Printer Friendly

Nursing home costs, Medicaid rates, and profits under alternative Medicaid payment systems.

This analysis compares nursing home costs, Medicaid payment rates, and profits under three Medicaid nursing home payment system: case-mix, facility-specific, and class-rate systems. Data used were collected from 135 nursing homes in seven states. The association of case mix with costs, rates, and profits under the three payment systems was of particular interest. Case mix was more strongly associated (positively) with patient care cost and the Medicaid rate for the case-mix systems than for the other systems, particularly the class-rate systems. In contrast, case mix and profits were not associated in the case-mix or facility-specific systems, but were negatively associated in the class rate systems. Overall, the results suggest that case-mix systems have some important advantages over the payment systems, but further research is needed on larger samples and involving the newer case-mix systems.


Medicaid payment system issues continue in importance due to the rapid growth in nursing home expenditures and the major portion of those expenditures paid by Medicaid. Nursing home expenditures increased from $2.1 billion in 1965 to $35.0 billion in 1985, the main data collection year of this study. In 1985, Medicaid accounted for 42.4 percent of total nursing home expenditures (Health Care Financing Administration, Division of National Cost Estimates 1987).

This article presents cost, Medicaid payment rate, and profit findings from a larger research study of Medicaid nursing home payment systems conducted during 1983-1988 and funded by the Health Care Financing Administration (HCFA). (The overall study results are presented in Schlenker et al. 1988, and summarized in Schlenker 1991.) The major premise of the study was that the structural characteristics of a payment system strongly influence nursing home behavior and, ultimately, overall outcomes.


Three major nursing home payment system types were analyzed: "case-mix," "facility-specific," and "class-rate" systems. Case-mix systems represent the newest payment approach and tie payment more directly than the other systems to patients' assessed care needs. Facility-specific systems base a nursing home's payment rate in its average costs, while class-rate systems pay a fixed rate per patient day regardless of costs. The performance of the case-mix systems thus was of major interest in this study. Each system type creates different financial incentives for nursing homes and, correspondingly, responses are expected to differ by payment system type.

The initial study phase involved categorizing Medicaid nursing home payment systems and then selecting representatives states from each system category. The classification process focused on the Medicaid payment rate component that was most closely associated with direct patient care. This rate component typically covered nursing staff and related costs. The three payment system types and seven states selected for this study in 1984 are summarized in Table 1, and discussed next.


Case-Mix Systems

As noted above, case-mix systems are relatively new, but of growing importance. The basic concept of case-mix payment is that the payment for nursing home care is based directly on patient's care needs and on state-estimated costs of providing care appropriate to those specific needs. Four states had case-mix systems at the start of this study (Illinois, Maryland, Ohio, and West Virginia) and two more (Minnesota and New York) adopted them while the study was in progress. Several other states are implementing, developing, or considering such approaches. Case-mix systems also are encouraged under the Omnibus Budget Reconciliation Act of 1987 (P.L. 100-203). In addition, four states (Kansas, Maine, Mississippi, and South Dakota) are currently involved in case-mix system development as part of the HCFA-funded Multistate Nursing Home Case Mix and Quality Demonstration.

Maryland, Ohio, and West Virginia were selected in 1984 as the study's case-mix states. Although the following descriptions point out the differences among the three systems, their overall approaches to case-mix measurement and to translating case-mix information into payment rates are broadly similar. The approaches also set these states apart from the states representing the other two payment systems.

Maryland's patient case rate component is based on the patient's case-mix characteristics. Rates are set for four general care categories (light, moderate, heavy, and heavy-special care) based on five ADLs (activities of daily living) and three services. Separate rates also are set for ten specific services (such as turning and positioning, tube feeding, and injections). Maryland encourages nursing homes to accept heavy care patients by increasing the "profit margin" in rates paid for the heavier care patient categories. The individual ADL and services rates are determined prospectively each year (and vary according to three geographic regions) and are not tied to the costs of individual nursing homes. Nursing homes bill the state monthly for each resident's care according to these rates (these data are periodically audited by a state-contracted organization).

Ohio's case-mix payment methodology is facility specific and retrospective. In contrast to the Maryland approach, case-mix information is used in Ohio to set facility ceiling on reimbursement rates. Payment for patient care services is based on each facility's costs up to the case mix-determined ceiling. The case-mix ceilings are based on the state-estimated cost of 20 separate items (five ADLs, nine special services such as injections, and six rehabilitation services). Patient assessments are conducted every three to six months by state personnel to determine the need for services by each nursing home's residents. Ohio's selection of the case-mix services and the service cost estimation methodology drew on West Virginia's approach and experience.

