Incremental Cost of Postacute Care in Nursing Homes

doi:10.1111/j.1475-6773.2010.01189.x

Health Serv Res. 2011 February; 46(1 Pt 1): 105–119.

doi: 10.1111/j.1475-6773.2010.01189.x

PMCID: PMC3015022

NIHMSID: NIHMS234196

Incremental Cost of Postacute Care in Nursing Homes

William D Spector, Maria Rhona Limcangco, Heather Ladd, and Dana A Mukamel

Agency for Healthcare Research & Quality, 540 Gaither Rd., Rockville, MD 20850

Social & Scientific Systems Inc., Silver Spring, MD

Academy, Health Policy Research, University of California, Irvine, CA

Address correspondence to William D. Spector, Ph.D., Senior Social Scientist, Agency for Healthcare Research & Quality, 540 Gaither Rd., Rockville, MD 20850; e-mail: william.spector/at/ahrq.hhs.gov. Maria Rhona Limcangco, Ph.D., Analyst, is with the Social & Scientific Systems Inc., Silver Spring, MD. Heather Ladd, M.S., Research Associate, and Dana Mukamel, Ph.D., Professor and Senior Fellow, are with the Academy, Health Policy Research, University of California, Irvine, CA.

Abstract

Objectives

To determine whether the case mix index (CMI) based on the 53-Resource Utilization Groups (RUGs) captures all the cross-sectional variation in nursing home (NH) costs or whether NHs that have a higher percent of Medicare skilled care days (%SKILLED) have additional costs.

Data and Sample

Nine hundred and eighty-eight NHs in California in 2005. Data are from Medicaid cost reports, the Minimum Data Set, and the Economic Census.

Research Design

We estimate hybrid cost functions, which include in addition to outputs, case mix, ownership, wages, and %SKILLED. Two-stage least-square (2SLS) analysis was used to deal with the potential endogeneity of %SKILLED and CMI.

Results

On average 11 percent of NHs days were due to skilled care. Based on the 2SLS model, %SKILLED is associated with costs even when controlling for CMI. The marginal cost of a one percentage point increase in %SKILLED is estimated at U.S.$70,474 or about 1.2 percent of annual costs for the average cost facility. Subanalyses show that the increase in costs is mainly due to additional expenses for nontherapy ancillaries and rehabilitation.

Conclusion

The 53-RUGs case mix does not account completely for all the variation in actual costs of care for postacute patients in NHs.

Keywords: Health care costs, Medicare, instrumental variables

Postacute patients are increasingly being treated in nursing homes (NHs). Spending has risen every year since 2000 (Medicare Payment Advisory Commission [MedPAC] 2008). They are typically treated for conditions such as knee and hip replacements, fractures, strokes, and heart failure (MedPAC 2007).

The percentage of NH care funded by Medicare has increased nationally from 8.6 percent in 1999 to 13.1 percent in 2005 (Harrington, Swan, and Carrillo 2007). Medicaid pays about 43 percent of NH care with the remainder mostly private pay (Kaiser Commission on Medicaid and the Uninsured 2009).

Medicare and 35 State Medicaid programs use prospective systems based on case mix for NH reimbursement. This approach provides a fixed amount, typically set at the average cost for each patient group considered to be clinically homogenous. State Medicaid programs may also include other features in attempting to align payment and costs. Features include add-ons for certain types of residents, cost ceilings and floors on specific cost centers, and incentives for direct care spending (Schlenker 1986; Feng et al. 2006; Rudder, Mollot, and Mathuria 2009;). NHs with managed care contracts use a variety of reimbursement methods, including case mix reimbursement.

The Medicare payment system pays a fixed amount for predefined patient groups. With this approach, payment needs to be well aligned with the cost of caring for patients in these groups. If payment rates for each patient group are not aligned appropriately with the cost of each group, perverse incentives arise, leading to both access and quality issues. Some patients may be less profitable than others and some may represent net losses. These patients may face difficulty gaining access to NH care because they are financially less attractive. For those that represent net losses, once admitted, facilities may have difficulty meeting their clinical needs.

