SPARC
Redesigning the Medicare Prospective Payment System for Skilled Nursing Facilities
Funded by: Centers for Medicare & Medicaid Services
Subcontract from: The Urban Institute
Project Period: 7/02 - 6/05
Objective: In 1998, Medicare instituted a prospective payment
system for nursing home care, a fixed price system that assigned patients to resource
utilization groups (RUGs) (similar to hospital DRG concept, but not diagnosis-based).
The system has been criticized for two primary reasons: 1) it fails to accurately pay
providers for use of nontherapy ancillaries (NTA) such as prescription drugs and
respiratory therapy; and 2) it creates incentives to deliver high levels of physical,
occupational, and speech therapy, regardless of patient need or clinical effectiveness.
Our objective was to create a payment model based upon clinically meaningful factors that
improves the payment for NTA costs and that eliminates the undesirable incentives for
therapy.
Methods: We utilized a unique data set that combined information
from Medicare skilled nursing facility (SNF) and hospital claims, the Minimum Data Set,
and cost report information to create approximately 600 variables for each of the 1.7
million SNF stays in 2001. First, three conceptual categories of patients were created:
Rehabilitation patients – patients that are admitted for functional recovery or
deconditioning and can benefit from substantial therapy; Acute – patients with acute
medical conditions, exacerbation of a chronic condition, or surgical intervention; and
Chronic – patients with illnesses that lead to multiple hospitalizations and SNF
admissions each year. Second, within each of these three categories, models were developed
that incorporate information from both the SNF stay and the qualifying hospital stay.
Two-stage models were built for selected components of NTA and therapy, using several
variable selection methods. The first stage identified the presence of any resource use
during the stay, and the second stage predicted the extent of the resource use. Third,
refined and simplified models then were adapted to create a proposed new payment system.
Results:We found significant clinical predictors of high costs,
and the models were successful at identifying high cost patients. Characteristics such
as parenteral IV, HIV status, IV use, and care in a hospital-based facility were associated
with high NTA costs. Scores in the Activities of Daily Living were significantly
associated with therapy costs. Hospital costs appeared to transfer to the nursing home,
and increased hospital drug and lab charges were associated with increased NTA costs while
increased hospital therapy and overall charges were associated with increased therapy costs.
Discussion: Centers for Medicare & Medicaid Services (CMS) has
acknowledged the problems with the current payment system and is looking at the possibility
of adjusting or overhauling it. We have presented them with a new payment option for
nursing home care, based more on clinically meaningful factors.
Current Project Status: Three reports to CMS have been submitted,
with a final Report to Congress to be delivered in January 2005.
Key Staff: Kramer, Eilertsen, Epstein, Min, Schlenker
v