In West Virginia, rates are facility specific as in Ohio, but are set prospectively rather than retrospectively. West Virginia, like Ohio, uses case-mix information to determine facility rate ceilings. A nursing home's patient care rate component is determined from patient assessment data provided by the facility (and audited periodically by the state). The assessments are based on information for 15 services, and the state-estimated cost of providing each service (based on time studies) is used to determine the case-mix ceiling amount for each facility. The lower cost - between case mix ceiling or actual cost (with an inflation adjustment - becomes the facility's subsequent (prospective) patient care rate component.

Facility-Specific Systems

At present, most states use a facility-specific payment system. The common feature of such systems is that the payment to each nursing home is linked in some way to its costs. Most such systems are prospective, so that past costs are used to set future payment rates. A HCFA survey identified 13 retrospective and 25 prospective facility-specific state systems during the study's state selection phase (Jazwiecki 1984).

Colorado and Florida were the selected facility-specific states. Although their individual features differ in detail, the two payment systems are quite similar in their overall structures.

Colorado's system is prospective and facility specific, but it does not include case mix as a factor in determining payment rates. In setting the prospective rates, costs up to ceiling amounts are allowed. In 1985, a ceiling was used for patient care and raw food costs (combined), and no efficiency incentive payments were made for costs below the ceiling. (An efficiency incentive is the payment of some portion of the difference between a ceiling and actual cost per day to facilities with costs below the ceiling.)

Florida's system also is prospective and facility specific. In determining rates, patient care costs are treated separately, with a higher ceiling than for other cost components, but case mix is not directly incorporated into the rate determination methodology. Although efficiency incentive payments are possible in the patient care rate component, they are linked to quality ratings from the Medicaid certification survey in an effort to encourage quality. Higher amounts are paid to facilities with "superior" quality ratings.

Class-Rate Systems

In a class-rate system, nursing homes are paid a fixed amount per patient day by class or category of patient. Although class rates also can be based on nursing home characteristics such as size and geographic location, the focus in this study was on classes defined by patient characteristics, primarily the Medicaid skilled and intermediate levels of care. Class-rate systems were fairly common in the early years of Medicaid, but by the mid-1980s few states had such systems (the earlier-mentioned survey identified six class-rate states). Despite their limited current use, class-rate systems were important for this study because of the strong incentives they create for nursing homes to minimize the costs of providing care, possibly by admitting primarily light care patients or by limiting quality, or both.

Texas and Utah were selected as the study's class-rate states. Their essential features are basically the same. Although Texas implemented a case-mix payment system in 1989, during the study it used a class-rate methodology based on three levels of patient care: SNF (skilled), ICF (intermediate), and ICF-II (light intermediate). (The ICG-II category was quite rare, so that essentially two levels were in use.) Uniform statewide class rates were determined each year, with the rates based on the median plus seven percent of statewide costs per day by level of care.

Utah's class-rate methodology also includes three levels of care: SNF, ICC-I and ICF-II. (Utah, in addition, uses separate negotiated rates for patients with extremely costly care, needs.) The Utah class rates are established annually, based on negotiations between the state and the nursing home industry.

The categorization into case-mix, facility-specific, and class-rate systems was developed to highlight the major financial incentives the different payment methodologies create for nursing homes. Nursing home responses to those incentives were then hypothesized to determine overall system outcomes, as discussed next.


Table 2 highlights the study hypotheses on cost, Medicaid rate, and profit differences by payment system type. The hypotheses address expected differences across payment systems both in the levels of costs, rates, and profits, and in the associations of case mix with those variables. The major underlying assumption was that, despite each state's unique payment system features, providers' incentives and responses would be similar within a payment system category and different across categories. The null hypothesis in each instance was that no difference would be found among systems. The statistical analyses tested for alternative hypotheses in either direction.


Costs and Medicaid Rates

The levels of 91) patient care cost per day, (2) its share of total cost, and (3) the Medicaid payment rate per day were all hypothesized to be highest for case-mix systems, lowest for class-rate systems, and between the two for facility-specific systems. The associations of case mix with both patient care costs and Medicaid payment rates also were hypothesized to be strongest in the case-mix systems, weakest in the class-rate systems, and in-between for the facility-specific systems.

The cost and rate hypotheses built on earlier studies, but added the newer case-mix approach to the facility-specific and class-rate payment system variants. (Reviews of prior relevant research include Bishop 1980b; Palmer and Cotterill 1983; Stassen and Bishop 1983; Schlenker 1986; and Hawes and Phillips 1986.) The underlying rationale for these hypotheses was that payment rates under case-mix systems were expected to respond more directly than under the other systems to differences in patient care needs. Thus, case-mix systems were expected to facilitate greater access to nursing home care by high-cost intense case-mix Medicaid patients. The anticipated outcomes were higher expected costs and rates in those systems, and stronger associations of costs, and rates with case mix.