Medicare pays only for postacute, skilled care NH patients defined as those having had at least a 3-day hospital stay for medically necessary inpatient hospital care, and who require daily skilled care or rehabilitation services (Centers for Medicare and Medicaid [CMS] 2008). The daily rate depends on the care needs of the patient as measured by the Resource Utilization Groups (RUGs), a case mix index (CMI) for NHs, that is expected to cover operating and capital costs (MedPAC 2009b). A person is classified into a RUG based on expected minutes of therapy, activities of daily living, need for special services, and certain clinical conditions. There were 44 RUGs during the 1998–2005 period, which were expanded to 53 categories in 2006. Each RUGs category has a nursing and a rehabilitation weight that was derived from NH staff time studies performed during the 1990s. There is an “other” component that covers room and board and capital costs. The minimum data set (MDS) assessment is used to determine the RUG for each patient. The increase in the number of groups used for Medicare from 44 to 53 was an attempt to better capture the variation in nontherapy ancillary (NTA) service costs (such as drugs, laboratory expenses, and respiratory services) of patients classified into high therapy and extensive services groups (CMS 2005). A new RUGs, RUG-IV, being proposed for FY2011, would make adjustments to the groupings based on more recent time and motion studies and would increase the number of groups to 66 and expand the number of rehabilitation, special care, and complex care groups (CMS 2009).

The RUGs classification system has been criticized for having a number of inadequacies. Analyses of the explanatory power of the 44 categories RUGs shows that it explains about 40–55 percent of staff time costs for all NH residents, with the higher estimate when NTA costs are not included (Fries et al. 1994; White, Pizer, and White 2002;). The MedPAC criticizes the RUGs classification because it was developed from staff time studies, and it argues that the 53 category index still does not reflect much of the additional costs of NTAs because these are not strongly related to staff time. MedPAC suggests ways to improve diagnostic information using information to classify patients from the prior hospitalization, to improve reimbursement accuracy by developing a new method for reimbursing NTAs and therapies that would use patient and NH stay characteristics that reflect cost differences, and to implement an outlier payment system for high NTA and therapy costs (MedPAC 2008).

Analysis of NH cost reports can provide insights into how well the RUGs classification system explains the variation in NH costs. The annual costs of care at the facility level depends primarily on the case mix of its residents, the number of days of care provided, and the wages that the NH pays to provide that care. NHs that admit a greater percentage of postacute patients will generally have a higher overall case mix because postacute patients require more complex medical care and services. To the extent that the RUGs reflect the true cost burden of postacute care, the costs of NHs that provide a higher percentage of postacute care will be appropriately reimbursed. However, if the RUGs do not sufficiently capture the costs for all types of postacute care, then the percentage of postacute care provided by the facility will also explain some of the costs.

In this paper, we analyze whether the RUGs CMI sufficiently capture the cost burden of postacute patients. We estimate cost functions that include in addition to the RUGs CMI, inpatient days, ownership, and wage index, the percent of days due to Medicare skilled care days (%SKILLED). If costs are higher when facilities have a higher %SKILLED, even when controlling for RUGs (using the 53 RUGS CMI), then this suggests that the current RUGs do not sufficiently capture differences in costs of postacute patients. In addition to the impact on total costs, we also estimate the impact on the two cost categories that are expected to be affected, as argued by MedPAC—rehabilitation and NTAs.

METHODS

Sample and Data

We obtained 2005 Medicaid cost reports from the California Office of Statewide Health Planning and Development. These are annual financial reports that are mandated, audited, and used by the state to set Medicaid payment rates for all skilled NHs in the state. They include information about expenditures, wages, and outputs (e.g., inpatients days and admissions).

The 2005 MDS data were used to calculate the RUGs CMI for each NH. The MDS is an individual-level dataset with information about all residents in the facility, with demographics, physical and mental health status, and information about specific treatments. MDS data are collected by NHs upon admission and at specific time intervals following admission (e.g., every 90 days for long-term care patients and 5, 14, and 30 days for skilled care patients). Data collection is mandated by the CMS.

We used additional data sources for the instrumental variables (IVs). We used the 2002 Economic Census, Sector 62 to obtain revenue data of rehabilitation establishments by zip code. Driving time between NHs and community hospitals was calculated using Google map. A list of hospitals and their addresses was obtained from the 2005 California Healthcare Cost and Utilization Project State Inpatient Database (HCUP Databases 2005).

