David Foroughi runs operations at a PE firm that has been building an outpatient imaging platform in the Southwest for four years. He has acquired eleven freestanding centers across three states, and he can recite the volume distribution by modality without consulting a dashboard. He can also recite something that does not appear in his dashboard: the names of the four orthopedic groups whose loss would put the platform’s IRR underwater. He keeps that list in his head because the dashboards he has been buying for four years cannot tell him when one of those groups is starting to leave. They can tell him after the volume has dropped. By then the orthopedist has had six weeks of frustration with his scheduling line, the appointments have already been redirected, and the only question is how much revenue the platform absorbs before the pattern shows up in monthly reports.
The imaging portfolio carries the sharpest version of the PE relationship problem in clinical verticals. Patients do not choose imaging centers. Physicians do, and physicians choose based on operational signals the PE firm typically does not measure: how quickly the scheduling line picks up, whether the radiologist read came back inside 24 hours on a stat, whether the front desk recognized the patient by name when she walked in. When PE acquires an imaging platform and changes the front desk staff, the scheduling system, or the turnaround time targets, the referring physicians notice within weeks. The volume erosion follows by month, and by the time the quarterly review surfaces it, the referral relationship is often already gone.
The referral and relationship concierge has its most operationally critical deployment in imaging. The vertical also activates three other architectural challenges that distinguish it from other clinical settings: scheduling complexity that is modality-specific rather than slot-generic, a prior authorization burden that is the heaviest of any outpatient clinical service, and a compliance surface that varies by modality, by state, and by accreditation body.
The referral that left
A specific case from a real PE-owned three-location imaging company illustrates how fast the relationship can break. The firm acquired the platform in January. By March, central operations had decided that the scheduling experience was inconsistent across locations and consolidated all three sites to a single regional IVR running out of Phoenix. Each referring practice that called the prior direct line was redirected to the IVR with a recorded message.
For most referring practices, the change was an annoyance. For the orthopedic group that had been the largest source of MRI volume, it was operationally fatal. The group’s scheduler had been calling the direct line at 4 PM on Mondays to book the next week’s MRIs. The direct line had been answered in under thirty seconds for six years. The IVR, on average, held callers for four minutes. By the third week, the scheduler had switched to a competitor imaging center four miles away, where the direct line still worked.
The referral concierge running on the imaging platform’s operational layer detected the pattern in week two. It had a baseline for the orthopedic group’s call frequency, slot booking velocity, and case mix. By the second Monday after the IVR transition, the call frequency had dropped 60 percent against baseline. The concierge surfaced the finding to the local center manager with three data points: the call drop, the booking velocity drop, and the geographic alternative most likely to be absorbing the redirected volume. The center manager had 72 hours to act before the orthopedic group’s pattern would crystallize.
Without the concierge, the same pattern would have surfaced at the monthly volume review in early May, by which point the volume loss was already $112,000 cumulative against pre-acquisition baseline. Annualized at the orthopedic group’s contribution, the at-risk revenue was $380,000. The earlier detection let the platform restore a direct line at the Tucson location specifically for the affected orthopedic group, with a measured response time commitment that the group’s scheduler tested twice before resuming her booking pattern. The volume came back. The savings, against the counterfactual, were the difference between catching the problem in week two and catching it in week sixteen.
Modality-specific scheduling
Imaging scheduling is not slot scheduling. A brain MRI without contrast takes 30 minutes. A lumbar spine MRI with contrast takes 50 minutes. An abdominal MRI with sedation can take 90 minutes. A claustrophobic patient adds 20 minutes regardless of the indication. The same patient may need additional time if she has previously not tolerated a sequence and a different protocol has to be substituted on the day of the study.
The scheduling concierge running in an imaging center learns actual durations rather than scheduled durations. It learns them by physician, by magnet, by modality, by patient anxiety profile where that signal is available. A 1.5T magnet takes 15 percent longer than a 3T magnet for certain musculoskeletal sequences because the field strength affects acquisition time. The concierge does not assume this. It measures it from the actual case logs at this specific facility and adjusts the template scheduling accordingly.
