200 Million Rows of Data and Counting

At the University of Utah Hospitals and Clinics, the former cost-modeling effort was a bit of a crapshoot when it came to accuracy.


Two years ago, the University of Utah Hospitals and Clinics embarked on a Herculean data analytics task: the five-hospital system wanted to know its true service line costs. Like care delivery organizations across the industry, the academic medical center had only a rough idea of how much any given procedure cost, says Charlton Park, finance director.

Its former cost-modeling effort was a difficult undertaking. “We had allocated cost to the patient level, but it was based on cost-charge ratios,” Park says, meaning that bottom line expenses were stacked up against aggregate charge levels, or the full list-price of procedures. “At the macro level, it makes sense, because you’re allocating costs somewhere, but when you drill down to what we actually did and take into account supplies, pharmacy expenses, and lab costs at the visit level, you’re not seeing what something actually cost.” (Park will speak at Health Data Management’s upcoming Healthcare Analytics Symposium & Expo, July 15-17 in Chicago.”

That’s because the academic medical center’s charge list comprised numbers against which payers invariably applied major discounts and which patients rarely paid, even those without insurance. And the official charge amounts did not necessarily reflect the underlying cost of the good or service in question. Pharmaceutical mark-ups, for example, could vary widely depending on the drug, he says. “If you apply the same cost-charge ratio across all pharmaceuticals, you will be well over-costing some and under-costing others. If you apply the acquisition cost (of the pharmaceutical), what you actually paid for it, your cost calculations will be more accurate.”

To tackle the problem U-Utah convened a group of medical, I.T., and finance executives, which devised a patient visit cost allocation model for both facility and professional services. The model takes into account physician expenses, nursing and other salaries, pharmaceuticals, labs, supplies and a host of other factors that surround any given patient visit. The center identified nearly 900 business units, where patient care is delivered and where costs per given patient visit should be allocated. The allocation model does not include every piece of overhead, such as the cost of maintaining a finance department which serves the entire enterprise, but it does account for the labor and material costs that constitute the operating budget of each department. Each month, those costs are allocated to every patient visit that touched the department.

The analytics effort and data warehouse needed to support the cost-allocation effort were a complex undertaking, says Park, who will describe the project at the symposium. A homegrown data warehouse accepts data feeds from dozens of systems, including an Epic outpatient system, a Cerner inpatient record, a GE system in the operating room, and a PeopleSoft financial system. U-Utah has been delivering data reports for about one year, primarily to department chairs and administrators. Now every patient visit has costs allocated to it, which in turn may show up in different departments’ reports. A patient coming into the ED and winds up getting an emergency appendectomy, for example, would move from the ED, to the OR, perhaps the surgical ICU, and then a step-down unit before being discharged. All the costs from the different departments would be accounted for and allocated to that patient’s visit.

With 1.2 million annual visits in its inpatient and outpatient clinics to analyze, the resulting “data set is incredibly large,” Park says. “Close to 200 million rows of data, as wide as the eye can see. In the ED, you may get an IV, an aspirin, lab work, blood taken; every one of those charges triggers a cost allocation. Some are tiny, measured as cents, and some are very big. But every charge is on a different row and it is happening for every patient in the system.”

Of course, department heads do not need to pore though such massive amounts of data as the set-up enables self-service analytics and the capacity to generate department level cost reports, Park says. “Doctors can see how they compare to their peers at the average per-case level and drill down into their average pharmacy cost per visit, average nursing labor costs, and so forth.”

The analytics can provide insight into departmental performance, Park says. But its long-term value will be in negotiating better payer contracts that better reflect the actual cost of a given procedure. Park’s talk, “Analytics Tools to Determine Healthcare Delivery Costs and Value Driven Outcomes,” is set for Tuesday, July 16, at 4:15 p.m. For information about the event, visit www.healthdatamanagement.com/conferences/hcs/

 

More for you

Loading data for hdm_tax_topic #better-outcomes...