There were 1,083 free-standing skilled nursing facilities in California in 2005. We excluded 51 facilities because of missing case mix data and 42 because they were atypical, providing care to special populations such as subacute pediatrics and residents with developmental and psychiatric disorders. In addition we removed one outlier based on the Cook's D statistic that had implausible values. The final sample includes 988 free-standing skilled nursing facilities.

Variables

The dependent variable was total facility expenditures. Because some NHs produce other services in addition to inpatient days, such as home care, outpatient clinics, or day care visits, but the cost reports do not report these costs separately, inpatient days were adjusted upwards by the ratio of outpatient to inpatient revenues to account for the higher output level in facilities producing these services. To measure the proportion of postacute care provided by the NH, we used skilled nursing care days from the cost reports and calculated the percent of days due to skilled care days as the ratio of unadjusted skilled care days to inpatient days × 100.

A facility-level CMI was calculated using the 2005 MDS data. A resident assessment was assigned to one of 53 case mix groups using the RUGs grouper program v5.20 and the Index Maximized grouping option. We included all assessments that have the required items to calculate RUGs—admission, annual, significant change in status, and Medicare required Prospective Payment System (PPS) assessments. In California the quarterly assessment does not include all items to calculate RUGs and is not used. Each assessment was assigned a weight using the RUGs total rate from Table 4 of FY2006 SNF PPS Final Rule (CMS 2009). The total rate is the sum of the nursing, therapy, and non-case-mix components. These weights were rescaled to make the mean weight equal one. In calculating the facility average CMI, the RUGs weight for each resident assessment was weighted by the length of stay associated with that assessment, determined as the number of days from the assessment date to the next assessment date or discharge. For assessments that were not admissions and that occurred after the first of the year, days were counted from January 1. Thus, the CMI for the facility was calculated as

A mathematical equation, expression, or formula.
Object name is hesr0046-0105-m2.jpg

where days is the number of patient days in RUG group r, and w is case mix weight for RUG group r.

Table 4

Marginal Cost for a One Percentage Point Increase in Percent of Days Due to Skilled Care (%SKILLED) by Expense Category

For the two categories relevant for skilled care residents, rehabilitation costs were calculated as the sum of expenditures reported for physical therapy, occupational therapy, and speech pathology, and NTA costs were calculated as the sum of pharmacy, laboratory, respiratory therapy, and other ancillary services.

We used two IVs in the two-stage least-square (2SLS) estimation approach. A discussion of the IV approach is included in the estimation section below. The first IV is the annual revenue of rehabilitation establishments (defined as offices of physical, occupational, and speech therapists and audiologists) that provide services to ambulatory patients in the zip code in which the NH is located, as reported in the Economic Census data. Because establishments in the Census data are grouped into aggregate revenue strata, this variable was calculated as IV=ΣW_iN_i, where N_i is the number of rehabilitation establishments in revenue stratum i, and the weights, W_i, is the middle point of the annual revenue range for the revenue stratum (except for the last category), as follows:

A mathematical equation, expression, or formula.
Object name is hesr0046-0105-m3.jpg

The second IV is the mean driving time between the zip codes of the NH and 15 closest community hospitals. We first estimated the driving time between each NH and the 305 community hospitals in California using Goggle map (Zdeb 2010). We then estimated the mean driving time between an NH and the 15 closest hospitals.

Estimation of Cost Function and Marginal Cost

We estimated a hybrid cost function following Nyman (1988) and Mukamel and Spector (2000) as follows:

A mathematical equation, expression, or formula.
Object name is hesr0046-0105-m4.jpg

where C is annual total costs, W is the average staff wages in NHs in the county, PD is adjusted inpatient days, CMI is the case mix index, %SKILLED is the percent of skilled care days, and FP is an indicator variable for for-profit status. We also included squared and cubed terms of the adjusted inpatient days to allow for both economies and diseconomies of scale depending on the pattern of signs of the three coefficients (Grannemann, Brown, and Pauly 1986). We used the average wage of nurse aides in the county to account for staff cost differences across the state. Wages for other nurses are highly correlated with this measure, and the vast majority of staff in NHs are nurse aides.

If the RUGs reimbursement method is capturing the cost of providing postacute care, then the coefficient of %SKILLED in the cost equation would be zero. If it is not tracking all costs of postacute care, then the coefficient would be positive.

