Through our signature complex care management program, the Camden Core Model, our care teams — made up of nurses, social workers, and community health workers — engage frequently hospitalized individuals with complex health and social needs and work with them to reach their own health and well-being goals. We’ve learned that measuring both the inputs and outcomes of such a highly individualized intervention is, well, complex.
From 2014-2018, our staff worked with researchers from J-PAL to run our Core Model as a randomized controlled trial (RCT): patients who enrolled were randomized to either the Core Model intervention group or to the control group that got normal post-discharge care. Analysis showed no statistically significant differences in hospital readmission rates between the intervention and control groups.
But looking at the data for the intervention group, our research team noticed that there was a lot of variation in the number of hours staff spent interacting with and coordinating care for participants. Is it possible that there is an association between the amount of staff time and effort that participants received and how likely they were to be readmitted to the hospital?
“[The Camden Core Model] is not a simple medical intervention where the patient either takes the medication or does not,” says Qiang Yang, Data Scientist for Research & Evaluation at the Camden Coalition. “It requires long-term implementation to change patient behavior and lead to better outcomes.”
The amount of implementation or engagement that participants receive can be thought of as a type of dosage, allowing us to compare hospital readmissions for “high dose” and “low dose” groups of participants.
The results of this analysis were published this month in the American Journal of Managed Care. We found that (after controlling for participant demographics as well as clinical and social variables) a higher dosage of the Camden Core Model was in fact associated with a lower readmission rate at 30 and 90 days post-enrollment. There was no statistically significant difference at 180 days.
The relationship between dosage and outcomes
Because this study only looked at data from the intervention arm of the RCT, and looked at it retrospectively, it doesn’t tell us whether a higher intervention dosage caused participants to have lower readmission rates. It does, though, demonstrate an association between staff time interacting with and coordinating care for participants and those participants’ readmission rates.
We also saw that the strength of that association waned over time. At 30 days post enrollment, the high dosage group’s readmission rate was 21.6%, compared with 36.6% for the low dosage group. At 90 days, the rates were 41.7% for the high dosage group, and 55.2% for the low. The difference at 180 days — 57.5% vs 64.9% — was not statistically significant.
This shrinking gap could be because intensive care team engagement helps stabilize participants as they return from the hospital to the community, but the intervention is of limited duration, and over time the medical and social complexities the participants experience may outweigh the help they’re getting from the care teams.
“For some participants, longer-term engagement may be necessary, and they may need more resources,” says Dawn Wiest, Director of Action Research and Evaluation at the Camden Coalition. “The trick is to figure out who needs what early on in the intervention.”
Measuring participant engagement
So why does the intervention dosage vary so much in the first place? We have some guesses, but this question will ultimately need more research to answer.
“In these complex, behavior change-oriented interventions, there is always variability in the extent to which participants can engage fully in the intervention,” says Dawn. “The participant may be fully motivated when they’re first approached in the hospital, but when they’re back in the community, they may be returning to circumstances that make it difficult to impossible for them to engage. On the care team side, having more participants on a panel may make it harder to engage each participant.”
And if you’re not measuring the barriers that could be inhibiting engagement — as we were not — the data can’t tell you why the variability in engagement exists.
“That has implications for when you’re trying to assess whether an intervention was successful or not,” Dawn says.
It also has implications for program design and operations. For programs with large participant panels and limited resources, tracking barriers to engagement early may be key to ensuring that staff are able to implement the intervention faithfully to the design.
One key takeaway from this analysis is that dosage matters and needs to be accounted for in study design, including in RCTs.
“Randomization doesn’t wash away the variability of the intervention group,” says Dawn. “You need a design that will take levels of intervention exposure into account.”
Another is that, though they are rarely included in evaluation protocols, both assessing participant motivation and identifying barriers to delivering an intervention as intended are crucial.
As a result of this analysis, we have incorporated new fields into our data management system to help track participant engagement and identify barriers. Our redesigned system, like the Camden Core Model itself, is centered around participant goals, and allows both our care teams and data team to track participants’ progress toward reaching them.
“We will continue to use lessons from the RCT to shape our work improving care for people with complex health and social needs here in Camden, as well as the efforts of the field of complex care,” says our President and CEO Kathleen Noonan.