The Effect of Competition. under Prospective Reimbursement. on Nursing Home Expenditures. John A. Nyman

Transcription

1 The Effect of Competition on Nursing Home Expenditures under Prospective Reimbursement John A. Nyman Thefor-profit nursing home's incentive to minimize costs has been maligned as a major cause of the quality probkms that have traditionally plagued the nursing home care industry. Yet, profit-maximizing firms in other industries are able to produce products of adequate quality. In most other industries, however, firms are constrainedfrom reducing costs to the point where quality suffers by the threat of losing business to competingfirms. In the nursing home industry, competitionfor patients often does not exist because of the shortage of nursing home beds. As a result, one would expect that nursing homes located in areas where there is excess demand would spend less on patient care than homes located where the bed supply is relatively abundant. This hypothesis is tested using Wisconsin datafrom It is found that, in counties with relatively tight bed supplies, an additional empty bed in all the homes in the county willforce each home to increase expenditures by $. 62 per day for each patient in the home. Overall, the average nursing home located in underbedded markets would spend $5.12 more per patient day or about $240, 000 more annually (in 1983 dollars) if it were located in a market where it was forced to compete for patients. The implications for public policy are discussed. Nursing home policy in the 1970s was preoccupied with improving quality. One major cause of the quality problem, as identified by a comprehensive Senate report (1974), was that proprietary nursing homes were lowering the quality of care in order to reduce costs. Since the common form of Medicaid reimbursement in the early 1970s tended to be a flat rate, reducing costs meant greater profit margins. Many conduded from that report that the profit motive is not consis- Address correspondence and requests for reprints tojohn A. Nyman, Ph.D., Assistant Professor, Center for Health Services Research, The University of Iowa, Iowa City, IA

2 556 HSR: Health Services Research 23:4 (October 1988) tent with assurance of adequate quality nursing home care and that the federal government should only approve reimbursement for costs actually incurred. The latter conclusion was embodied in a federal requirement (U.S. Congress 1972) that nursing homes be reimbursed on a "reasonable cost-related basis." While it is true that the profit motive in nursing homes is an incentive to minimize costs, it is also true that firms in most markets face similar incentives. In typical markets, however, if a firm reduces costs to the extent that quality suffers, it will generally lose business to rival firms. Thus, typical firms are prohibited-by the presence of competitors willing and able to serve new customers-from reducing costs to a point where quality is jeopardized. Indeed, this competition is required in free markets to offset the cost-minimization incentive and ensure that the products supplied on the market are of adequate quality. In many of the nursing home care markets of the 1970s (as well as the 1980s), the requisite competition seems to have been lacking because of an excess demand for beds. Scanlon (1979) has argued that excess demand is a pervasive feature of nursing home markets in the United States. Vladeck (1980) alludes to a shortage of beds at many points in his historical review of nursing home policy. When excess demand for beds exists, nursing homes can reduce costs - and thus quality - with impunity because prospective patients (especially Medicaid patients) are forced to accept the first home with an empty bed, regardless of quality. If a shortage exists to the extent that Scanlon argues is true and has persisted to the extent that Vladeck appears to support, then a more fundamental cause of the quality problems that nursing homes traditionally face is a lack of competition caused by an excess demand for beds. If excess demand permits nursing homes to lower costs with impunity, then we would expect that homes located in markets where the bed supply is tight will have lower per unit costs than those located in markets where there is a surplus of beds. Although the health services literature contains many studies that estimate nursing home cost functions (e.g., Ullmann 1985; Schlenker and Shaughnessy 1984; Palm and Nelson 1984; Meiners 1982; Birnbaum et al. 1981; Ruchlin and Levey 1972; Lee and Birnbaum 1983; Bishop 1980; Smith and Fottler 1981; Koetting 1980; Caswell and Cleverly 1983; Walsh 1979; Ullmann 1984; Christianson 1979; Smith et al. 1985; Schlenker 1986), none of them has included an explanatory variable measuring the degree of excess demand in the market in which the nursing home is located.' This article explicitly tests for a relationship between excess