Potential Endogeneity of Case Mix and %SKILLED

Costs may be endogenous with both CMI and %SKILLED for the following reasons. NHs often consider the financial implications of admitting residents when they develop their admission strategy. If the reimbursement system does not closely track actual costs of care for all patients, some patients are more profitable than others. In free-standing NHs, margins have been consistently higher for postacute than long-stay residents, creating an incentive to prefer postacute patients in general, and certain types of postacute residents may be preferred over others in particular (GAO 2002). Despite current occupancy rates in the mid–1980s, NHs can influence who applies for admission through their staffing decisions and provision of services geared toward the more complex and costly postacute care. Through marketing and development of relationships with other health organizations like hospitals and ambulatory rehabilitation offices, they may encourage a stream of potential residents who are more financially attractive. Furthermore, NHs can influence the RUGs score itself by determining the amount of services they provide, especially therapy minutes and therapy disciplines used, which also affects costs. Thus, %SKILLED, RUGs CMI, and costs are simultaneously determined. If the cost equation is estimated using ordinary least square (OLS), the endogeneity of CMI and %SKILLED will result in biased estimated coefficients. Consequently we estimated the cost equation with 2SLS treating both CMI and %SKILLED as endogenous variables. We perform the Hausman test of endogeneity and present estimates of both the OLS and the 2SLS to demonstrate the bias introduced by endogeneity.

IVs

At least one instrument is needed per endogenous variable. Good instruments are defined by two characteristics: first, they must meet criteria for strong IVs by being correlated with the endogenous variables; and second, they must meet the exclusion criterion, that is, they are not correlated with the cost equation error term. We chose two plausible instruments: (1) the average driving time between the NH and the 15 closest community hospitals (we performed sensitivity analyses using closest 5, 10, and 20 hospitals and as the results were similar, we present results using 15); (2) the weighted revenue of establishments that provide rehabilitation to ambulatory patients in the same zip code as the NH.

We expect these IVs to be strong, that is, correlated with both the CMI and %SKILLED. NHs with higher average driving time from hospitals will have less access to postacute residents, and hence will be less likely to be able to optimize their case mix. This should lead to a negative correlation of CMI and %SKILLED with this IV. NHs in areas with more demand for rehabilitation services are likely to have more ambulatory rehabilitation establishments and more NH rehabilitation. Consequently, we expect a positive relationship of CMI and %SKILLED with this IV. We use the minimum eigenvalue test proposed by Stocks and Yogo (2005) to test the hypothesis of weak IVs.

The exclusion criterion is not empirically testable, but the following arguments suggest that our IVs meet this criterion as well. With respect to the rehabilitation establishment instrument, the most likely threat to the exclusion criterion is the omission of rehabilitation staff wages from the cost equation. A high concentration of rehabilitation establishments locally may theoretically impact therapy wages. However, the IV is defined as the revenues of rehabilitation services at the zip code level, while wages are typically determined at the county or MSA level, due to worker mobility. Furthermore, other businesses besides rehabilitation offices employ therapists, including hospitals and home care agencies. Therefore, it is very unlikely that as defined, our IV is correlated with therapists' wages, and hence we believe that it meets the exclusion criterion. Furthermore, therapists account for <1 percent of NH costs, such that the impact on NH costs is likely to be negligible.

For the second IV, the mean driving time to the closest 15 hospitals, there may be similar concerns that RN wages and rehabilitation wages could be affected. Differences in the value of the instrument may affect the demand for RNs and therapists and consequently have impact on the wages faced by NHs. This in turn may affect hiring decisions and thus impact on costs. The impact on wages is likely to be small because hospitals are only one source of demand for these workers. The impact on cost would be negligible because RNs and therapists represent small shares of NH costs—rehabilitation is <1 percent and RN costs are 5 percent of total NH costs. Minimum NH staffing regulation also limits the ability of NHs to adjust staffing.

We estimate the marginal cost of an additional percent of skilled days using 2SLS regression with the IVs as described above. If the coefficient is significantly different from zero, this suggests that the CMI does not sufficiently account for all difference in costs associated with skilled care.

We calculate the marginal cost of increasing the %SKILLED by one percentage point. Because we used the log transformation of cost as the dependent variable, log scale predictions may provide a biased estimate of the impact of explanatory variables on mean cost. We retransformed the log cost predictions obtained from the second stage equation using the method described by Baser (2007). This method appropriately retransforms costs and accounts for possible bias due to heteroskedasticity in the error term.

