Making teams work well together is a challenge in any workplace. In a healthcare setting, miscommunication, clashing personalities, or uncertainty about responsibilities can lead not just to increased costs but to mistreatment, hospital readmissions, and even deaths.
In a recent study, Marissa King, a professor of organizational behavior at Yale SOM, and Ingrid Nembhard of the Wharton School set out to use new technology to understand how healthcare teams operate. Traditionally, teams have been studied using retrospective surveys; King and her team added wearable sensors, allowing them to gather real-time data about when and how team members interacted. Applying advanced analytical techniques to the massive amounts of data generated by the sensors, they were able to tease out some of the team dynamics that led to better outcomes.
Part of the goal was to test the effectiveness of a frequently used tool for making healthcare function better: the addition of a care coordinator, who is responsible for tracking and integrating various aspects of a patient’s care. But the study also yielded insights into how changes to status differences within a team can become disruptive.
Q: You used wearable sensors to gather data about a healthcare team operates. Tell us about the project.
We’re interested in understanding healthcare coordination within teams. And the sensors were one way of understanding how team dynamics impact health outcomes.
The project focused on 13 different healthcare centers in Connecticut. Within those 13 centers, we were studying team performance among 66 healthcare teams. Our primary goal was to understand how to improve healthcare coordination and reduce gaps in care. We utilized wearable sensors along with more standard assessment tools, like surveys, in order to assess how teams were performing, and how our intervention, which implemented a healthcare coordinator role, changed dynamics within the team.
Q: What were you trying to learn?
First and foremost, we were interested in whether or not having a nurse serve as care coordinator changed healthcare outcomes.
Healthcare coordination failures are one of the biggest healthcare problems in the United States; care coordination failures cost between $25 billion and $45 billion annually. The standard way of addressing this problem in primary care organizations is to have a nurse care coordinator—one person on the team who is responsible for the patient’s care both inside this organization and out.
So our first goal was to understand whether having a care coordinator actually improves healthcare outcomes. There are a lot of mixed findings in the literature.
We worked with an organization that was implementing a program that designated a nurse within each care team to help create continuity in care for complex patients. This is one of the standard approaches to implementing care coordination. Nurses involved in the intervention underwent a 24-hour training on how to care for complex patients, received a playbook, and had access to a dashboard that allowed them to enroll and track patients. We studied dynamics within the teams that the nurses were working in before and after the implementation of the program and also observed control teams that did not receive the training until after our study was completed.
Our second goal was to understand how those team dynamics unfold over time, with an idea that by studying these team dynamics more closely we can start to understand some of the problems and how to resolve them.
And then third, we wanted to see whether or not, just in general, teams have signatures that are associated with better patient outcomes. So, for instance, the way conversations flow, or how much time a doctor spends listening to a patient—does that affect, in the long term, their patient’s outcomes?
Q: What does a care coordination failure look like?
A care coordination failure is anytime there’s a gap in your care. For instance, if you go to the hospital and your hospital visit was not communicated to your doctor, that would be a care coordination failure. Because you really need that continuity from the hospital back to your primary care, so you aren’t readmitted.
You can also think about care coordination between specialist and generalist. If you’re seeing a psychiatrist, are the medications you’re getting from your psychiatrist being communicated back to your primary healthcare clinician? Because the patients that are taking up the most healthcare resources have a lot of chronic conditions simultaneously, coordination among their different providers is really key, both to the patient’s health and to keeping costs down. Coordination failures are associated with overtreatment, undertreatment, hospital readmissions, and also, ultimately, because of the inability to control chronic conditions, fatality.
Q: How have researchers looked at how teams work together in the past? What do you miss using conventional methods?
Most of the time, researchers will utilize survey instruments to look at team dynamics, asking teams how they perceive their team’s ability to function and work together as a group. Occasionally you’ll see researchers also utilizing network instruments, asking who’s connected to who, how frequently they talk to each other. The problem with both of those instruments is that they’re really subject to recall bias—our memories are just not that accurate.
So utilizing wearable sensors is one way to start to address some of the problems that are endemic to survey methods. Using sensors also is beneficial because we can watch the process unfold second by second in real time. We can see the entire dynamic unfolding, versus surveys, which by nature are going to be cross-sectional, so we can only observe responses in very limited periods rather than watching processes unfold.
Q: How did you utilize the sensors to illuminate these larger issues?
For both our treatment and control group, we ask them to wear sensors, which are about the size of your phone, and collect second-by-second data on conversational characteristics—how much time you’re spending listening, how often you’re interrupting, how long your speaking segment length is—as well as who you’re interacting with. We tracked throughout the course of a two-week pre-period, as well as the post-period after the intervention, when we elevated a nurse into the care coordinator role—both their interactions as well as the characteristics of those interactions.
Q: Were the results what you expected?