3 Competition and Nursing Home Expenditures 557 market demand and nursing home expenditures using 1983 data from prospectively reimbursed Wisconsin nursing homes. In the first section, the data sets and empirical model are described. In the second, the results are reported. And in the concluding section, the implications for public policy are discussed. DATA AND MODEL Data are from the 1983 Wisconsin Annual Survey of Nursing Homes, which Wisconsin requires nursing homes to complete as part of the annual requirements for Medicaid recertification. This data set contains information on 475 nursing homes (according to the number of unique federal identifiers). These data have been supplemented with information from the 1983 Medicaid cost reports and 1983 data on the violations of Wisconsin's Medicaid certification code. In merging these three data sets, about 125 observations were lost because the identifiers used to link the different data sets did not match. In addition, the data were constrained to include only observations that showed a (meaningful) skilled nursing facility (SNF) Medicaid reimbursement rate. This excluded about 80 observations representing mostly the nursing homes that contained only intermediate care facility (ICF) patients. Ultimately, 269 nursing homes were used in the statistical analysis. Regression analysis was used to test for a relationship between nursing home expenditures and excess demand. The dependent variable was the cost of nursing home care per patient day, that is, average costs. This figure was calculated by adding figures from the four cost centers in Wisconsin's Medicaid cost report: direct care costs, fuel and utility expenditures, property tax or municipal fees, and social services costs, all per patient day. This aggregated cost variable was regressed on the 13 independent variables described below. Excess demand, the first independent variable, was not measured directly since the data set did not contain information on the number of different patients wishing to enter homes in a market. Instead, the average number of empty beds in the county in which the nursing home was located was used as a proxy measure for excess demand. This variable was intended to capture the relative need to compete for patients based on exogenous market conditions. Homes located in counties where all homes were completely occupied would not need to compete for patients, but homes located in counties where there were many empty beds on average would need to compete. The average number of empty beds, rather than the average per-

4 558 HSR: Health Services Research 23:4 (October 1988) centage, was used because the number of empty beds may be a better measure of the need to compete from the perspective of the total number of beds available in the market. For example, one county might have five 50-bed homes with 2 empty beds in each, and another might have five 500-bed homes with 20 empty beds in each. Although the percentage of empty beds is the same for homes located in each county, 4 percent, the total number of empty beds available, 10 and 100 respectively, differs significantly. Other things being equal, we would expect it to be easier for firms to find patients in the former county than in the latter; thus, nursing homes located in the former county would be less likely to feel the need to compete for patients. The average number of empty beds, 2 and 20 respectively, would reflect this difference, whereas the average percentage of empty beds would not. It is not necessary that all beds be filled for excess demand to exist. Normal turnover means that some beds are empty regardless of demand conditions. Furthermore, nursing homes may leave some beds empty rather than filling them with Medicaid patients, because they want to be able to accommodate the more lucrative private patients (Bishop 1979). The mean and standard deviation of the average county occupancies for these homes is about 6.5 and 3.75 empty beds, respectively. The mean number of actual beds is about 135. Given that a substantial number of nursing homes are located in counties that averaged less than three empty beds per home, it would be reasonable to assume that excess demand did exist in at least some Wisconsin counties in The average number of empty beds in the county as a measure of the need to compete for patients is a characteristic of the market environment in which the firm is located. An individual home's excess capacity will contribute to this variable, but the number of empty beds in an individual home is likely to be different, either larger or smaller, than the average excess capacity of all the homes in the county. Nevertheless, the two variables may be correlated to some extent, and a greater number of empty beds in the individual homes might directly contribute to increased costs per patient day because the fixed costs in such homes are spread over fewer patients (and patient days). It is, therefore, necessary to control for this factor in order to ensure that the average number of empty beds truly represents an exogenous market variable. Accordingly, a variable representing the number of empty beds in the individual home was included in the regression. In the following discussion, the average number of empty beds in the county will be referred to as "average excess capacity," while the number of