In contrast, the fixed payments by broad level of care category under class-rate systems were expected to lead nursing homes to admit primarily light care patients within each class and to minimize the cost of their care. Rates were expected to be low in class-rate systems, both because an objective of low rates often underlies a state's decision to use a class rate methodology, and because the low nursing home costs encouraged by class rates may reinforce a state's tendency to set low rates. These factors were expected to result in lower cost and rate levels in class-rate states, and also were expected to weaken the association of costs and rates with case mix.

Facility-specific systems were expected to be more responsive (in terms of rates) to care need and cost differences than class-rate systems but less responsive than case-mix systems. Thus, case-mix intensity and therefore cost and rate levels in facility-specific systems were hypothesized to be between those of the class-rate and case-mix systems. The association of case mix with both costs and rates in facility-specific systems also was expected to be stronger than in class-rate systems but weaker than in case-mix systems.


Several prior studies have used the concept the nursing homes operate in two primary markets: a private pay and a Medicaid market. (See, for example, Scanlon 1980; Bishop 1980a; Cotterill 1983; and Palmer and Vogel 1983.) Conventional supply and demand forces are assumed to operate in the private pay market, while in the Medicaid market the state's buying power allows it to purchase care at a lower rate than in the private pay market, creating a situation of excess demand for Medicaid care. The Medicaid payment system thus represents only one of several important influences on nursing home profitability. In this study, therefore, the relationships between profits and Medicaid payment system type were not expected to be as strong as those hypothesized for costs and rates.

The profit hypotheses dealt with the level of profits (measured by the profit ratio, defined further on) and with the association between case mix and profits. Specifically, profit levels were expected to be highest in class-rate systems, lowest in case-mix systems, and between the two for facility-specific systems. Further, the association between case mix and profits was hypothesized to be positive in case-mix systems and negative in class-rate systems. No strong association of case mix with profits was hypothesized for facility-specific systems.

The hypothesis on profit level assumed that the strong class-rate incentives to minimize cost would lead to higher profits than under the other systems. In case-mix systems, rate are designed to vary with the case mix-determined costs of care, making it more difficult to achieve high profits (at least from the direct patient care portion of the payment rate). Similarly, rates in facility-specific systems are closely tied to facility costs, thereby limiting profitability.

The hypothesized positive association of profits with case mix under case-mix systems was based on concerns that such systems may encourage providers to obtain profits by keeping patients debilitated or providing them only minimal services (Smits 1984; Kane and Kane 1988). However, this hypothesis was considered to be fairly weak. For example, in some case-mix states (e.g., Ohio and West Virginia), payment rates are based on costs up to a case mix-determined ceiling. Higher payments for heavier care patients in such circumstances are unlikely to lead to higher profits since payments will be made for the lesser between costs or the case-mix ceiling.



A "basic" sample and an "urban profit" subsample of facilities and patients were selected from the seven study states for primary data collection. The facility selection process for the basic sample obtained around 20 facilities per state. The facility samples were not intended to represent each state's entire nursing universe. Rather, stratified random samples were designed to represent the most common types of nursing homes providing care to Medical recipients. The intent of the analysis was to identify behavioral responses to payment system incentives for specific categories of nursing homes (particularly urban profit facilities). The study was not designed to derived impact estimates generalizable nationally or to any individual state.

As a result, freestanding, Medicaid-certified, nongovernment general care facilities were selected, and proprietary ownership nursing homes located in urban areas (i.e., in metropolitan statistical areas) were emphasized. Termed "urban profit" nursing homes in this study, such facilities represent the dominant category of nursing homes, and they were expected to be the most responsive to the financial incentives of Medicaid payment systems. For these reasons, the urban profit subsample was chosen to comprise two-thirds of the entire basic sample.

Further selection criteria included a bed size between 40 and 240 beds, a relatively high occupancy rate (over 70 percent), and a fairly high percentage of Medicaid patients (at least 50 percent). Once all facilities in a state meeting the criteria were identified, random samples were drawn within four facility categories (i.e., urban profit, urban nonprofit, rural profit, and rural nonprofit) to meet prespecified samples quotas by stratum.

For the patient-level data collection, random samples of 25-30 patients were selected from each sample facility. All payer categories were included, with the requirement of a minimum of ten Medicaid patients in each nursing home's sample.

The resulting basic sample including 135 nursing homes and 3,508 sample patients, of which 2,320 (66 percent) were Medicaid patients. The urban profit subsample included 91 facilities with 2,376 sample patients, of which 1,603 (67 percent) were Medicaid patients. Descriptive sample characteristics are presented in Table 3. In terms of the percent Medicaid patient days and occupancy rate, the sample were broadly similar to national averages of 60-70 percent for the percent Medicaid and over 90 percent for the occupancy rate, except for the somewhat lower occupancy rates in the class-rate samples (National Center for Health Statistic 1988a, 1988b; Hawes and Phillips 1986). Average facility bed size (number of beds) varied considerably among the state sample (from 89 in West Virginia and Utah to 122 in Florida and 123 in Maryland), but was generally in the range of the national average of around 100 beds. Sample average for cost per day and the Medicaid rate also are presented in Table 3, and highlight the considerable range in these variables across states.


