Since 2020, we have dived deep into the data collected as part of our randomized controlled trial (RCT) to not only improve our person-centered care management model, but also add to the evidence base for what works to improve the health and well-being of people with complex health and social needs.

We are excited to announce the results of a new study published today in JAMA Network Open that re-analyzed the RCT data using a “distillation” methodology, an award-winning novel method developed by researchers at Kaiser Permanente that provides additional insights into RCTs. This new analysis shows that greater likelihood of participation in the Camden Core Model was associated with 1) lower hospital readmission rates 30 and 90 days after discharge, and 2) fewer hospital readmissions 30 and 180 days after discharge.

The study authors, which include members of our research team as well as partners from Kaiser Permanente Bernard J. Tyson School of Medicine, concluded that the Camden Core Model does in fact reduce hospital readmissions when the intervention can be delivered as intended. These findings suggest two crucial avenues for improving the impact of our programs for the hardest-to-reach patients:

  • Strive to understand and measure engagement barriers, and continue to tailor our programs to support participants’ active engagement in them
  • Work with our ecosystem partners to strengthen and broaden the services available with the goal of connecting the individuals we work with to the resources they need

What is the distillation method?

Developed by researchers at Kaiser Permanente, the distillation methodology is a way of compensating for low levels of participation in the program being evaluated by an RCT — without introducing new biases into the evaluation. The methodology was awarded the 2023 James F. Burgess Methods Article-of-the-Year Award by Health Services Research.

The Camden Coalition’s RCT modeled the average impact of our Camden Core Model across all individuals randomized to treatment, regardless of the amount of time our care team staff worked with the individual. We have recognized that there is considerable variability in how much time our care team spends with clients: some we struggle to engage off the bat and receive very little support, while and others who receive many hours of time with care team staff.

Because the RCT employs an “intent-to-treat” approach, if fewer people in the intervention arm participate in the program than was expected, the RCT is no longer directly comparing people who received an intervention with people who did not, because the intervention arm is diluted by non-participants. However, simply removing the non-participants from the intervention arm also compromises the integrity of the results, because it means that the intervention and control groups can no longer be directly compared.

Enter distillation: a three-step process that first determines which baseline characteristics are predictive of program participation, then applies that predictive model to both the intervention and control arms of the study, creating subsets with higher likelihoods of program participation that are still able to be directly compared. The final step is to compare the outcomes between the intervention and control arms within this “distilled” population subset, using similar statistical models as the ones used in the original RCT.

This method allows researchers to see if there is a difference in outcomes when people actually engage in the intervention they were randomized into. It also provides clues about what drives variation in engagement levels  in the first place, and how barriers to engagement can be addressed.

What we learned when we distilled our RCT population

Camden Coalition Data Scientist Qiang Yang, PhD; Director for Research & Evaluation Dawn Wiest, PhD; and Senior Director of Data Analytics & Quality Improvement Aaron Truchil worked directly with the Kaiser Permanente researchers who developed the distillation method — Anna Davis, PhD and John Adams, PhD — to apply their method to our RCT data.

The distillation process itself produced interesting results. Some medical conditions (e.g., kidney disease, chronic obstructive pulmonary disease) were more prevalent among the subsets of patients predicted to have higher engagement. In addition, patients more likely to engage were less likely to have been arrested prior to enrollment, hospitalized three or more times in the six months prior to enrollment, or to have unstable housing.

Each of these insights provides potential avenues for further investigation and improvement. For instance, on a webinar with the Better Care Playbook, Dawn noted that after finding that prior arrests were a barrier to participation, we worked with our partners in the criminal justice system to ensure that our care teams have access to up-to-date criminal justice data and the ability to follow-up with program participants who are incarcerated in the Camden County Correctional Facility. Hopefully, these adjustments will make it less likely that criminal justice-involved participants are lost to follow-up.

Addressing barriers on a program level

The finding that criminal justice involvement and housing instability were associated with lower probabilities of program participation reinforced what we have heard from our care team members and program participants for years: addressing health and wellness is extremely difficult without a baseline level of stability.

Those observations led to the launch of two of our most impactful programs. Our Housing First program pairs permanent, affordable housing with wrap-around services and supports, and housed participants have seen a 62% average reduction in emergency department utilization. Our Medical-Legal Partnership with Rutgers Law School helps participants address the civil and criminal legal issues that stand in the way of their health and recovery.

From intervention to ecosystem

When we began collecting data for the RCT in 2014, we hoped that an intensive, 90-day care management intervention    would allow us to connect individuals with multiple, complex health and social needs with ongoing care that could keep them out of a cycle of crisis. As we learned very quickly, the issues our participants face are often too acute and too broad to be solved in 90 days — or to be solved by a single organization alone. This realization has been the impetus for our growing focus on building and strengthening community-based ecosystems of care.

As we use the data from this new study to ensure that our care management model is engaging those who will benefit from it the most, we are also working with partners in Camden and South Jersey to ensure that people with complex health and social needs have many options for quality, ongoing care.

“Short term care management for this population is not enough,” says Camden Coalition President and CEO Kathleen Noonan. “The finding that patients who were experiencing homelessness and other social needs were among the least likely to engage in our care management program only underlines the need we see for broader investment in social services and other resources.”