Identifying High-Need Patient Profiles That Respond to Intensive Care Management: Insights from the Camden Health Care Hotspotting RCT

This study utilized latent class analysis on a randomized controlled trial to identify distinct high-need patient subgroups, revealing that tailored intensive care management significantly reduced readmissions and emergency visits for specific profiles, thereby suggesting that segmenting patients by medical, behavioral, and social risk factors can enhance the effectiveness of complex care interventions.

The Gist

Imagine the healthcare system as a massive, busy airport. Most passengers (patients) just need a quick boarding pass and a seat on a standard flight. But there's a small group of "frequent flyers" who are constantly running through the terminal, missing connections, and landing in the emergency room. These are the high-need patients.

For years, doctors and hospitals have tried to help this group by giving them a "personal concierge" (a care management team) to guide them, help with meds, and find housing. The problem? Sometimes this works great, and sometimes it seems to do nothing at all. It's like giving everyone the same map, even though some people are lost in a forest, some are stuck in a swamp, and others are just trying to find a coffee shop.

This paper is like a detective story where the researchers decided to stop looking at the crowd as one big group and started looking at the individual travelers to see who actually needed what kind of help.

The Investigation: Sorting the Passengers

The researchers took data from a big experiment in Camden, New Jersey, involving nearly 800 people who were frequent hospital visitors. Instead of treating them all the same, they used a special statistical tool (called Latent Class Analysis) to sort them into four distinct "traveler profiles."

Think of it like sorting a messy pile of luggage into four specific bins:

1. Bin 1: The "Lost & Struggling" Group (Behavioral Health & Housing Instability)

   • Who they are: These passengers are dealing with heavy emotional storms, substance use issues, and often have no place to sleep. They are the most chaotic travelers.
   • The Result: They needed the most help from the concierge team (the most hours of support). While they didn't stop going to the ER immediately, they started going less after a while. It took time to get them settled, but the help eventually worked.

2. Bin 2: The "Broken Engine" Group (Multi-system Medical Complexity)

   • Who they are: These passengers have serious, complicated physical problems affecting many parts of their bodies (heart, diabetes, kidneys, etc.). Their "engines" are just very complex and hard to fix.
   • The Result: Even with the concierge team helping, they kept landing in the hospital. Their medical needs were so deep and complex that a standard care team couldn't stop the frequent trips. They might need a completely different, more specialized type of mechanic.

3. Bin 3: The "Breathing Trouble" Group (Pulmonary Health & Substance Use)

   • Who they are: These travelers are struggling with lung issues (like asthma or COPD) mixed with addiction problems.
   • The Result: The standard care team didn't seem to change their pattern much. They kept visiting the hospital at the same rate. This suggests they might need a very specific type of help focused on lungs and addiction, rather than general care.

4. Bin 4: The "Quiet Strugglers" Group (Lower Overall Complexity)

   • Who they are: These passengers have some health issues, but they aren't as severe or complicated as the others. They are the "hidden" group that often gets overlooked because they don't scream for help as loudly.
   • The Result: This group actually saw the biggest improvement! With a little bit of support, they stopped going to the hospital as much. It turns out, a lighter touch was all they needed to keep them from spiraling.

The Big Lesson: One Size Does Not Fit All

The main takeaway from this paper is that you can't use a "one-size-fits-all" approach.

• If you give the "Lost & Struggling" group a light touch, they might drown. They need heavy, sustained support.
• If you give the "Broken Engine" group the same light touch, it won't fix the engine. They need a specialist.
• If you ignore the "Quiet Strugglers" because they seem "fine," they might crash later. A little help goes a long way for them.

Why This Matters

Imagine a hospital administrator trying to fix traffic at the airport. Before, they were handing out the same "Helpful Guide" pamphlet to everyone. Some people threw it away, some couldn't read it, and some needed a wheelchair, not a pamphlet.

This study says: "Stop handing out the same pamphlet!"

By identifying these four specific groups, hospitals can now:

• Send a heavy-duty support team to the "Lost & Struggling" group.
• Send a specialist medical team to the "Broken Engine" group.
• Focus on lung and addiction experts for the "Breathing Trouble" group.
• Give a light, preventive check-in to the "Quiet Strugglers" to keep them healthy.

The Bottom Line

This research is a step toward precision medicine for social and medical care. It teaches us that to fix the healthcare system, we need to stop treating "high-need patients" as a single blob. Instead, we need to look closely at who they are, what their specific struggles are, and give them the exact right kind of help. That's how we save money, save lives, and make sure everyone gets the care they actually need.

Technical Summary

1. Problem Statement

Complex care management programs aim to reduce healthcare utilization (readmissions, ED visits) and costs for "high-need, high-cost" patients. However, evaluations of these programs have yielded mixed results, with many randomized controlled trials (RCTs) showing no overall reduction in hospitalizations.

• The Core Issue: Traditional "one-size-fits-all" interventions fail to account for the heterogeneity within high-utilizer populations.
• The Gap: Aggregate analyses often mask differential treatment effects. Patients with distinct medical, behavioral, and social risk profiles may respond differently to the same intervention. Furthermore, standard risk stratification often relies solely on claims data, overlooking critical social determinants of health (SDOH) that drive engagement and outcomes.
• Objective: To use Latent Class Analysis (LCA) to identify distinct patient subgroups within the Camden Coalition's Core Model RCT and determine if specific profiles respond differently to intensive care management.