Data were collected primarily for 1985 and included facility-level cost, rate, and profit variables; facility characteristics; area or market factors; and extensive patient-level information on case mix and quality.

Dependent Variables

The three major facility-level dependent variables in the analyses were (a) patient care cost per patient day, (b) the Medicaid payment rate per patient day, and (c) the ratio of patient care-related revenues to expenses as the profit measure (termed the profit ratio). Both the cost and rate variables were adjusted for geographic-area wage differences, using wage index data for using nursing homes obtained from HCFA (non-wage geographic indexes were not available).

Patient care costs were defined as nursing staff (including aides) salary, wage, and benefit costs, as well as nursing supply costs. Also included were other direct patient care staff costs, such as those for social services, activities, and therapies. The mean and standard deviation of patient care cost per patient day (wage index adjusted) for the basic sample of 135 nursing homes were, respectively, 19,000 and 4.569.

In the rate analyses, the total Medicaid payment rate was used as the dependent variable rather than the rate component related to direct patient care (i.e., that portion of the rate intended to cover nursing staff time and associated costs) The use of the total rate was necessary because in the class-rate states a paint care rate component could not be separated from the total Medicaid rate. The basic sample mean and standard deviation for the Medicaid rate variable were 42.071 and 7.949.

Profits are variously measured in the literature, and the accounting measure used here differs from the economic concept of profit as the amount paid to a resource over what it would receive in its most productive alternative use (McCaffree, Malhotra, and Wills 1979). However, only accounting data were available, and the most useful profit measure was the ratio of patient care-related revenues to expenses. This was a before-tax measure that excluded revenues from other sources such as investment and cafeteria operations, and also excluded expenses not considered allowable by the state Medicaid programs. The profit ratio is also referred to as the "mark-up" in the literature (see Alexander and Lewis 1984; for other profit measures see Meigs and Meigs 1981; and Lev 1974). The mean and standard deviation for this variable were 1.016 and 0.158.

Independent Variables

Case mix and quality were measured by patient-level variables aggregated to the facility level. Case mix was measured based on simulated case-mix payment rates for the three case-mix payments systems. The variables was created using a simulation model of case-mix payment rates developed for this study. The model utilized patient-level primary data calculate the case-mix payment rates that would result for each sample patient and nursing home under the Maryland, Ohio, and West Virginia systems. The resulting simulated patient-specific payment rates under each system were converted to ratios relative to the average rate for that system for all sample patient. The ratios for all three simulated systems were then averaged for each patient and aggregated to the facility level. The resulting "average rate ratio", also denoted as the "case-mix index," was used as an indicator of relative case-mix intensity (it was positively correlated with other case-mix measures, such as ADL indexes and various indicators of long-term care problem prevalence and severity). Although quality could not be well measured in this study, the indicator used was the proportion of patients with ulcerations (other quality-related variables - such as catheterization, urinary tract infection, and restraint rates - also were tested, with similar results).

Area/market and facility characteristic variables were selected to control for additional factors potentially affecting the dependent variables. The selection was based on conceptual nursing home cost models, reviews of prior nursing home cost studies (several reviews were cited earlier), and preliminary correlation and regression analyses. The area/market factors had little effect on the results, so only the country nursing home beds per hundred elderly variable was included in the final equations (as a rough measure of area nursing home supply relative to demand). The final facility characteristics were profit or nonprofit ownership (nonprofit = 1), urban or rural location (rural = 1), Medicaid and Medicaid shares of patient days, bed size (number of beds), occupancy rate, and chain affiliation. Dichotomous variables were used to test for state and payment system effects, as explained in the next section.

Medicaid cost reports were the source of the cost and profit data, while the state Medicaid agencies provided the Medicaid payment rate information. Other independent variables used in the analyses were obtained from several sources. The case-mix and quality variables were aggregated to the facility level from the patient-level primary data collected for the study. Facility characteristics were obtained from cost reports and a survey of all sample nursing homes. Area/market factor data were obtained from various secondary source, primarily the Area Resource File.


All analyses presented in this article used the facility as the unit of analysis. The descriptive comparisons of states and payment systems included an analysis of high-profit versus low-profit nursing homes (termed the high/low profit analysis). Thereafter, regression techniques examined multivariate relationships.

