The Case for a Modern Nursing Valuation Framework
Part of Nurse.org’s Nursing AI Watch, our ongoing investigation into how artificial intelligence is reshaping nursing practice. In his first essay, nurse and clinical AI strategist Mark Smith asked how healthcare learned to value software before nursing. In this second of three, he goes looking for the answer in the research record, and finds the problem was nearly solved forty years ago.
Two patients can have the same diagnosis.
Same hospital. Same unit. Same DRG.
On paper, they can look almost identical.
At the bedside, they may be nothing alike.
One gets up, takes the medications, understands the discharge plan and goes home.
The other is confused. Unsteady. Something about the breathing changes. The nurse notices it, reassesses, watches a little longer, calls, comes back, checks again.
Maybe the plan changes.
Maybe it doesn’t.
That second possibility matters.
A nurse can absorb new information, reconsider what is happening and decide the current plan is still right.
Clinical judgment occurred.
The financial record barely noticed.
Healthcare knows what both nurses cost.
It has a much harder time describing what they contributed differently to each patient.
That was where I thought this article was going.
Then the history got in the way.
Want to see more Nurse.org articles in your Google results? Add us as a preferred source.
After the first article was published, John Welton, PhD, RN, FAAN pointed me back toward some of the earlier work on nursing intensity.
I expected to find a few conceptual papers.
Instead, I found researchers measuring nursing care patient by patient.
In a New Jersey study conducted between 1979 and 1981, nursing activity was recorded in actual minutes for individual patients, by shift and by skill level. Researchers built a case-sensitive measure of nursing resource use and used it to allocate nursing costs by patient rather than simply by hospital day. (Caterinicchio, 1984)
That is not a vague attempt to prove nursing matters.
It is costing.
It is measurement.
It is patient-level resource allocation.
And it was happening more than forty years ago.
The researchers were blunt about the problem. A flat patient-day charge was too crude because patients do not consume the same nursing resources every day. Nursing demand changes with illness, procedures, recovery and the clinical problems that appear along the way. (Caterinicchio, 1984)
In other words, the basic assumption underneath nursing finance was already being challenged:
A hospital day is not a unit of nursing care.
The early DRG work makes the story even more interesting.
The Yale case-mix accounting model did not simply lump everything together. It separated nursing accounts from hotel costs and ancillary services. Nursing was treated as its own cost category inside the architecture used to understand hospital resource consumption. (Thompson et al., 1979)
By the early 1990s, John Thompson and Donna Diers were synthesizing a substantial body of work on nursing intensity, patient classification, staffing and nursing cost. Their chapter shows that the field was not merely asking whether nursing intensity existed. It was asking how to measure it well enough to use for staffing, costing and eventually payment. (Thompson & Diers, 1991)
That changed my understanding of the problem.
The history of nursing valuation is not a story of an idea that failed. It is a story of an idea that never became infrastructure.
That distinction matters.
Because it means healthcare did not simply overlook nursing.
People were building the machinery to see it.
By the late 1980s, the question had moved beyond whether nursing intensity could be measured.
Researchers were asking what would happen if the payment system actually used it.
A Yale study supported by the Health Care Financing Administration tested several ways of weighting DRGs for nursing intensity. One approach relied on experienced clinical nurses to group DRGs according to the nursing care they required and then estimate the time associated with those groups.
According to Donna Diers and Janis Bozzo, the methods performed well against more complicated approaches.
But they created another problem.
Every method tested would have increased Medicare program costs.
The model was not pursued for payment. (Diers & Bozzo, 1997)
That changes the history.
Nursing intensity was not simply sitting outside the payment system because nobody could figure out how to measure it.
Making nursing visible had a price.
A year later, Jerry Cromwell and Kurt Price approached the question from another direction. They tested how much DRG weights would change if nursing intensity were incorporated into them.
Under the narrowest assumptions, the answer looked small.
When the adjustment was applied primarily to direct nursing salaries, most DRG weights changed very little. That could have made nursing intensity look economically unimportant.
But the result depended heavily on what counted as nursing cost.
When broader routine costs were included, the effect became larger. Cromwell and Price also left unresolved which nursing classification system should be used, what cost base should be adjusted, and whether some DRGs required more patient-specific treatment. (Cromwell & Price, 1988)
So, the problem had changed again.
It was no longer:
Can nursing be measured?
It was:
Measured how?
Against which costs?
At what level of detail?
And what happens to payment once we do?
Those are not measurement questions alone.
They are financial choices.
That distinction matters because it helps explain how an idea can be recognized, studied and technically workable without ever becoming part of the infrastructure that follows.
The research had made nursing more visible.
The payment system now had to decide what to do with what it could see.
And at least in one federally supported effort, seeing more of nursing meant paying more for it.
There is another problem buried in those early models.
Much of the work operated at the DRG level.