5 Competition and Nursing Home Expenditures 559 empty beds in an individual home will be called simply 'excess capacity." The dependent variable, costs per patient day, will vary with the quality of care provided and the dependency of the patients served. In order to control for cost differences along these dimensions of the output, measures of quality and case mix were included in the regression. Quality is represented by the number of violations of the Medicaid code weighted according to severity. The severity weights represent the relative magnitudes of the maximum fines-1, 10, and 50-that are associated with each violation type- C, B, and A, respectively. The characteristics of quality reflected by the 1983 violations measure, however, may not be representative of all of the qualitydefining characteristics of the nursing home. This is because the fines associated with the violation characteristics force the home to be more concerned about these characteristics than about the many others that patients might find desirable. For example, a home is more likely to have the requisite staffing ratios than to have staff who treat patients with kindness, because the former characteristic is subject to fine whereas the latter one is not. Therefore, we cannot assume that we have adequately controlled for quality of care by having included this variable in the regression. Case-mix differences were measured by two variables: the home's average activities of daily living (ADL) score and the average length of stay of patients in the home. The home's average ADL score was calculated by adding the number of patients who are dependent in each of the eight ADL categories (bathing, bowel continence, urine continence, mobility, dressing, feeding, toileting, and transferring) and dividing by the number of patients in the home. For example, a value of 8 for this variable would represent a home with all patients dependent in all eight categories. Average length of stay, the second case-mix variable, was constructed by taking the number of patients residing in the home less than one year, the number of patients in the home between one and two years, between two and three years, three and four years, four and five years, and over five years, and multiplying each number by.5, 1.5, 2.5, 3.5, 4.5, and 5.5, respectively. These products were then summed and divided by the total number of patients, yielding a variable that measures the average length of stay of patients in the home on a given date. Homes that have a larger number of short-stay patients may have more acutely ill patients whose health can be expected to improve. As patients improve, they consume fewer care resources; therefore, homes with more of these kinds of patients may have lower

6 560 HSR: Health Services Research 23:4 (October 1988) costs. Because chronic patients do not get better, it is expected that the longer the average length of stay, the greater the number of chronic patients and the greater the costs. Average costs are assumed to vary with the size of the output. We adopt the convention of the health services cost-function literature and measure output by the number of beds in the home rather than by the number of patient days in the year or the average number of patients served. This is a useful convention because it allows for the direct estimation of the optimal bed size in nursing homes. It is expected that a U-shaped relationship will exist; therefore, a squared term was also introduced into the regression in order to accommodate this functional form Ṫhe eighth explanatory variable is a dummy variable representing whether the nursing home was or was not chartered as a for-profit nursing home. For-profit nursing homes have explicit reasons for minimizing costs and are therefore likely to have lower costs than nonprofit homes. Nonprofits are constrained to have costs equal revenues in the long run. Nevertheless, in the short run, nonprofits may desire revenue surpluses in order to build up a fund for an investment, such as a capital improvement, that would enhance the prestige of the nursing home or achieve some other objective. If nonprofit firms do want surpluses, then their desire to minimize costs may differ only qualitatively from the similar desire in for-profits. They may, therefore, take advantage of the same market failures that for-profits do, although to a lesser extent. The Medicaid reimbursement rate was included as another explanatory variable. The 1983 reimbursement rates for Wisconsin nursing homes were based on the 1981 costs of the home, adjusted forward to account for inflation (State of Wisconsin 1983, 1984). Therefore, the rate is prospective and exogenous, since the individual nursing home cannot influence the size of its rate in the present period. The reimbursement rate is induded in order to determine what proportion of an additional Medicaid dollar is spent on patient care. The tenth explanatory variable is the percentage of Medicaid patients in the home. Under excess-demand conditions, there are two reasons why costs are expected to be lower in homes with more Medicaid patients. First, nursing homes can choose among patients when the bed supply is tight; given that choice, they will admit the more lucrative private patients first, then the less dependent (and therefore less costly) of the Medicaid patients, until they run out of beds. A home with a larger percentage of Medicaid patients will, therefore, exhibit relatively lower costs. Second, with excess demand, nurs-

