants in our program can end up staying in programs for highly varying amounts of time, ranging from a few weeks to many months or even years.  In fact, one of the major lessons learned from the randomized controlled trial (RCT) we participated in is that many people need to be engaged longer than our original program design of 120 days.

That all means that the staffing needs of complex care programs are also significantly different — and often more complicated to figure out — than those of traditional healthcare or social services models. As a data-driven organization, we wanted to tackle the question of care team workload and staffing needs in a quantitative way.

“There’s a lot of uncertainty over what it looks like to do this work in terms of staffing, in terms of how patients move through the care management model,” says Aaron Truchil, Senior Director of Data and Quality Improvement at the Camden Coalition. “We really wanted to understand that patient movement and be able to quantify it, and to come up with ways to reduce unnecessary burdens on the care team and avoid under-utilization of staff hours.”

To do that, we teamed up with researcher Dr. Hari Balasubramanian, Associate Professor of Mechanical and Industrial Engineering at the University of Massachusetts Amherst. While industrial engineering might not sound at all related to complex care, it turns out that their methodology can be quite useful for thinking about workload.

In industrial engineering, Hari says, “we use mathematical models, quantitative methods, and data science approaches to improve systems and make them more efficient.” Hari started out studying scheduling problems in semiconductor manufacturing, and got into the healthcare field when he joined the Mayo Clinic for a postdoctoral position. In 2008, he joined UMass Amherst to start a program working with hospitals and primary care providers to determine appropriate patient panel sizes for providers who were increasingly burnt out and burdened with managing electronic health record data.

When he heard about the Camden Coalition, Hari was intrigued by our unique care management model — and also by our data. “In general in healthcare, it’s very hard to get data,” he says. “You can build all sorts of models, but if they’re not grounded in practice, then it’s really an academic exercise. Additionally, in hospital settings, you have very physician-centric data. You may have some data on nurses and their work patterns, but even there you see a drop-off.”

“The Camden Coalition’s dataset is unique,” he continues. “There are not many places that collect data on how much time a community health worker spent helping an individual and what they helped them with. But at the Camden Coalition, the staff actually record through their system what they do and how much time they spend on it. So you can piece together for a single patient — what happened over time? Who did they engage with? What was the outcome? How long did they stay in the intervention? How many hours did they need with the registered nurse, the community health worker, the social worker, and so on?”

Modeling staff time

The first paper that Hari’s team and our research team co-published simply used the data we had collected to analyze what our participants’ trajectories through the care management program looked like, and how much care team time they needed at different points.

The second paper was published in June 2023 in Health Systems. This time, Hari’s team, led by doctoral student Ekin Koker, actually built a model that could predict how many hours of each staff type’s time would be needed to be spent with patients per week depending on how many participants were being enrolled into the program per week.

“For example, if you have three patients enrolled into the program per week on average, then it turns out from our model that on average you need 37 hours of community health worker time,” says Hari. “But ideally, you don’t want to staff at the mean, because that means most of the time you will be over your planned capacity. If you staff at the 80th percentile instead, which in this case is 45 hours of CHW time, you might sometimes be over capacity, but you won’t have too many backlogs or delays.”

Of course, not all programs are in what Hari calls a “steady state” of new enrollments. So the paper also simulates staffing for a program that has varying numbers of participants being enrolled each week. “You can use the methodology in the paper to transition your staffing needs as operations ramp up or down,” he notes.

New research questions

In June 2022,  Hari and the Camden Coalition were awarded a new, three-year grant from the National Science Foundation to continue this work. One direction that this new research might take: leveraging other work of our research team exploring how the “dosage” of care team time received by participants is associated with higher or lower hospital readmission rates. “Can we lay out how [under]-staffing leads to delays in care, which can lead to readmissions?” Hari asks.

The research will also be affected by how our care management model itself evolves. “One of the things we’ve come to recognize from the RCT is that these short-term interventions aren’t going to move the needle on things like readmissions,” says Aaron. “Our participants are very sick and have been underinvested in their entire lives. We need to have longer-term relationships. So there are people that we’ll continue to work with as long as they continue to want to be in our program.”

The length of program enrollment may also be an area of study. “We see in the data that there is tremendous variation between participants in terms of the amount of time they need in the program,” says Hari. “So, with how much accuracy can we predict who needs longer interventions?”

What this means for the complex care field

Though the staffing model described in the Health Systems paper is based specifically on the workings of our care management programs, Hari and Aaron hope that other complex care programs can see the utility in these methods and will consider applying them to their own programs to help make their staffing as efficient as possible.

“This is a roadmap of how we did it that others can hopefully follow with their data,” says Aaron.

What should other programs consider as they scale up their operations? “One, you should be tracking the data in a way that allows you to see interaction levels between [participants and different] staff types,” he continues. “Second, you have to know that for each staff type there will be weeks where they’re a little over capacity and weeks where they’re a little bit under. You have to make some decisions about how you’re going to structure that and what a reasonable expectation is.”

In complex care, says Aaron, “there are a lot of factors that are outside of our control. How do we coordinate everything so that our care team can be working cohesively with the highest volume of patients possible so we can be producing the highest level of impact on our community?”

“At the core of it, whether for semiconductor chips or care management staff, we’re trying to optimize how we use our limited resources to the best of our ability.”

Interested in learning more?