But nursing happens at the patient level.
Two people with the same DRG can consume very different amounts of nursing care.
That difference becomes easy to lose once everyone is averaged together.
Welton’s later work makes that visible.
In a 2006 study of 12 medical-surgical units, nursing intensity was recorded for individual patients in a computerized database. The distribution of daily nursing cost was broad and skewed, not clustered neatly around an average. Mean direct nursing cost was about $429 per day, but the standard deviation was roughly $160. Across units, average nursing intensity ranged from 8.1 to 19.9 hours per day, with estimated direct nursing costs ranging from $255 to $619.
The billing system did not reflect that variation very well.
One cardiac step-down unit generated a large share of higher intermediate-care charges despite nursing costs below the overall mean. Another unit had some of the highest nursing costs but relatively little use of the higher billing category. (Welton et al., 2006)
The room charge saw location.
The nursing data saw work.
They were not telling the same story.
The problem isn’t that nursing leaves no data behind. The problem is that the data rarely follows nursing into the room where money gets allocated.
There is a trap here.
Suppose tomorrow every hospital could measure nursing intensity perfectly.
Then what?
A dashboard is not a valuation model.
And counting more things does not necessarily get us closer to value.
A medication administration is easy to count.
So is an assessment.
A dressing change.
An hour worked.
But nursing also happens between those events.
Watching.
Connecting.
Reconsidering.
Recognizing that two small changes suddenly mean something together.
Sometimes the result is an intervention.
Sometimes the result is that nothing bad happens.
No fall.
No ICU transfer.
No rapid response.
No readmission.
Healthcare is much better at counting the complication than the complication that never occurred.
That does not mean every prevented event belongs to nursing. Care is delivered by teams, and attribution is messy.
But healthcare already accepts messy attribution when evaluating technology.
A software system rarely creates an outcome by itself.
Clinicians, workflows, patients and operations all matter.
Yet the technology still gets an ROI.
Nursing gets a labor cost.
Now imagine an AI tool saves a nurse twenty minutes.
The business case will capture those twenty minutes.
But what happens next?
Does the nurse spend them watching a deteriorating patient?
Teaching a family?
Coordinating a discharge?
Taking another patient?
Or does the organization simply remove twenty minutes of nursing capacity because the software made the workflow more efficient?
Those are not the same outcomes.
On a spreadsheet, they can look exactly the same.
If we can measure the minutes AI gives back but not the nursing value created with those minutes, the technology will always win the spreadsheet.
The software has an ROI.
The nurse has a labor cost.
The irony is that nursing has been measurable for decades.
Researchers measured patient-level nursing activity in the early 1980s.
They tested nursing-intensity adjustments against DRGs. They built classification systems.
They studied cost.
They kept refining the methods.
The work never became a durable national financial infrastructure.
That is the problem this piece leaves us with.
Not whether nursing can be measured.
It can.
Not whether nursing intensity varies.
It does.
The harder question is whether hospitals are willing to connect what they already know about nursing to the decisions that determine what nursing is worth.
They do not have to wait for Medicare to begin.
And neither do nurse leaders.
That is where part three begins.
More from Nurse.org’s reimbursement series:
🤔If an AI tool gave you back twenty minutes on your next shift, what would you spend them on, and would your hospital’s spreadsheet ever know? Tell us in the comments below.
References
- Caterinicchio, R. P. (1984). Relative intensity measures: Pricing the inpatient nursing services under diagnosis-related group prospective hospital payment. Health Care Financing Review, 6(1), 61–70.
- Cromwell, J., & Price, K. F. (1988). The sensitivity of DRG weights to variation in nursing intensity. Nursing Economics, 6(1), 18–26.
- Diers, D., & Bozzo, J. (1997). Nursing resource definition in DRGs. Nursing Economics, 15(3), 124–130.
- Thompson, J. D., Averill, R. F., & Fetter, R. B. (1979). Planning, budgeting, and controlling—one look at the future: Case-mix cost accounting. Health Services Research, 14(2), 111–125.
- Thompson, J. D., & Diers, D. (1991). Nursing resources. In R. B. Fetter, D. F. Brand, & D. Gamache (Eds.), DRGs: Their design and development (pp. 121–183). Health Administration Press.
- Welton, J. M., Fischer, M. H., DeGrace, S., & Zone-Smith, L. (2006). Hospital nursing costs, billing, and reimbursement. Nursing Economics, 24(5), 239–245, 262.
If you have a nursing news story that deserves to be heard, we want to amplify it to our massive community of millions of nurses! Get your story in front of Nurse.org Editors now – click here to fill out our quick submission form today!
-
Published on
August 12, 2026
Written by
Mark Smith, MBA, BSN, RN Critical-Care Nurse | Nursing Value, Clinical Judgment & AI | Research & Advisory