To gain insights into what contributes to the costs of postacute patients that are not explained by the 53-RUG CMI, we performed additional analyses in which the dependent variables were the costs of rehabilitation care and NTAs. We use the same IV regression method as we do for the total cost analysis and estimate the same models. Following the analyses of MedPAC, we expect that most of the additional costs will be associated with rehabilitation and NTAs.

RESULTS

Table 1 presents descriptive statistics. On average, 11 percent of NH days were skilled care days, but the variation was large (25th percentile—5.2 percent; 75th percentile—14.8 percent). The average facility had 100 beds, annual costs of U.S.$5.7 million, and over 32,000 adjusted inpatient days. The facilities excluded from the analyses were not significantly different from the study sample in terms of total costs, bed size, wages, or ownership. However, they had significantly lower percentage of skilled care days (9 percent versus 11 percent).

Table 1

Descriptive Statistics (N=988)

We did the Hausman endogeneity test and confirmed the endogeneity of both %SKILLED and CMI (F_{(2, 922)}=770, p<.00001) and thus proceeded with the 2SLS analyses. Table 2 shows the first-stage equations. The dependent variables are %SKILLED and CMI. The incremental F-tests were high—F_{(2, 922)}=20.06 for %SKILLED and F_{(2, 980)}=18.56 for CMI equations, respectively. The minimum eigenvalue was 10.81, which is above the critical value of 7.03 for a nominal 5 percent Wald test for a 10 percent rejection rate, indicating that we can reject the hypothesis of weak instruments (Stocks and Yogo 2005). We find a negative and significant correlation between CMI and the mean driving time to the closest 15 hospitals. We also find significant and positive correlations between CMI and %SKILLED with revenues of ambulatory rehabilitation establishments in the same zip code as the NH.

Table 2

First-Stage Instrumental Variable Equations: Dependent Variables: Percent Skilled Care Days and Case Mix Index (p-Values in Parentheses); N=988

Table 3 compares the results of the regression of the log of total cost comparing the OLS with the 2SLS results. The estimated coefficients are very similar except for the coefficient for %SKILLED and CMI. Both models show the expected relationship between adjusted inpatient days and costs. The positive, negative, and then positive coefficients produce a typical S-shaped curve, suggesting increasing returns to scale followed by decreasing returns. As expected, costs also increase with wage levels and for-profit NHs have lower costs.

Table 3

Results from 2SLS and OLS: Dependent Variable Is Log Total Cost (p-Values in Parentheses); N=988

Although both coefficients are significant (p<.001) in the OLS, for %SKILLED the IV estimate in the 2SLS is almost twice the OLS estimate—0.012 compared with 0.007; for CMI the IV estimate is about 1/3 higher—0.911 compared with 1.245.

Table 4 shows the marginal costs from the 2SLS estimation of total rehabilitation and NTA costs. The marginal cost of increasing the %SKILLED by one percentage point, for the average facility, was U.S.$70,473, or 1.2 percent of the annual cost. For the subanalysis, the instruments remained strong in each of these 2SLS procedures. R²s were generally high, ranging from 0.27 for rehabilitation to 0.42 for NTAs. Sixty-two percent of the increase in costs due to a higher %SKILLED was due to higher rehabilitation costs and 32 percent was due to higher NTA costs.

DISCUSSION

NHs reimbursement for both postacute and long-term care attempts to cover the costs of treating the variety of patients that NHs serve. Case mix reimbursement method dominates reimbursement for both skilled and long-term care. It is important that these case mix reimbursement systems be well aligned with the costs of caring for patients. If they are not, they create incentives or disincentives to attract specific types of patients and may create barriers to access.

It has been difficult to design NH case mix reimbursement systems to properly meet the costs of care for all postacute patients. MedPAC suggests that the complex needs of postacute patients and their associated costs have been especially difficult to match for NTA services, which include expensive antibiotics, intravenous medications, and respiratory care. MedPAC raised concerns about the impact on access for these patients (MedPAC 2009a). To the extent that reimbursement is not paying appropriately for these patients, NHs treating a higher proportion of these patients will be disproportionately affected. This paper assesses whether the 53-RUGs CMI adequately captures the costs of care in NHs. It specifically examines the issue of whether the discrepancy is greater in facilities that have a higher proportion of skilled care days. The analyses confirm that this CMI does not capture all the costs and that additional costs are associated with skilled care. These additional costs are about 1.2 percent higher for each additional percentage point of skilled care days. The extra costs are mainly associated with rehabilitation and NTAs. These findings are broadly consistent with MedPAC's suggestions that with the 53-RUG index there is an underpayment for NTA expenses and some rehabilitation patients who have very high therapy costs.