For the high/low profit analysis, each nursing home's profit ratio was divided into the following components (for which data were available in Medicaid cost reports):

p = (M/C)m + (N/C)(1 - m) = xm + y(1 - m)


p = the profit ratio (revenues/expenses); M = Medicaid revenue per Medicaid patient day; N = non-Medicaid revenue per non-Medicaid patient day; C = Patient care cost per patient day (all payers combined)

equals facility average cost per patient day; m = Medicaid patient days as a proportion of total (all-payer)

patient days (i.e., the Medicaid share); x = the Medicaid revenue/cost ratio = M/C; and y = the non-Medicaid revenue/cost ratio = N/C.

The nursing home in each state sample were divided into high-profit and low-profit groups, based on the median profit ratio for the state's sample/. The high- and low-profit facilities were then pooled by payment system (due to the small facility samples in each state), and two-sample statistical tests were applied to each payment system pool.

The regression analyses used one basic model to examine the association between case mix and each dependent variables. The model was designed to allow for a different association between case mix and the dependent variables under each payment system through use of the following functional form (for the portion of the equation involving the case-mix variable):

Z = a + b(CM) + [b.sub.i](CM x [SYS.sub.i])

where: Z = dependent variables; CM = case=mix index (average rate ratio); [SYS.sub.i] = payment systems (a 0:1 variable for each system i); and a,b,[b.sub.i] = regression coefficients.

This functional form incorporate an overall linear association between case mix and the dependent variable, represented by the "all systems" coefficient b for the case-mix index (CM). The possibility of a different linear association between case mix and the dependent variable under each payment system is then incorporated through the use of interaction terms combining the case-mix variables with a dichotomous variable for each payment system. The resulting payment system interaction coefficients [b.sub.i] can then be combined with the overall case-mix coefficient (b) to obtain the total case-mix effect on the dependent variable under each payment system. (Technically, it is impossible for all three interaction terms to be included in a regression equation with the overall case-mix variable, since the interaction terms would sum to the case-mix variable and the estimation of individual coefficients would be impossible. Thus, in the estimation procedure, either the overall index or one of the interaction terms was delete.) Both full model and stepwise regression techniques were used. Stepwise regression results are presented in this article, and thus show the equations after eliminating statistically insignificant variables.

A dichotomous variable for each state also was included, to allow for other state-level factors to affect the levels of the dependent variables. The model thus tested for a different association of case mix with the dependent variable for each payment system, but allowed for a different effect for each state on the dependent-variable level. This was done in order to examine a major premise of this analysis: that the general structure of each payment system, even with individual state differences, would yield a unique association case mix and the dependent variables under each payment system type.






Descriptive cost and rate data by state for the basic and urban profit samples are presented in Table 4. These data were used to address the hypotheses on levels of costs and rates. Significance tests were not carried out due to the small sample sizes.

Average patient care cost per day (after adjustment for geographic wage rate differences) was higher in the case-mix states than in the class-rate states. The result for the facility-specific states were mixed. Florida's patient care cost per day was the highest of all study states, while Colorado's was between the averages for the case-mix and the class-rate states. The average patient care shares (or proportions) of total cost per day were higher for the case-mix states than for the class-rate states in both samples. However, the shares for the facility-specific states were higher than those for the class-rate states and for two of the three case-mix states (i.e., except Ohio).

The same general pattern was found for the Medicaid payment rate per patient day (also after geographic wage rate adjustment). The average Medicaid rates were higher for the case-mix states than for the class-rate states. For the facility-specific states, Florida's rate was the highest of all seven states, and Colorado's was between the rates of the two class-rate states (higher than the Texas rate and slightly lower than the Utah rate).

These descriptive results are consistent with the hypotheses on case-mix-class-rate system comparisons, but are relatively uninformative without the multivariate analyses. As a prelude to those analyses, the case-mix index (i.e., the average rate ratio described earlier) is shown in Table 4 to follow the same general pattern as the cost and rate variables, suggesting that positive associations may be found between case mix and the cost and rate variables in the regression analysis.



The profit analysis included 89 urban profit nursing homes (two of the original 91 had incomplete profit data). The urban profit category was expected to be the most strongly affected by and responsive to financial and market incentives. Their overall profit (revenue:expense) ratio was 1.02, indicating that revenues exceeded expenses on average by 2 percent. Average profit ratios by payment system (shown by state in Table 4) were 1.07 for the urban profit sample case-mix and facility-specific nursing homes, and 0.91 (representing a loss) for the class-rate facilities. The high-profit/low-profit comparisons (by payment system) are presented in Table 5.

The high-profit groups' average revenue:expense ratios were similar for the case-mix and facility-specific systems (1.15 and 1.14), as were the low-profit ratios (0.98 and 1.00). However, both ratios were lower for the class-rate system (1.07 and .75). In fact, the low-profit class-rate group had revenues averaging only three-fourths of expenses. Clearly, these results did not support the hypothesis of higher profits under class rate systems.