No, everything was contrary to what we anticipated. We set up the experiment really to try to improve care through this care coordination intervention. We quickly realized that the intervention wasn’t nearly as successful as we anticipated going in. In fact, the elevation of the nurse’s position within a network actually induced status conflict. Increasing the nurse’s status within her team increased interruptions within the team; it decreased the amount of time that doctors and physicians and nurses were spending listening to each other. The interaction dynamics that ensued from elevating the nurse, rather than being better, actually got much worse.
Q: What do you attribute that to? Were the doctors trying to preserve their power?
Well in any human interaction, we all have a space hierarchy. And either consciously or unconsciously, we’re geared to maintain that relative ordering. Everybody, I think, finds comfort in knowing where they sit. So when you change that, everybody is disrupted. It wasn’t necessarily that the doctors consciously were trying to keep the nurses down, but there’s just a perturbation to the hierarchy. Everyone’s trying to see where they sit.
Q: What did you learn about the relationship between how teams function and the effectiveness of their care?
Una Lee, who was a graduate student here and is now an assistant professor at Columbia, led this piece of the study. She found that the greater the amount of time a physician spent listening to their patient, the longer their speaking segment length and the less that they interrupted, the more the patients had increased hypertension control and diabetes compliance. And the effects were quite large. We were observing changes in patient outcomes that were quite large based on these very simple conversational characteristics.
Q: What do you find particularly interesting about healthcare teams? Can you apply what you learn about healthcare to other types of teams?
I think that healthcare teams are particularly interesting because you have this clear status hierarchy. And this is one thing that drew me to this project, in addition to the fact that healthcare is incredibly important. Because you have this clearly delineated hierarchy, you can understand status dynamics that, in a way that’s much more difficult in teams in which the status order is more concealed. We know that there’s always a hierarchy and there’s always a status order, but because it’s explicit in the medical context it makes it much easier to see.
It also means, though, that generalizing the results to other teams is slightly different. I think that our findings apply—anytime that you’re going to expand the status of someone in the middle of a hierarchy, you’re likely to see status conflict like we observed. But it might be slightly attenuated in teams where the hierarchy isn’t as clearly established.
Q: Are there other aspects of team performance that you’d like to look at using the same methods?
There are two studies that I have ongoing that I’m really interested in. One is how spatial layout affects interactions. So with a graduate student at University College in London, we’ve been understanding how different hospital ward layouts—whether or not they’re structured as just a corridor or whether they’re U-shaped or whether they’re a racetrack—how that affects interaction dynamics, and whether or not that translates into healthcare outcomes. How physical space structures interaction dynamics is one aspect I’m interested in. I’m particularly interested in what happens in the break room, which I think as researchers we oftentimes don’t see.
Then the second part I’m interested in is understanding team formation and how sticky people’s behavior is in different teams. For instance, how much do I have one conversational style that always sticks across teams? Does my behavior change when I’m interacting in different groups? I think that has a lot of applications if we’re trying to get people to change their behavior—how much of that change is going to have to come from that individual versus how much can we actually utilize social dynamics to change behavior.
Q: How does big data change your research?
Because we can watch team dynamics unfold in a way that we’ve never been able to see before, we can actually start to understand, both structurally and behaviorally, how these things shift and how shiftable they are over time. Our intervention was actually over a long period of time—it was a three-month intervention. That really changes the way that we’re doing social science. Rather than just putting you in an experiment and being able to see if you change your behavior over an hour, we can now watch this unfold over weeks and months to see how sticky that behavior is and what we can actually do to reinforce it to make positive changes.
Q: How can managers in other contexts benefit from these insights?
The first thing I’d say is, status conflict is rife in organizations. Even in this well-intentioned intervention, where we anticipated that health outcomes would improve, no one was attentive to the possibility that this would actually generate status conflict and interaction dynamics that would be deleterious to the group’s performance. Just be aware that status conflict exists and that managers often don’t recognize that it’s happening. If you think about where a manager sits, they’re often outside the team or the unit that’s going to be impacted. Those hierarchies are really sort of developing within the team and often times there aren’t physical markers like there would be in a healthcare organization. Being aware that that’s a possibility is important.
Two, going into any sort of instance in which you’re going to give some sort of external validation or promotion, realizing that you may not be attentive to status conflict, and in those cases potentially using peer recognition. To the extent that peers are conveying back to you information about who should be elevated or be recognized, that may be less likely to incite status conflict than externally you intervening on a group without understanding the internal dynamics of that group. Peer recognition, I think, can often be a more effective way of intervening in a group performance that’s not as disruptive.
And then, three, we also know in general what makes teams less likely to be subject to status dynamics. Trying to create psychological safety in teams—that’s one way to help guard against status conflict, because people are actually willing to express status differences and say, “Hey, I realize somebody needed to take on this role, but why wasn’t it me?” Status tends to be, in organizations, something like religion or politics—we just don’t talk about status. By creating an environment with psychological safety where people can air these grievances, then I think you can address them, rather than them have team members expressing themselves by not listening to you or disengaging from interacting with you.