2. Methodology

This study is a post-hoc exploratory analysis of the Camden Coalition's Core Model RCT (ClinicalTrials.gov: NCT02090426).

• Study Population:

  • Sample: 780 adults (aged 18–80) from Camden, NJ, with multiple chronic conditions and frequent hospitalizations.
  • Groups: Randomized into Treatment (n=391, receiving multidisciplinary care management) and Control (n=389, usual care).
  • Intervention: A multidisciplinary team (nurses, social workers, community health workers) provided care coordination, coaching, and social service navigation for 3–4 months post-discharge.

• Data Sources:

  • Clinical/Social Data: Pre-enrollment survey data (housing, employment, language) and hospital encounter data from the Camden Coalition Health Information Exchange (HIE).
  • Variables: 79 clinical conditions (mapped via Clinical Classification Software - CCS), 4 social factors (housing, marital status, education, support), and prior utilization metrics.

• Statistical Approach:

  • Latent Class Analysis (LCA): Implemented using the Stepmix package in Python. LCA identifies unobservable subgroups based on patterns of observed variables.
  • Model Selection: Models with 1–10 classes were tested. The 4-class model was selected based on the Bayesian Information Criterion (BIC), average posterior class membership probability, and clinical interpretability.
  • Outcome Measures: Inpatient readmissions and Emergency Department (ED) visits over two consecutive 6-month periods post-discharge.
  • Analysis: Differences between intervention and control groups were tested using two-sample proportion Z-tests. No multiple testing corrections were applied as the study is hypothesis-generating.

3. Key Contributions

• Methodological Innovation: Demonstrates the utility of LCA in healthcare research to move beyond aggregate risk scores, integrating clinical, behavioral, and social risk factors simultaneously to define patient profiles.
• Subgroup Discovery: Identifies four distinct, clinically interpretable patient archetypes that explain the variability in intervention effectiveness.
• Equity Focus: Highlights how social determinants (e.g., housing instability, language barriers) create distinct subgroups that require tailored engagement strategies, addressing limitations of claims-based stratification.

4. Key Results

A. Identified Patient Classes

The LCA revealed four distinct subgroups:

1. Class 1: Behavioral Health & Housing Instability (n=169)
   • Profile: High prevalence of mental health conditions, substance use, HIV, hepatitis, and homelessness. Younger mean age (47.8 yrs), higher proportion of White patients.
   • Utilization: Highest pre-enrollment ED and inpatient use.
   • Intervention Response: Received the highest staff hours. Showed a trend toward reduced ED visits in the second 6-month period compared to controls.
2. Class 2: Multi-system Medical Complexity (n=221)
   • Profile: Broad spectrum of complex conditions across multiple organ systems (circulatory, endocrine). Oldest mean age (63.1 yrs), high Charlson Comorbidity Index.
   • Utilization: High utilization.
   • Intervention Response: Received the lowest staff hours despite high complexity. No significant reduction in readmissions or ED visits compared to controls.
3. Class 3: Pulmonary Health & Substance Use (n=129)
   • Profile: Respiratory conditions combined with anxiety and substance use disorders.
   • Utilization: Moderate utilization.
   • Intervention Response: No measurable benefit observed in readmissions or ED visits.
4. Class 4: Lower Overall Complexity (n=261)
   • Profile: Moderate circulatory/endocrine issues but lower overall disease complexity. Highest employment rate.
   • Utilization: Moderate utilization.
   • Intervention Response: Showed a significant reduction in inpatient readmissions in the second 6-month period (-11.18 percentage points, 90% CI: -20.86 to -1.2) compared to controls.

B. Overall Trial Outcomes

• Aggregate Level: The full study population showed no statistically significant difference in readmissions or ED visits between intervention and control groups in the first 6 months. A trend toward reduced readmissions appeared in the second 6 months (-6.4 pp), but significance was marginal.
• Subgroup Level: Significant heterogeneity was found. The "Lower Complexity" group (Class 4) drove the readmission reduction, while the "Behavioral Health" group (Class 1) showed reduced ED usage. The "Multi-system" group (Class 2) showed no improvement.

5. Significance and Implications

• Precision in Complex Care: The study validates that "average treatment effects" are misleading in heterogeneous populations. Success depends on matching intervention intensity and type to specific patient profiles.
• Resource Allocation:
  • Class 4 (Lower Complexity): Can benefit from "lighter-touch" interventions that prevent escalation, suggesting current programs may be over-investing in this group or under-investing in others.
  • Class 1 (Behavioral/Housing): Requires sustained, high-intensity outreach to overcome initial engagement barriers; benefits may be delayed.
  • Class 2 (High Medical Complexity): May require specialized, highly medicalized care models rather than standard care coordination, as current interventions failed to reduce their utilization.
• Equity: The findings suggest that excluding patients based on perceived barriers (e.g., language, housing instability) exacerbates disparities. Conversely, focusing only on the "most visible" high-need patients may miss the "Lower Complexity" group who are also responsive to intervention.
• Future Directions: The authors advocate for segmentation-informed implementation, where healthcare systems use similar LCA approaches to tailor outreach, engagement strategies, and service intensity to local patient populations, moving toward equitable, evidence-based precision medicine.