Profit Components in Case-Mix and

Facility-Specific Systems

The profit component patterns were similar for the case-mix and facility-specific systems. As indicated in Table 5, the average Medicaid revenue and total cost per patient day variables were similar between the high- and low-profit groups. In contrast, the high-profit groups in both systems had significantly higher non-Medicaid revenue per patient day than the low-profit groups. As a result, the Medicaid revenue:cost ratios were not significantly different between the high- and low-profit groups in either system, but the non-Medicaid revenue: cost ratios were considerably higher for the high-profit than for the low-profit groups (even though the high-profit/low-profit difference was statistically significant only for the facility-specific system).


The Medicaid share was also considerably lower, on average, for high-profit compared to lower-profit under both systems, although the difference was statistically significant only for the case-mix system. Neither the nursing cost per patient day nor the case-mix index was significantly difference between the high- and low-profit groups.

These results suggest that higher profits in these systems were obtained through a combination of (a) higher non-Medicaid revenues per (non-Medicaid) patient day, and (b) lower participation in Medicaid (i,e., lower Medicaid shares). That is, it appears that high-profit nursing homes achieved their higher profits outside the Medicaid system.

Profit Component in Class-Rate Systems

The class-rate results differed considerably from the other two systems. The differences in Medicaid and non-Medicaid revenue per day and in the Medicaid share were not statistically significant (although the mean values of both revenue variables were higher for the high-profit group). Total cost per day, however, was significantly lower for the high-profit group, and as result, both revenge:cost ratios were significantly higher for the high-profit group. Nursing cost per day also was significantly lower for the high-profit group, but the case-mix index was approximately the same for both groups.

Thus, profitability in the class-rates states appeared to be associated with lower overall costs, and lower nursing costs for approximately the same case mix. Whether the lower costs reflect greater efficiency or lower quality deserves further study, particularly in view of the lower nursing costs without lower case mix in the high-profit group.




The final cost, rate, and profit regressions for the basic sample of 135 facilities are presented in Table 6. As noted above, both full model and stepwise regression were estimated, with the stepwise regression results presented here. Of relevance to assessing the results for all three equations is that the case-mix index (i.e., the average rate ratio) was derived from a sample of patients in each nursing home, which introduces additional measurement error. This typically results in biased and inconsistent coefficient estimators, with asymptotic bias toward zero (Johnson 1972). The coefficient estimators in Table 6 thus may be understated (both positively and negatively), so that the true coefficients may actually be greater in absolute value than indicated in Table 6, which would strengthen effect of case mix in equations.


The patient care cost per day regression accounted for about two-thirds of the dependent variable variation ([R.sup.2] = .661). All listed variables were included in the analysis, but coefficients and significance levels are presented only for variables that were significant are the p < .10 level by the stepwise regression procedure. The equation shown in Table 6 indicates a positive overall association between case mix and cost (the case-mix index coefficient is 12.307). The case-mix interaction terms reveal a weaker case mix-cost relationship for the class-rate system (indicated by the negative class-rate coefficient of -4.482). The total case-mix effect under each payment system is the sum of the case mix co-efficient and the relevant interaction coefficient, as follows:


This result in consistent with the hypothesis of a stronger case mix-cost association in the case-mix states than in the class-rate states. The results also imply that the association in facility-specific states is similar to that in the case-mix states. The only state variable to enter this equation was Ohio, with a positive coefficient. This suggests higher patient care costs in Ohio than in the other states, after controlling for the remaining variables in the equation.

The other variables to enter the equation will only be briefly mentioned. The quality indicator had a significant negative coefficient. Since a higher proportion of ulcerations is assumed to reflect lower quality, this result suggests that lower quality is associated with lower patient care cost. The nursing home beds per elderly variable was not significant, and the significant facility characteristic variables were nonprofit ownership, rural location, the percent Medicare patients (all positive coefficients), and chain affiliation (negative coefficient). The area/market and facility characteristic results generally agree with the findings of other nursing home cost studies, such as those cited earlier. The most important finding for this study was the stronger association between case mix and patient care cost in case-mix systems (and in facility-specific systems) than in class-rate systems.



The Medicaid rate regression results are also presented in Table 6. The functional form and independent variables used for the rate regressions were the same as for the cost analysis. However, for the rate equation, two case-mix indexes were tested - one for all patients and one for only Medicaid patients, because in theory it is primarily the Medicaid case mix that should affect the Medicaid payment rate. However, the two variables were highly correlated and yielded similar results in the regressions. Therefore, due to the larger patient samples (in each facility) underlying the all-patient measure, it was used in the final analyses.


