In Nashville, in a room full of hospitalists who spend their days thinking about flow, safety, and scale, this session landed in a familiar but underexamined space. The focus was practical and grounded: how advanced practice practitioners (APPs) are actually deployed on the ground, and what that means for care delivery. The session, led by Christopher Bruti, MD, MPH, a division chief in hospital medicine, and Erik McIntosh, DNP, a lead APP with a focus on integration and team-based care (both at Rush University Medical Center in Chicago), brought a dual-lens perspective that felt appropriate for the topic.
There was no attempt to oversell a single model. Instead, the discussion moved through what most of us recognize as lived experience. Different structures were tried over time, each solving one problem while introducing another. It had the feel of Nashville itself. Layered, iterative, built over time, less about a single headline act and more about how the group comes together.
The conversation in hospital medicine has shifted from structure to function, no longer being “Do you have APPs?” but “What are they actually doing in your system, and what does that do to care?” Most programs did not design their current state; rather, they inherited it, layer by layer, under pressure from volume, workforce constraints, and the steady demand to move patients through beds without breaking the system.
This origin story matters. APPs entered the workforce to address physician shortages, first in primary care and then in inpatient settings as hospital medicine matured. What persisted alongside that expansion was a flawed mental model. APPs were treated either as fractional physicians or as task absorbers. That ambiguity shows up operationally, in morale, and in throughput.
The early model was task delegation. One physician, one APP, shared rounding, a defined census. The physician owned the plan and documentation. The APP handled orders, coordination, pages, and updates. It is clean on paper, but can be inefficient in practice. Cognitive work ends up being duplicated. Documentation burden concentrates on the physician. Probably most importantly, a trained clinician is underused. The result is predictable: physicians feel like managers, APPs feel underutilized, and throughput improves less than expected.
The pendulum then swings to independence. APPs carry their own panels with minimal oversight. Capacity expands. Coverage gaps close. But the system pays in a different currency. The attending of record may not know the patient. Team cohesion degrades. Variability increases, especially in higher acuity settings where calibration matters. Programs often stall here, choosing between efficiency and integration as if they cannot coexist.
The collaborative model attempts to resolve that tension. APPs manage patients. Physicians review and co-own the plan in a structured way. On paper, this is the right balance. In practice, it only works if the interface is defined: Who sees the patient first? What gets presented? When should decisions be escalated? How should this disagreement be handled? Without that, collaboration introduces delay.
Performance data across mature programs shows a consistent pattern. Quality outcomes are comparable across team structures, with collaborative models often demonstrating greater consistency. Patient experience remains stable. Safety metrics such as falls and hospital-acquired conditions do not worsen. The value proposition is not about superiority but reliability at scale.
Operationally, the signal is similar. Length of stay and readmissions do not meaningfully differ between collaborative teams and physician-only teams. Early discharge rates may be modestly lower for collaborative than for physician-only teams, but substantially higher than in trainee-driven models. The value proposition is clear: the model does not create dramatic gains in a single metric, but it does stabilize performance across multiple domains while expanding capacity.
Cost is where the model becomes more concrete. Physician salaries remain higher than APP salaries by a meaningful margin. When APPs carry defined panels and physicians supervise within a structured framework, the cost per patient decreases. There is an optimal range. A single physician working with two to three APPs tends to balance cost efficiency with manageable oversight. Beyond that, gains flatten. The constraint is no longer salary. It is attention.
This is where most implementations fail. The physician’s experience highlights it directly. Communication quality is generally strong. Confidence in APPs varies with experience. Scope concerns are uncommon. Role tension is rare. The limiting factor is cognitive load. When an attending carries a personal panel while supervising multiple APPs, especially with complex patients or additional learners, the system strains. This is not a cultural issue, but a bandwidth problem.
From the APP perspective, the drivers are consistent across institutions. Engagement is tied to autonomy, trust, and team culture. Retention depends on clarity, growth, and recognition. APPs want to practice as clinicians, not as extensions. APPs want defined expectations, structured development, and visible advancement pathways. When those elements are present, the model stabilizes. When they are absent, turnover becomes the default pressure valve.
The collaborative model is not a staffing solution. It is an alignment solution. Talent alone is insufficient. The system depends on how individuals interact under time pressure. Misalignment shows up as delays, redundant work, and missed signals. Alignment shows up as flow.
For programs trying to evolve, several operational principles emerge:
Define the unit of work. A patient panel needs clear ownership. If the APP owns the panel, that ownership includes synthesis, plan, and communication. The physician’s role is oversight, calibration, and escalation. Without that clarity, the system reverts to task delegation regardless of intent.
Cap cognitive load explicitly. There is a finite number of decisions a physician can supervise well in a given time frame. Staffing ratios should reflect that constraint. A lower-cost model that degrades decision quality is not efficient.
Standardize the handshake. Presentation style, timing of review, and escalation thresholds should be consistent. These are core operational processes. Variability at this interface drives delay and rework.
Invest in onboarding and progression. Experience gaps are predictable. Newer APPs often struggle with synthesis, prioritization, and discharge planning. Autonomy should be staged to match capability, with clear milestones and feedback.
Integrate leadership. APP leadership should be embedded within divisional and system structures. Participation in scheduling, quality work, and committees reinforces accountability and aligns incentives. It also signals that APPs are part of the clinical enterprise, not an adjunct to it.
Lastly, measure the right variables. Length of stay and readmissions are necessary but insufficient. Variability, early discharge reliability, and communication quality are more sensitive indicators of whether the team is functioning as intended.
The field has already answered the question of whether APPs belong in hospital medicine. The remaining question is how to build a system that uses them well. The answer is not found in a single model. It is found in how the model is executed. Programs that define roles clearly, respect cognitive limits, and invest in development achieve consistency and scale. It’s a bit like Broadway: get the structure right and the handoffs are tight, the timing holds, and the whole thing moves without anyone needing to force it.
Dr. Migliore is an assistant professor of medicine at Columbia University College of Physicians and Surgeons and director of general medicine consult and perioperative services, as well as a medicine attending physician, at Columbia University Medical Center, both in New York.