The health-system utilization forecasting conundrum: one size definitely doesn't fit all.

Recently we’ve been chatting with several health system strategy leaders who seem to be dealing with a similar problem: the forecasting products they license are polished and expensive but increasingly challenged to be useful at the local market level where real decisions are made. If one scans across the national-level estimates from the usual major health system forecasting providers, many of these forecasts are broadly fine. They capture the big trends, line up the conventional wisdom, and produce something that looks credible in a board pack. But a forecast can be both perfectly respectable in aggregate and completely whiff-it in a specific market depending on some of the ground level forces at play. That can be a bit of a problem, if you’re using intel to help drive capital projects, service line expansion, physician recruitment, bed planning, ambulatory strategy, or trying to unpack where future demand is likely to land.

The challenge, as we are increasingly hearing from strategy and service line leaders, is straightforward. A fair number of third-party models lean heavily on national utilization rates and standardized assumptions about how care will shift over time. They appear to assume that market dynamics are similar enough from one region to the next that broad adjustment factors will do the job.  But what seems to be emerging in our conversations is just how often this approach has been missing the mark in recent years. The worst examples find some trying to plan with forecasts that are directionally inconsistent with where their market has been trending.

Urban and rural markets behave differently. Regions with very different payer mixes behave differently. Affluent suburbs with strong preventive health uptake behave differently from poorer communities where people delay care, struggle with transport, and bounce in and out of coverage among many contributors too poorer overall health. Use rates vary by race, by socioeconomic status, by access, by local provider supply, and by how fractured the insurance picture has become. When you’re boots on the ground in those markets none of that stuff is marginal, it goes straight to the heart of your demand equation.

By the way, such a challenge hasn’t been the sole domain of health system strategists.  Take a glance at the recent payer MLR trends and you’ll quickly come to the realization that actuarial modelers at health insurers haven’t been able to easily figure it out either. These are smart people, with sophisticated analytical approaches and robust data at their disposal, whose job is to understand utilization inside and out. Forecasting is an inexact science at best.  Moving targets, controlled approximation, call it what you will. But utilization demand forecasting is often grounded in historical data, with future assumptions, and probabilities that don’t cleanly fit future state timelines in all markets.

Here’s one concrete example that we chatted about with a health system strategic planner: Outpatient shift has been a blanket assumption that many forecasters are driving at hard as they prognosticate absolute volume decline. Yes, there has been a long-run migration in that direction. Yes, technology and reimbursement models have enabled (or incentivized, or both) care moving out of the acute care setting. Yes, GLP-1s may ultimately cannibalize some “would-have-been” procedures. Yes, the IPO list got unceremoniously dumped by CMS after roughly a quarter century.  But some planners feel they’re being handed forecasts that treat this as a dependable path in every market, when that plainly is not how their world looks. In just the last fortnight, we’ve been shown data from three health systems who cite a perfect storm of rising comorbidity burden, worsening obesity rates, deferred care, and limited access to primary and preventive services driving inpatient use to double digit growth rates in some of their sub-markets for a substantial stretch. And that includes a favorable procedural mix, where surgeons are increasingly reluctant to do a procedure in an ASC for fear of patient complications requiring hospital level care. 

And the concern about blind spots doesn’t end there. Our discussions with health system leaders reveal concern about the robustness of current forecasting assumptions for tomorrow, with several calling out off-the-shelf forecasts as “the bleeding obvious”, meaning they’re much better at extending visible, in-vivo trends than they are at spotting what could materially shift demand – and consequently health system strategy - five or ten years down the track. Drug pipelines, device development, clinical trial signals, and shifts in consumer health behavior all matter here, along with the interactive effects with known forces. Are the real signals getting picked up early enough?  Semaglutide is an obvious example that is starting to hit home, but for many wasn’t a consideration until recently. And yet, human trial results were available years before many strategic planning models treated it as something worth worrying about. There will be more examples like that in the years to come.

So, health systems need to push their forecasting vendors harder. Which assumptions are national defaults? Which are calibrated to consider factors that actually matter? Local disease burden, local payer friction, local access barriers, and local patterns of care? Are there scenario ranges for markets that diverge from the neat version of outpatient migration? Are there analyses of whom in the market is getting healthier, who is getting sicker, and who is being priced out of timely care?  

Remember this: “Just take off these filters and it’ll get the numbers closer” is an incredibly poor substitute for “This submarket is likely to behave differently, due to these specific factors. And here’s how we’d recommend approaching it”. What we're describing isn't a particularly novel realization; the imperfect fit between a generic set of assumptions feeding a national model and one that fits any given market has been something health systems have wrestled with before. But it's one that takes on a new degree of importance in an environment where margins are especially thin, as health systems allocate resources to plan for an increasingly uncertain future. To be clear, vendor forecasts still have value as a baseline. They can be a useful starting point that save some "grunt work" by combining a few data variables together. But as health system leaders have told us, rarely are they fit-for-purpose right out of the box. Planners should be ready to go deep on exactly what any model contains, prepare to cross examine scenarios (and the researchers who built it), pressure-test the relevance of assumptions for their market and weave their own view of their population into the projections. 

Union Healthcare Insight works with health system leadership to cut through the noise — bringing clarity and genuine insight into the decisions that matter most. If the challenges outlined here resonate with what your team is navigating, we’d welcome a conversation. We've scheduled a series of Strategy Dinners throughout 2026 — intimate, off-the-record sessions with senior health system executives working through exactly these kinds of issues.

Go here if you'd like to register for one of our upcoming strategy dinners in a city near you. Or you can reach out to us directly at Info@unionhealthcareinsight.com, we’d love to connect with you!