Physical rehabilitation following Medicare prospective payment for skilled nursing facilities

Health Services Research, Oct, 2004 by Walter P. Wodchis

This research was supported by a Centers for Medicare and Medicaid Services dissertation fellowship grant. The author wishes to thank dissertation committee members B. E. Fries, R. A. Hirth, J. B. Cullen, and J. P. Hirdes, as well as two anonymous reviewers.

NOTES

(1.) Grimaldi (1999, 2002) provides a thorough review of RUG-III and the Medicare PPS payment system.

(2.) For example, under Medicare case-mix payment in 1999, the payment rate for residents in the highest rehabilitation category was $142.32 for the nursing component and $186.01 for the therapy component (Health Care Financing Administration 1998).

(3.) Although excess demand for public-pay (particularly Medicaid) residents has given facilities substantial power over the severity of public-pay total patient mix historically, more recent evidence points to excess capacity and thus less opportunity for arbitrary selection of patients (Grabowski 2001; Harrington et al. 2000).

(4.) Several ranges for the dependent variable are explored including (1) exact levels of nodal therapy (to the minute), (2) therapy levels within 10 percent and within 15 percent of nodal levels, and (3) therapy levels within 5, 10, and 15 minutes of nodal levels. These models are tested both including zero and not including zero as a nodal level of therapy, and using only Medicare residents. Nearly all of these models provide results with higher AOR estimates than that for Table 3 for the effect of PPS for Medicare residents. The lowest AOR is 1.31 (C.I. 1.18-1.45), and is found for therapy within 10 percent of nodes and including zero in the flail sample; the highest AOR is 1.99 (C.I. 1.77-2.23), and is found in the model using exact nodes of therapy not including zero in the full sample. In all specifications, parameter estimates for all other payers are consistent except for Medicaid payment, which is positively associated with nodal levels of therapy any time zero is included as a nodal level of therapy.

(5.) Additional models examined whether the included diagnoses and conditions adequately account for resident need for rehabilitation. Two additional models are run for each analysis using constructed rehabilitation potential variables. These models do not change any estimates presented here. The results presented in Table 4 are not sensitive to unobserved resident rehabilitation potential and the estimates generally provide a lower bound on the effect size.

REFERENCES

Ai, C., and E. C. Norton. 2003. "Interaction Terms in Logit and Probit Models." Economics Letters 80 (1): 123-9.

American Nurses Association. 2000. "Prospective Payment System for Long Term Care" [accessed on January 12, 2003]. Available at http://www.nursingworld. org/mods/archive/mod90/kcfull.htm.

Centers for Medicare and Medicaid Services. 2002. "National Health Care Expenditures: Table 7" [accessed on July 4, 2002]. Available at http://www.cms. hhs.gov/statistics/nhe/historical/t7.asp.

Coburn, A. F., R. Fortinsky, C. McGuire, and T. P. McDonald. 1993. "Effect of Prospective Reimbursement on Nursing Home Costs." Health Services Research 28 (1): 45-68.