Per-magnet optimization is the operationally meaningful unit. A three-location platform with five magnets has five distinct scheduling problems, not three. The scheduling concierge maintains a model per magnet, per modality, per radiologist availability window, and proposes template adjustments that improve effective capacity without changing the slot count visible to schedulers. The capacity gain typically runs 8 to 15 percent in the first six months, driven primarily by accurate duration prediction and reduced gap time between cases that previously ran long.
The concierge does not, in this phase, take direct booking action. It surfaces template recommendations and capacity adjustments to the center manager. The center manager approves or declines. The earned autonomy progression to operational scheduling typically takes nine to twelve months at an imaging center, longer than at a physician practice, because the scheduling decisions touch radiologist workflow more directly and the radiologists’ tolerance for autonomous booking changes is lower.
The prior auth burden
Virtually every advanced imaging study requires prior authorization. MRI, CT with contrast, PET, nuclear medicine studies, and most ultrasound studies above a certain complexity threshold all trigger payer authorization requirements. For Medicare beneficiaries undergoing advanced imaging, the Appropriate Use Criteria consultation requirement adds an additional documentation step that the ordering physician must complete or the imaging center cannot bill. For commercial payers, eviCore and AIM Specialty Health together handle authorization for a large share of the commercial population, each through its own portal with its own documentation requirements and its own approval logic.
The prior authorization concierge handles the volume that, in a typical imaging center, consumes 1.5 to 2.5 full-time-equivalent staff positions per location. The concierge processes the routine authorizations directly, attaches the documentation the payer requires for the specific study type, and submits through the appropriate portal. It escalates the cases that require physician peer-to-peer review back to the ordering physician’s office with the specific information the peer-to-peer is likely to require, rather than leaving the ordering physician to discover the requirement when the call comes in.
The portfolio-level intelligence emerges in this domain quickly. The approval pattern for a brain MRI ordered for headache by a primary care physician under United Healthcare is the same whether the patient is being imaged at Center A or Center B. The denial pattern for the same study with the same indication under Anthem in Texas is also consistent. The portfolio’s prior authorization concierge layer aggregates these patterns across centers and surfaces the patterns that affect submission strategy. When a payer policy changes, the change shows up first at one center, then within days at others. The aggregation lets the platform respond to the change at the portfolio level rather than discovering it eleven times in eleven separate locations.
Radiologist workflow intelligence
Imaging operations are not complete without the radiologists, and radiologist workflow is increasingly hybrid. On-site radiologists handle interventional procedures and complex cases requiring real-time consultation with the technologist. Teleradiology handles overnight reads, weekend coverage, and volume balancing across locations. Hybrid arrangements, where the same group of radiologists rotates between on-site and remote work, are now the dominant pattern at PE-owned imaging platforms.
The scheduling and staffing concierges coordinate on this domain. Modality scheduling drives radiologist requirement: an MRI booked at 2 PM on Thursday requires a radiologist available to read the study within the turnaround time committed to the referring physician. The reverse is also true: radiologist availability drives what can be scheduled. The concierges share state. The scheduling concierge does not book a case the staffing concierge knows cannot be covered by a subspecialty-appropriate radiologist within the committed window.
Subspecialty matching is the harder problem. A musculoskeletal MRI ideally reads by a musculoskeletal-trained radiologist. A neuro study ideally reads by a neuroradiologist. The match is not always possible, and the platform’s radiologist mix determines what is feasible at what time. The staffing concierge maintains the subspecialty mix per location per shift, and the scheduling concierge respects the constraints when proposing bookings.
Read queue management runs continuously. Stat reads require turnaround targets measured in minutes. Routine reads run on hour-level or day-level targets depending on the contract with the referring practice. Peer review, the secondary read on a sample of cases for quality assurance, runs on a separate queue and must be completed without disrupting the primary read workflow. The quality concierge tracks these queues and surfaces the patterns that predict missed targets before they manifest as escalations.
