Testimonials are broken: Why healthcare’s last bottleneck still lacks innovation
An 82-year-old stroke patient was hospitalized in an acute care bed at a cost of $2,000 a night, where she remained for six days after being discharged because no one could determine which skilled nursing facilities had empty beds, accepted Medicaid and offered stroke rehabilitation services. The discharge planner has made 23 calls and 14 faxes so far, but has received zero responses. This is not an outlier; This is a $150 billion referral problem in health care. We kept throwing technology at it without fixing the underlying architecture.
Numbers tell stories
U.S. clinicians make more than 100 million specialty referrals each year; however, studies show that 50% of these referrals are never completed. Things get even worse for post-acute care placements; the length of stay for patients waiting to be discharged to post-acute care increased by 24% between 2019 and 2022. In Massachusetts, one in seven medical-surgical beds is occupied by patients who no longer need acute care but have nowhere to house them.
The economic impact is devastating—health care systems lose 10-30% of revenue due to referral leakage and patients seeking care outside their network. This represents an annual revenue loss of $821,000 to $971,000 per physician. California hospitals report that boarding patients as they prepare to be discharged costs the state $2.9 billion annually. Additionally, by 2024, more than 75% of North American healthcare providers will still rely on fax machines for referrals.
Three structural failure technologies cannot be repaired individually
Why “AI-driven” solutions keep failing
The typical approach treats AI as an add-on. There's OCR for scanning paper recommendations, autofill widgets for EHR fields, and predictive algorithms for risk scoring. These tools all solve a small problem but ignore the macro disaster. Not surprisingly, more tools often create more physical work and cause fatigue rather than relief.
The global patient referral management software market will reach US$16.14 billion in 2025 and is expected to reach US$67.92 billion by 2034. Although 87% of hospital executives said referral leakage is a top priority, 23% have no plans to monitor it. What's obviously missing? AI solves the real coordination problem, the workflow gap between sending referrals and seeing patients.
What actually needs to be built
Effective recommendation innovation treats recommendations as a constrained optimization problem. This will involve matching patients with specific clinical needs, insurance requirements and geographic restrictions with providers who can meet their needs with real-time, two-way confirmation.
Recent market analysis shows that 40% of healthcare organizations have adopted predictive analytics for provider matching, with real-time referral tracking dashboards increasing processing efficiency by 45% while reducing patient misses by 30%.
Privacy-preserving initial matching
Currently, coordinating referrals means sending complete medical records before anyone has confirmed capacity. This creates regulatory friction and slows everything down. A smarter approach would be to first match the anonymity criteria, “Stroke patients within 10 miles who need PT, Medicaid coverage,” and only share personally identifiable information once common interest is confirmed. AI-driven solutions can consolidate siled data while maintaining privacy during the matching stage.
Real-time status visibility
The recommendation black hole exists because no one knows what will happen after a recommendation is sent. Improved referral coordination should function like package tracking, with both sender and recipient seeing the same timeline. This real-time tracking will help health organizations improve processing efficiency. It's not technically complex; we do it to deliver food, but it requires breaking down information silos.
results-oriented learning
Current recommendation systems do not have memory retention capabilities. If a facility accepts a referral but the patient is readmitted within 30 days, the facility should be ranked lower in future competitions. Research shows that referral leakage can be reduced by up to 60% through AI-enhanced workflows that include outcome tracking. Smart systems will track readmission rates, wait times and patient satisfaction, then adjust recommendations accordingly.
Neutral infrastructure without vendor lock-in
The problem of fragmentation cannot be solved by tools that only work within Epic or only cover Medicare patients. What is needed is an effective referral infrastructure with universal access regardless of EHR vendor or payer, real-time data exchange, minimal barriers to entry, and transparent quality metrics.
An Inconvenient Truth
Contrary thinking – recommendations are still broken, not because of technical incompetence, but because people with the ability to fix them benefit from keeping them broken. Health systems make money by preventing outward leakage, not by fixing referral black holes. EHR vendors sell expensive modules, locking customers in, while payers negotiate network exclusivity, limiting choice. A recommended leakage rate of 55-65% incurs consulting fees, software licenses, and internal planning. Everyone optimizes against their own metrics, while patients and frontline coordinators suffer the consequences.
what happens next
Techniques to fix the recommendations are currently being deployed. Currently, technologies to fix recommendations are being deployed, but most implementations are still in the pilot phase. AI-enabled referral systems show significant reductions in processing times, faster authorization turnaround, and significant reductions in referral leakage.
Prescription delivery became automated in the 2000s. Laboratory ordering became automated in the 2010s. We've automated everything but the one workflow that most directly determines whether a patient actually gets the care they need. Understanding the gap between effective approaches and large-scale implementation reveals the real problem, which is that healthcare views referrals as an administrative burden to manage rather than a critical workflow that needs to be optimized.
Every day we wait, patients pay the price: acute beds are occupied unnecessarily, specialist appointments never happen, and families print out facility names on the phone tree. For more than a decade, the data has been clear. The technology is ready. The question is whether we want to ultimately fix the pipes rather than slap on another Band-Aid.
Photo: porcorex, Getty Images
Naheem Noah is a PhD student in computer science at the University of Denver's Ritchie School of Engineering and Computer Science, where his research spans privacy-preserving systems, security, artificial intelligence, and healthcare coordination.
As co-founder and CEO of Carenector, Naheem is transforming research into practice by building an AI-driven referral infrastructure for patients and care organizations. Carenector operates a real-time consumer platform to help families understand post-acute care options while establishing an institutional coordination platform with care facilities to address referral breakdowns across the continuum. The platform uses privacy-preserving matching, real-time status tracking, and outcome-informed learning to address coordination gaps.
This article appeared in Medical City Influencers program. Anyone can share their thoughts on healthcare business and innovation on MedCity News through MedCity Influencers. Click here to learn how.