7 Competition and Nursing Home Expenditures ing homes need to compete only for the higher-paying private patients. They can attract Medicaid patients with only minimal-quality care. Homes with a higher percentage of private patients will, therefore, exhibit higher quality. To the extent that differences in expenditures reflect differences in quality, homes with a greater percentage of Medicaid patients will spend less on patient care.2 Three additional variables were included in the regression: the percentage of SNF patients in the home, a dummy representing whether or not the home was located in the Milwaukee area, and a final dummy representing whether the nursing home was operated in conjunction with a hospital. SNF patients have greater staffing requirements than other patients. As a result, we would expect that as the percentage of SNF patients increases, so will average costs. Homes located in urban areas such as Milwaukee face higher wage and other input prices; therefore, they are likely to have higher costs. Finally, homes operated in conjunction with hospitals may use upgraded inputs (e.g., registered nurses instead of licensed practical nurses); thus, their costs are expected to be higher than those operated without a hospital connection. RESULTS 561 Table 1 lists the means and standard deviations of the variables included in the regression. The regression results are reported in Table 2. The R2 indicates that the variables included in the regression explained over three-quarters of the variation. As expected, the costs were greater for those nursing homes located in counties where the average number of empty beds was greater. The coefficient, significant at the 5 percent level, indicated that, for every additional empty bed in all the county's homes, each home in the county would experience a S. 19 increase in costs per day for every patient in the home. This was the case controlling for the effect that the excess capacity of an individual home might have on costs. Neither the violations nor the ADL score coefficients had the expected sign, but neither was significant. The coefficient of the average length-of-stay variable, however, had the expected sign and was significant. The number of beds and beds-squared variables showed signs representing the conventional U-shaped average cost curve, and both were significant. According to these coefficients, the minimum of the long-run average cost curve occurs at about 272 beds. Proprietary homes have costs that are about $4.15 lower than those of nonprofit

8 562 HSR: Health Services Research 23:4 (October 1988) Table 1: Means and Standard Deviations (N = 269) Variabk Mean Standard Deviation Total costs per patient day Direct patient care costs per patient day Average excess capacity of homes in county Excess capacity of home Total violations weighted by severity Average ADL score of patients in home Average length of stay of patients in home Number of beds in home Number of beds in home squared For-profit status (- 1) Medicaid reimbursement rate Percentage of Medicaid patients in home (in percentage points) Percentage of SNF patients in home (in percentage points) Milwaukee-area location (- 1) Operated with hospital (- 1) homes, and the Medicaid reimbursement rate was the most significant predictor of average costs, as might be expected. The percentage of Medicaid patients was negatively and significantly related to costs. The coefficient indicates that, as the percentage of Medicaid patients increases from 0 to 100 percent, the costs per patient day decline by about $7. Of the three remaining variables, only the Milwaukee dummy variable was insignificant, although it showed the expected sign. The coefficient for the percentage SNF patients variable indicates that as the home's percentage of SNF patients increases from 0 to 100 percent, costs per patient day increase by about $ Nursing homes operated in conjunction with a hospital have about $6 higher costs per patient day. The relationship between costs and the average number of empty beds may not be linear. There may be some threshhold level of empty beds beyond which additional empty beds will not add to the need to compete for patients. This was tested by dividing the data set into two parts according to the average excess capacity variable and running the regression separately on each subsample. This was tried at three different cutoff levels (average excess capacities of 5.5, 6.5, and 7.5), and all pairs of regressions yielded similar results.