The explanatory power of the rate equation was higher than for the cost equation, with an [R.sup.2] of .760. Because the Medicaid rate covers all rate components, whereas the cost analysis focused on the patient care cost component, the association between case mix and the total Medicaid payment rate was expected to be weaker than the case mix-cost association. This expectation was confirmed; the overall case mix-rate association for all payment systems was positive and marginally significant (p = .051). A stronger positive association was indicated for the case-mix systems by the significant case-mix system interaction coefficient of 6.977, and the insignificant coefficients for the other system interaction variables.

Thus, the cost and rate regressions both revealed major differences between the case-mix and class-rate systems in the role played by case mix. In contrast, the facility-specific systems showed a strong case mix-cost association (like the case-mix systems) but a weak case mix-rate association (like the class-rate systems).

Three of the state variables had significant coefficients in the rate equation. The results implied that, compared to the other four states (Maryland, West Virginia, Colorado, and Utah), Florida had a higher Medicaid rate (positive coefficient), and Ohio and Texas had lower rates (negative coefficients), after controlling for case mix and other factors captured by the equation. The negative Ohio coefficient must be considered in conjunction with the positive coefficient for the case-mix system interaction variable. Together, the two coefficients implied a lower overall Medicaid rate in Ohio than in the other two case-mix states, but a rate that was still positively associated with case mix.

The quality variable was not significant in the equation, nor was the nursing home beds per elderly variable. The significant facility characteristics suggest that after controlling for case mix and other included factors, rural location and high occupancy rates were associated with higher payment rates, and high participation in Medicaid was associated with lower payment rates.

Overall, the rate regression results were consistent with the hypotheses. In particular, the findings suggest a strong positive association between case mix and the Medicaid payment rate under a case-mix system, and weaker positive (and only marginally significant) associations under the facility-specific and class-rate systems.


The profit regression equation (see Table 6) accounted for only slightly over one-third of the variation in the profit ratio ([R.sup.2] = .358), considerably less than was accounted for by the cost and rate equations. However, a low [R.sup.2] was expected due to the many factors affecting profits that could not be adequately incorporated into the regression.


The only significant case-mix variable was the class-rate system interaction variable, with a negative coefficient. This implies that case mix and profits were negatively associated under class-rate systems, but were not associated under the other two systems. The only significant state variable was Texas. Its positive coefficient implies higher profits for Texas than for the other study states. However, the negative class-rate interaction coefficient denotes a lower profit ratio under the class-rate system compared to the other systems (at a given case-mix index value). Thus, within the context of lower class-rate profits, the positive Texas coefficient indicates only that the profit ratio was somewhat higher in Texas than in Utah.

The quality indicator was marginally significant (with a p-value of .099) and negative, which suggests a positive but weak association between quality and profitability that warrants further study in future work. With regard to the other significant variables, the nonprofit coefficient was negative, suggesting that lower profits were associated with nonprofit ownership. The Medicaid and Medicare variables also were significant and negative, implying that greater participation in Medicaid or Medicare was associated with lower profits. These results agree with findings of studies such as those cited earlier.

The profit regression results thus point to (a) no association between case mix and profit in case-mix and facility-specific systems, and (b) a negative association in class-rate systems. In particular, there was no evidence that under case-mix payment higher profits were obtained by facilities with more intense case mix.


The main findings of these analyses were that patient care costs and Medicaid payment rates were more closely associated with case mix under case-mix systems than under the other two systems types, particularly class-rate systems. Also, profits were negatively associated with case mix in class-rate systems but not associated in the other systems.

These results suggest that case mix-systems can be more effective than other systems in linking Medicaid payment to patient care needs, and also that the resulting cost structure of nursing homes is more in line with care needs in case-mix systems than in the other systems. Further, the case-mix systems may facilitate access for higher-cost Medicaid patients by avoiding the negative correlation between higher case mix and lower profits present in the class-rate systems.

The facility-specific systems shared some positive features with case-mix systems, largely because facility-specific rates are responsive to individual facility costs. However, cost differences may be due to many factors other than case mix, so that the connection of case mix to costs and rates is weaker in facility-specific systems than in case-mix systems (but stronger than in class-rate systems).

Overall, the results suggest that case-mix systems have several advantages over other nursing home payment systems. Additional analyses from the larger study supported the advantages of case-mix systems but also highlighted important caveats. In particular, both access and quality need to be studied over longer time periods, with an emphasis on tracking outcomes for patient cohorts over time. In addition, studies involving larger samples of states, facilities, and patients are needed.

The trend toward case-mix systems was noted at the outset of this article. This study examined the earlier case-mix systems, and research on the more recent systems is warranted. The newer systems usually employ the resource utilization group (RUG) patient classification methodology, which, compared to the case-mix systems covered in this study, classifies patients into broader categories and does not tie payments as directly to the provision of specific services. The evaluation of the current HCFA Multistate Case Mix and Quality Demonstration, as well as comparative longitudinal studies of states with and states without case-mix payment systems, therefore can further our understanding of nursing home payment policies and their effects.