Compliance surface
The compliance surface in imaging is layered. ACR accreditation applies per modality per center: an MRI accreditation, a CT accreditation, a mammography accreditation, an ultrasound accreditation, each with its own renewal timing, its own documentation requirements, and its own equipment-specific protocols. MQSA, the Mammography Quality Standards Act, governs mammography specifically with federal oversight, annual inspections, and consequences that include suspending the center’s ability to read mammograms if findings exceed thresholds. State radiation safety regulations vary significantly: California, Texas, and New York each have distinct radiation safety regulations that apply on top of federal requirements. Dose tracking for CT scans is required by ACR accreditation and is the subject of growing regulatory attention. Equipment inspection schedules vary by modality and by state. Personnel qualification requirements differ between technologist categories and between states.
The compliance concierge tracks all of this per center, per modality, per state, per accreditation body. It maintains the document repository, surfaces renewal deadlines before they become urgent, and flags personnel certifications approaching expiration. The architectural value is not that the concierge handles the documentation, though it does. The value is that the compliance state is observable. At any moment, the operating partner can ask what the compliance status is across the platform, and receive an accurate answer at the granularity of center, modality, regulator, and document. The current alternative, which is the spreadsheet maintained by an administrator at corporate with quarterly updates, does not produce reliable answers at any granularity.
Consumer connection
Margaret has a breast MRI ordered. Her health concierge, running on the consumer side of the BlueMirror architecture documented in BMT-01.02, receives the order from her primary care physician’s portal. Through the membrane described in BMT-03.01, the health concierge initiates coordination with the imaging center her primary care physician has referred her to, which happens to be a center in the platform David’s firm owns.
The imaging center’s scheduling concierge receives the referral through a structured exchange. The exchange carries the clinical context the imaging center needs to schedule appropriately: the indication, the prior imaging history if available, the contrast tolerance from prior studies, and Margaret’s stated mobility constraints. The exchange does not carry the full medical record. The membrane enforces the constraint.
The prior authorization concierge at the imaging center obtains the authorization using the documentation the health concierge has shared. The scheduling concierge selects a slot that matches Margaret’s mobility profile, choosing a morning appointment because she has indicated that fatigue increases through the day, and selecting the location closest to her home. The scheduling confirmation flows back to her health concierge, which surfaces it to her with the time, the location, the preparation instructions, and the transportation plan if she has indicated she will need NEMT.
After the study, the read result flows back through the membrane to her health concierge. The health concierge interprets the result in the context of her broader medical history, surfaces the relevant elements to her in language she has indicated she prefers, and coordinates the follow-up appointment with her primary care physician if the result requires one.
From Margaret’s experience, the interaction is seamless. She received an order, an appointment got scheduled, transportation arrived, the study happened, and she received the result with appropriate context. Behind the experience, four operational agents at the imaging center coordinated with two consumer agents on her side, across two organizational entities, through a membrane that limited information exchange to what each side needed and recorded everything for audit. The seam is invisible because the architecture is designed to make it invisible.
Cross-References#
Prior Authorization (BOI-01.03). The agent architecture for prior authorization, deployed in its highest-volume clinical setting at the imaging center.
Scheduling (BOI-01.06). The scheduling concierge architecture, instantiated for modality-specific imaging scheduling complexity.
Referral and Relationship (BOI-01.16). The referral concierge that monitors the referring physician signals whose erosion is the primary failure mode in imaging acquisitions.
The Integration Layer (BOI-02.03). The RIS and PACS integration architecture that wraps around the imaging center’s existing systems.
The Health Concierge (BMT-01.02). The consumer-side agent that coordinates imaging through the membrane on Margaret’s behalf.
Technical Appendix BOI-03.02-A is available to partners and investors at partners.bluemirror.tech.