The ambulatory intelligence gap

The Ambulatory Intelligence Gap | Databricks Blog Skip to main content

Summary

The gap: Ambulatory growth stalls when access, referral, capacity, and financial data sit in disconnected systems.

The solution: Health Catalyst's Ambulatory Intelligence combines AI with nearly two decades of healthcare improvement expertise to explain what's driving the numbers and where to act.

The outcome: Prebuilt metrics deliver same-week visibility and real results.

Ambulatory care, the outpatient clinics and physician practices where most patients interact with a health system, is where growth is won or lost. Yet many health systems struggle with the operational constraints that limit access to care: provider capacity, referral retention, and downstream financial performance. Access is moving in the wrong direction. A new patient trying to book a routine appointment waits an average of 31 days to be seen, a number that has climbed 19 percent since 2022, according to  AMN Healthcare's 2025 survey of physician access. Referrals are another source of pressure. U.S. health systems lose an estimated $150 billion annually to referral leakage, according to  ReferralMD , while  MGMA's 2025 Referral Coordination Benchmarking Study found that 38 percent of referrals never close the loop. For ambulatory leaders, these aren't isolated problems. Access, provider capacity, referral retention, panel management, and financial performance are deeply interconnected. The challenge isn't recognizing the problems as much as it is understanding where constraints in the network exist, how decisions in one area affect outcomes in another, and where to focus first. Robbie Hughes, Chief Product Officer at Health Catalyst, sees the same pattern across the health systems his teams work with: "Every day, patients wait weeks for an appointment only to find their doctor isn't taking new patients. Referrals get lost. Providers are busy, but they still can't make the economics work," Hughes said. "It isn't because they don't care. The information they need to improve is scattered across five or six disconnected systems, and by the time it's pulled together, it's already out of date." The Gap Between Data and Action "The time from understanding you have a problem to actually doing something about it can be many months," Hughes said. "That isn't a lack of data. As often as not, it's a lack of alignment around what the data means. Everyone has dashboards. Everyone has an opinion.” The usual toolkit doesn't close that gap: EHR reports, a Tableau view sitting on a stale extract, a consultant engagement that produces a snapshot and then expires. Each one answers a question once at the time of extract. None of them produces a single, current picture that an operations leader and a finance leader can look at together, trust, and agree on. Even when organizations have confidence in the numbers they are seeing, they often struggle to understand the operational drivers behind them. A long wait time may indicate insufficient provider capacity but may also be the result of  referral bottlenecks, misaligned scheduling practices, or uneven panel distribution. Lower provider productivity may signal excess capacity in one clinic and access constraints in another. The same metric can point to a variety of very different underlying problems. That complexity is what creates the ambulatory intelligence gap. Access, provider productivity, referrals, panel management, and financial performance do not operate independently. They influence one another. Improving one area can shift constraints to another. Understanding what is happening is important. Understanding why it is happening, what is driving it, and where intervention will have the greatest impact is what enables improvement. Turning Expertise Into Software Recent advances in AI have expanded what's possible. Today’s AI models can now analyze complex operational relationships, surface patterns across large volumes of data, and rationalize over information in ways that were previously difficult or impractical. But AI alone doesn't solve the problem. AI can identify patterns in data, but it doesn't inherently understand ambulatory operations and the complex interrelations between components in the network. It doesn't understand which interventions have historically improved access, increased provider utilization, or reduced referral leakage. Those answers require operational context and improvement expertise, something that’s long been the domain of long-tenured experts in the field. This is where Health Catalyst's approach differs. Health Catalyst’s Ambulatory Intelligence solution combines modern AI capabilities with nearly two decades of healthcare improvement expertise, embedding healthcare-specific knowledge, operational best practices, and proven improvement methodologies directly into the solution. But connecting an organization’s sensitive patient data to external resources has always been problematic for healthcare providers. For this reason, Health Catalyst deploys its AI solution directly into the customer's own Databricks workspace. Health Catalyst brings its expertise to the data, ensuring it never leaves the health system's environment. That choice follows a shift Hughes has watched accelerate. Customers increasingly want their data hosted on their own lakehouse, he says, and in the age of AI, they are focused on governance and control. They don't want it, as he puts it, "used as a training ground for someone else to derive value they never consented to." Brian Eliason, who leads strategic vendor partnerships at Health Catalyst, says the market made the direction obvious: "Customers have shifted and want to control their own data. But they've also told us they want to continue the relationship with Health Catalyst, with them owning it. That's why we moved forward with the shipped model." Built for the Customer's Environment Getting there took real engineering. The medallion architecture, Hughes says, let the company turn its services expertise into a common semantic layer, doing "in weeks what would previously have taken months or years." Building inside the customer's environment required balancing governance, performance, and usability. Health Catalyst's solution relies on several Databricks components, each serving a specific...