Several strengths and limitations of this study should be noted. An important strength is the ability to account for the endogeneity of the postacute care percent and case mix. A limitation of the study is that it is based on data from one state. Although Medicare reimbursement is applied to states uniformly, Medicaid rates and reimbursement methodologies vary. Other supply factors may also differ. This may differentially affect the cost structure of NHs. Although the California RUGs case mix measure in this study is not based on quarterly assessments, calculations of the case mix in one state with “RUGable” quarterly assessments, with and without the quarterly assessments, showed little difference in the facility case mix scores. Furthermore, as a cross-sectional study, the analysis is vulnerable to omitted variable bias. Sensitivity analyses were carried out to determine whether alternative controls were needed to minimize missing variable bias. Competition as measured by the Herfindahl–Hirschman Index and indicators of regional location were included in alternative specifications. Competition was not statistically significant and region affected the size of the wage rate variable, suggesting that that variable was picking up regional differences in labor supply and other cost factors. In both cases the coefficient of interest, %SKILLED, and the strength of the IV, were not affected. Therefore, we report the cost equations without competition and regional dummies.

On August 11, 2009, CMS published the skilled nursing facility prospective payment Final Rule for FY2010 and 2011 (CMS 2009). With this rule it introduced a revised case mix classification system, the RUG-IV, which is based on time and motion study data collected in 2006–2007. The 53-RUGs were based on staff-time studies from 1997. With the final rule, CMS expanded the RUGs classification system from 53 to 66 groups. The RUG-IV expands and makes technical adjustments to rehabilitation, clinically complex, special care, extensive care, and rehabilitation groups. The revised index is intended to be applied using v3.0 of the MDS. Although these changes ought to better reflect the costs of some of the highest cost rehabilitation patients, the approach to paying NTAs, however, remains the same and is proportional to nursing staff costs. MedPAC has criticized this approach because staff time does not sufficiently explain the variation in the cost of NTAs. It appears that CMS is interested in exploring alternative approaches to reimbursing NTAs but did not include a new approach in this rule (CMS 2009). Although the changes made in the Final Rule are likely to improve the explanatory power of the RUGs, future research should determine to what extent the new changes in RUGs sufficiently account for differences in the actual costs of skilled care patients in NHs. The study we present here should be repeated in several years time, when data from the new RUGs become available, to determine whether the 66 RUGs system has corrected the shortcomings and resulting disincentives in the current system, or if further refinements, such as an outlier methodology for high rehabilitation patients and new approaches to reimbursing NTAs, are needed.

Acknowledgments

Joint Acknowledgment/Disclosure Statement: The research was partially supported by grant AG027420 from the National Institute on Aging.

Disclosures: None.

Disclaimers: None.

SUPPORTING INFORMATION

Additional supporting information may be found in the online version of this article:

Appendix SA1: Author Matrix.

Click here to view.^{(1.9M, pdf)}

Please note: Wiley-Blackwell is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