9 Competition and Nursing Home Expenditures Table 2: Regression Results -All Observations (N = 269) Variabk Coefficient Standard Error Significance Level Intercept Average excess capacity of homes in county Excess capacity of home Total violations weighted by severity Average ADL scores of patients in home Average length of stay of patients in home Number of beds in home Number of beds squared For-profit status ( = 1) Medicaid reimbursement rate Percentage of SNF patients in home Percentage of Medicaid patients in home Milwaukee area location (= 1) Operated with hospital (- 1) R2 =.79 F-value = Dependent variable: average costs (costs per patient day) 563 Results appear in Table 3 for the regressions run on those observations in counties with fewer than 6.5 empty beds on average. Results for counties averaging 6.5 empty beds or more appear in Table 4. Taking first the regression run on observations where the bed supply is tight (<6.5 average excess capacity), the coefficient for the average empty-bed variable is again significant, but now over three times larger than in the all-observation case. This indicates that for every additional empty bed in this range, costs increased by $.62 per patient day. The assumption of linearity may be unsupportable since the values for the same coefficient in the other regressions seem to indicate that as the sample's bed supply becomes tighter, the effect of one additional empty bed on costs becomes greater. The analogous coefficients in the 5.5 and 7.5 cases were.83 and.44, respectively, both significant at the 10 percent level or better.

10 564 HSR: Health Services Research 23:4 (October 1988) Table 3: Regression Results -Tight Bed Supply Observations (Average Excess Capacity Less Than 6.5 Beds) (N = 165) Variabk Coefficint Standard Error Significance Level Intercept Average excess capacity of homes in county Excess capacity of home Total violations weighted by severity Average ADL scores of patients in home Average length of stay of patients in home Number of beds in home Number of beds squared For-profit status (= 1) Medicaid reimbursement rate Percentage of SNF patients in home Percentage of Medicaid patients in home Milwaukee area location (= 1) Operated with hospital (= 1) R2 =.80 F-value = Dependent variable: average costs (costs per patient day) The regression results from the > 6.5 empty beds subsample show that the average excess capacity variable is now insignificant. This is also true of the analogous cases for the 5.5 and 7.5 cutoff subsamples. This may indicate that there are homes located in counties where the average excess capacity is so great that differences in that variable do not reflect differences in the homes' need to compete. For example, the need to compete for patients may be virtually the same for homes located where the average excess capacity is 20 as where the average excess capacity is 15. Because the subsample is likely to contain these homes, a weaker relationship exists. This would seem to indicate that a threshhold level of competition can be reached, beyond which increases in the average number of empty beds in a county have a

11 Competition and Nursing Home Expenditures Table 4: Regression Results - Surplus Beds Observations (Average Excess Capacity at Least 6.5 Beds) (N = 104) Variable Coefficient Standard Error Significance Level Intercept Average excess capacity of homes in county Excess capacity of home Total violations weighted by severity Average ADL scores of patients in home Average length of stay of patients in home Number of beds in home Number of beds squared For-profit status (= 1) Medicaid reimbursement rate Percentage of SNF patients in home Percentage of Medicaid patients in home Milwaukee area location (= 1) Operated with hospital (= 1) R F-value = Dependent variable: average costs (costs per patient day) 565 negligible effect on costs. This analysis, however, gives us only sketchy information regarding the location of that threshhold. The average ADL score variable shows an interesting reversal of signs across the two regressions. In the regression where the bed supply is tight, an increase in the average ADL scores of the homes leads to significantly lower costs, while the same coefficient where the availability of beds is greater indicates that increases in the dependency of patients leads to the expected finding of significantly increased costs. One possible explanation is that nursing homes can choose among patients when the bed supply is tight. Those homes that choose high ADL cases may choose the exceptional cases that are not as costly. For example, a comatose patient may require less attention than one who has the same number of dependencies (or even fewer), but who is also