Alexander, J., and B. L. Lewis. "The Financial Characteristics of Hospitals

under For Profit and Nonprofit Contract Management." Inquiry 21,

no. 3 (Fall 1984): 230-42. Bishop, C. E. Nursing Home Behavior under Cost-Related Reimbursement. Waltham,

MA: Brandeis University, University Health Policy Consortium,

1980a. ____. "Nursing Home Cost Studies and Reimbursement Issues." Health Care

Financing Review 1, no. 4 (Summer 1980b): 47-64. Cotterill, P. G. "Provider Incentives under Alternative Reimbursement Systems."

In Long-Term Care: Perspectives from Research and Demonstrations.

Edited by R. J. Vogel and H. C. Palmer. Washington, DC: U.S.

Department of Health and Human Services, 1983. Hawes, C., and C. Phillips. "The Changing Structure of the Nursing Home

Industry and the Impact of Ownership on Quality, Cost and Access." In

For-Profit Enterprise in Health Care. Edited by B. H. Gray. Washington

DC: National Academy Press, 1986. Health Care Financing Administration. Division of National Cost Estimates.

Office of the Actuary. "National Health Expenditures, 1986-2000."

Health Care Financing Review 8, no. 4 (Summer 1987): 1-36. Jazwiecki, T. "How States Pay for Long-Term Care Facility Services under

Medicaid." Healthcare Financial Management 38, no. 4 (April 1984):

76-80. Johnston, J. Econometric Methods. 2d ed. New York: McGraw-Hill, 1972. Kane, R. A., and R. L. Kane. "Long-Term Care: Variations on A Quality

Assurance Theme." Inquiry 25, no. 1 (Spring 1988): 132-46. Lev, B. Financial Statement Analysis: A New Approach. Englewood Cliffs, NJ:

Prentice-Hall, Inc., 1974. McCaffree, K. M., S. Malhotra, and J. Wills. "Capital and the Reimbursement

of the Costs of Nursing Home Services." In Reform and Regulation in

Long-Term Care. Edited by V. LaPorte and J. Rubin. New York: Praeger,

1979. Meigs, W. B., and R. F. Meigs. Accounting: The Basis for Business Decisions. 5th

ed. New York: McGraw-Hill, 1981. National Center for Health Statistics. Public Health Service. Health, United

States, 1987. Washington, DC: Government Printing Office, 1988a. ____. Nursing and Related Care Issues as Reported from the 1986 Inventory of Long-Term

Care Places. Huntsville, MD: 1988b. Palmer, H. C., and P. G. Cotterill. "Studies of Nursing Home Costs." In Long-Term

Care Perspectives from Research and Demonstrations. Edited by R. J.

Vogel and H. C. Palmer. Washington, DC: U.S. Department of Health

and Human Services, 1983. Palmer, H. C., and R. J. Vogel. "Models of the Nursing Home." In Long-Term

Care: Perspectives from Research and Demonstrations. Edited by R. J. Vogel

and H. C. Palmer. Washington, DC: U.S. Department of Health and

Human Services, 1983. Scanlon, W. J. "A Theory of the Nursing Home Market." Inquiry 17, no. 1

(Spring 1980): 25-41. Schlenker, R. E. "Case Mix Reimbursement for Nursing Homes." Journal of

Health Politics, Policy and Law 11, no. 3 (Fall 1986): 445-61. ___. "Comparison of Medicaid Nursing Home Payment Systems." Health

Care Financing Review 13, no. 1 (Fall 1991): . Schlenker, R. E., J. D. Stiles, T. Carlough, and P. A. DeVore. A Multi-State

Analysis of Medicaid Nursing Home Payment Systems. Final Report. Vol. 1.

HCFA Cooperative Agreement 18-C-98306/8. Denver, CO: University

of Colorado Health Sciences Center, Center for Health Services

Research, 1988. Smits, H. "Incentives in Case Mix Measures for Long-Term Care." Health

Care Financing Review 6, no. 2 (Winter 1984): 53-59. Stassen, M., and C. Bishop. Incorporating Case Mix in Prospective Reimbursement

for SNF under Medicare: Critical Review of Relevant Research. Waltham, MA:

Brandeis University Center for Health Policy Analysis and Research,

COPYRIGHT 1991 Health Research and Educational Trust
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1991 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Schlenker, Robert E.
Publication:Health Services Research
Date:Dec 1, 1991
Previous Article:Uncompensated care before and after prospective payment: the role of hospital location and ownership.
Next Article:Using medical care: the views and experiences of high-risk mothers.

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters