The Workload Score Drops Exactly When Your Nurses Need It Most

Aug 13

3 min read

You've seen this shift. The floor is slammed, admits are stacking, two nurses are down due to a call-out, and everyone is running flat-out delivering care. And somewhere in your Epic instance, the workload score for that unit is falling.

Not because the work got lighter. Because no one had a free second to chart it.

If you're a CNE, CNO, or nursing operations leader, you already know this in your gut. You've looked at the numbers on a brutal day and thought, that's not what I saw out there. You're right — and there's a structural reason, one that isn't your build, your team, or your implementation.

The uncomfortable truth

The Epic workload score measures what nurses document, not what nurses do. When a shift gets busy, nurses do exactly what we train and trust them to do: they prioritize the patient over the keyboard. Charting gets compressed, delayed, or skipped — and the score drops at the precise moment true workload peaks.

As one enterprise nursing leader at a large multi-state health system put it: "When we're busy, we're delivering care, these numbers are inaccurate, and then we're planning off of inaccurate numbers".

It's not you — it's the design

This isn't a configuration miss or a change-management gap. It's an inherent limitation of any workload tool built on nurse-entered data. The peer-reviewed evidence is now hard to ignore:

OHSU researchers (Womack et al., 2021) found only a weak correlation (−0.09 to −0.23) between the EHR workload score and nurses' actual felt workload.
UCLA Health's own annual CareConnect validation showed 10–13% score mismatches per unit, traced directly to late or missing documentation.
A retrospective ICU cohort study found perceived workload was higher than the score suggested in 53% of cases.
Thate et al. (Applied Clinical Informatics, 2025) confirmed that structured flowsheet interfaces increase documentation burden and produce inaccurate data — with real implications for any AI built on top of it.

Your instincts on the floor have been more accurate than the dashboard.

The cost nobody budgets for

This isn't only a measurement problem. Inaccurate workload data drives the assignments that quietly burn nurses out — and turnover is where that shows up on the ledger. The average cost to replace a single RN runs about $60,090, against a national turnover rate near 17.6%. When the score underestimates a nurse's actual load from shift to shift, the system keeps asking more of the people already closest to the edge.

Protecting nurses and protecting the budget aren't competing goals here. They're the same goal.

You're not alone in this

Nearly every leader running Epic's workload framework has felt this — most just haven't named it this clearly yet. The systems that have run it longest, and even those who built thoughtful custom layers on top of it, keep hitting the same wall: the score can only see what got typed into the chart. This is a market-wide reality, not a local failing.

What comes next

The encouraging part: the nursing informatics literature is converging on a genuinely better path — workload captured passively from clinical orders, audit logs, and system interactions that accumulate automatically, with zero added documentation burden on the nurse.

One large academic health system offers a clear proof point. They have deployed a workload model built on 400+ clinical order signals pulled from Epic, refreshed every few hours, requiring no nurse charting at all. The workload gets measured without asking an already-stretched nurse to stop and account for how stretched they are.

That's the shift worth watching: from tools that add work to nurses, to intelligence that removes it.

A question worth sitting with

So here's what I'd ask you to consider: what would change for your nurses if your staffing data got more accurate on the hardest shifts — not less?

A handful of Epic health systems are already helping shape what comes next — pairing passive workload signals from the EHR with the qualitative side of the nurse experience. I'm not going to pitch it here. But if you lead nursing at an Epic system and this problem sounds familiar, I'd welcome a quiet conversation about what a better measurement layer could look like for your teams.

You've been right about this longer than the dashboards have.

— Greg Paulus, WonWay Health

Sources: Womack et al., OHSU (2021); Chen et al., Journal of Nursing Scholarship (2024); Thate et al., Applied Clinical Informatics (2025); UCLA Health CareConnect validation data.