REFERENCES

Baser O. Modeling Transformed Health Care Cost with Unknown Heteroskedasticity. Applied Economic Research. 2007;1:1–6.
Centers for Medicare and Medicaid (CMS) Medicare Program; Prospective Payment System and Consolidated Billing for Skilled Nursing Facilities for FY 2006: Final Rule. Federal Register. 2005;70(149):45025–127. [PubMed]
Centers for Medicare and Medicaid (CMS) 2008. “Chapter 8—Coverage of Extended Care (SNF) Services under Hospital Insurance. Medicare Benefit Policy Manual” [accessed on May 4, 2010]. Available at: http://www.cms.gov/manuals/Downloads/bp102c08.pdf.
Centers for Medicare and Medicaid (CMS) Medicare Program; Prospective Payment System and Consolidated Billing for Skilled Nursing Facilities for FY 2010; Minimum Data Set, Version 3.0 for Skilled Nursing Facilities and Medicaid Nursing Facilities. Federal Register. 2009;74(153):40288–395. [PubMed]
Feng Z, Grabowski D, Intrator O, Mor V. The Effect of State Medicaid Case-Mix Payment on Nursing Home Resident Acuity. Health Services Research. 2006;41(4, part 1):1317–36. [PMC free article] [PubMed]
Fries BE, Schneider DP, Foley WJ, Gavazzi M, Burke R, Cornelius E. Refining a Case-Mix Measure for Nursing Homes: Resource Utilization Groups (RUG-III) Medical Care. 1994;32(7):668–85. [PubMed]
GAO. Skilled Nursing Facilities: Medicare Payments Exceed Costs for Most But Not All Facilities (GAO-03-183) Washington, DC: U.S. Government Printing Office; 2002.
Grannemann W, Brown S, Pauly MV. Estimating Hospital Costs: A Multiple Output Analysis. Journal of Health Economics. 1986;5(2):107–27. [PubMed]
Harrington C, Swan JH, Carrillo H. Nurse Staffing Levels and Medicaid Reimbursement Rates in Nursing Facilities. Health Services Research. 2007;42(3):1105–29. [PMC free article] [PubMed]
HCUP Databases. Healthcare Cost and Utilization Project (HCUP) Rockville, MD: Agency for Healthcare Research and Quality; 2005. Available at: http://www.hcup-us.ahrq.gov/databases.jsp.
Kaiser Commission on Medicaid and the Uninsured. 2009. “Medicaid and Long-Term Care Services and Supports” [accessed on November 4, 2010]. Available at: http://kff.org/medicaid/upload/2186_06.pdf.
Medicare Payment Advisory Commission (MedPAC) 2007. “Report to the Congress: Promoting Greater Efficiency in Medicare” [accessed on November 4, 2010]. Available at: http://www.medpac.gov/documents/Jun07_EntireReport.pdf.
Medicare Payment Advisory Commission (MedPAC) 2008. “A Data Book: Healthcare Spending and the Medicare Program” [accessed on November 4, 2010]. Available at: http://medpac.gov/documents/Jun08DataBook_Entire_report.pdf.
Medicare Payment Advisory Commission (MedPAC) 2009a. “Comment Letter to CMS Re: Revising the SNF Prospective Payment System” [accessed on November 4, 2010]. Available at: http://www.medpac.gov/documents/MedPAC%20to%20CMS.Feb%2002%2009.pdf.
Medicare Payment Advisory Commission (MedPAC) 2009b. “Skilled Nursing Facility Services Payment Basics Payment System” [accessed on November 4, 2010]. Available at: http://www.medpac.gov/documents/MedPAC_Payment_Basics_09_SNF.pdf.
Mukamel DB, Spector WD. Nursing Home Costs and Risk Adjusted Outcome Measures of Quality. Medical Care. 2000;38(1):78–89. [PubMed]
Nyman JA. The Marginal Cost of Nursing Home Care. Journal of Health Economics. 1988;7(4):393–412. [PubMed]
Rudder C, Mollot R, Mathuria B. 2009. “Modifying the Case-Mix Medicaid Nursing Home System to Encourage Quality, Access and Efficiency” [accessed on November 4, 2010]. Available at: http://nursinghome411.org/documents/finalreport_000.pdf.
Schlenker RE. Case Mix Reimbursement for Nursing Homes. Journal of Health Politics Policy and Law. 1986;11(3):445–61. [PubMed]
Stocks JH, Yogo M. Testing for Weak Instruments in Linear IV Regression. In: Andrews DWK, Stock JH, editors. Identification and Inference for Econometric Models: Essays in Honor of Thomas Rothenberg. Cambridge, England: Cambridge University; 2005. pp. 80–108.
White C, Pizer SD, White AJ. Assessing the RUG-III Resident Classification System for Skilled Nursing Facilities. Health Care Financing Review. 2002;24(2):7–15. [PubMed]
Zdeb M. 2010. “Driving Distances and Times Using SAS® and Google Maps” [accessed on March 10, 2010]. Available at: http://support.sas.com/resources/papers/proceedings10/050-2010.pdf.

Articles from Health Services Research are provided here courtesy of
Health Research & Educational Trust