12 566 HSR: Health Services Research 23:4 (October 1988) conscious. Homes located in areas where beds are available cannot pick and choose among the more dependent patients; therefore, their costs reflect the generally positive relationship between ADLs and cost that would be expected on average. The percentage Medicaid variable shows a similarly interesting loss of significance in the regression where beds are generally available. This is again consistent with the hypothesis that homes can choose among Medicaid patients where the bed supply is tight, choosing the patients that require the least costly care. When there are many beds available, the nursing home cannot be as selective and must take Medicaid patients in the order that they apply for admission. This means that the relationship between cost and percent Medicaid patients, though again negative, is apparently less systematic than when the nursing homes are able to choose. This explanation would hold only to the extent that other variables do not adequately control for the effects on costs of case-mix differences. The loss of significance in the percent Medicaid variable is also consistent with the hypothesis that, when nursing homes do not need to compete for patients, it is the Medicaid patients who suffer most from the lack of competitive quality care. Since private patients are a more lucrative source of revenue, nursing homes prefer them to Medicaid patients. When there is a shortage of beds, nursing homes will only compete for the private patients. As a result, nursing homes catering to a more private clientele will have higher expenditures than homes catering primarily to Medicaid patients. When there is a surplus of beds, however, nursing homes must compete for both types of patients and, in that case, lower expenditures per patient day will not necessarily be associated with Medicaid-dominated homes.3 It should also be noted that the number of weighted violations is directly related to costs when the bed supply is tight, and that the expected inverse relationship exists when there are beds available. Perhaps the direct relationship indicates that homes with more violations have greater costs because of fines and the costs associated with corrective measures. This pattern again obtained for the two other pairs of subsetted regressions. Finally, the size of the Medicaid reimbursement rate coefficient changed across the two regressions. When the bed supply is tight, an additional dollar of reimbursement results in an additional S.81 spent on patient care; however, when ample beds are available, an additional dollar of reimbursement results in $1.04 more of expenditures, ceteris paribus.4 If this difference is significant, it indicates that the taxpayer's Medicaid dollars are more likely to be spent on patient care when there

13 Competition and Nursing Home Expenditures 567 is competition for patients. A Chow test (Kmenta 1971, 373) was performed and showed that the structure of the two regressions is significantly different at the 5 percent level, indicating that differences in any of the coefficients across the two regressions must be taken seriously. The declining coefficients of the average excess capacity variables in the less than 5.5, 6.5, and 7.5 average empty-bed subsetted regressions may indicate a nonlinear relationship between costs and the average number of empty beds. Accordingly, the natural logarithm of the average empty-beds variable (plus 1) was substituted into the allobservation regression. The resulting coefficient for that variable was 1.70, significant at the 1 percent level. Although the fit was slightly better and the average excess-capacity variable was more significant, the results of this regression were not appreciably different from the original all-observation results and are therefore not reported. CONCLUSION INTERPRETATION OF RESULTS These findings appear to support the hypothesis that a lack of competition caused by an excess demand for beds was an important reason why costs were allowed to drop low enough to jeopardize the quality of nursing home care under prospective and flat-rate reimbursement systems. Indeed, the importance of the effect of excess demand on costs rivals that of the profit-maximization incentive itself. It was found that for-profit nursing homes have $4.15 lower costs than nonprofits. If excess demand were to decrease from a level represented by 0 average empty beds to a lower level represented by 6.5 average empty beds, and if a linear relationship between costs and average empty beds is assumed, each nursing home would have about $4.03 ($.62 x 6.5) greater costs.5 The magnitude of the effect of reducing excess demand to this extent ($4.03), therefore, is similar to the magnitude of the cost reduction from having a for-profit rather than a nonprofit charter ($4.15). Although similar in magnitude, the lower expenditures in forprofit nursing homes may have less sinister implications for quality than the lower expenditures of homes located in underbedded markets. Lower costs in for-profit firms may stem from differences in efficiency between them and nonprofit firms. To the extent that reduced expenditures in proprietary firms represent efficiency gains, they are desirable.

14 568 HSR: Health Services Research 23:4 (October 1988) Although this only represents a possibility, it is difficult to ascribe any similarly positive reason for nursing homes in underbedded markets to have lower costs. Therefore, the effect of excess demand on quality may be greater than the effect of the firm's charter status on quality, even though the effect of both on firm expenditures is the same. Furthermore, the effect of the excess demand is not simply reflected in the size of the average excess-capacity coefficient. The subsetted regressions are distinguished by the kvel of the average excess-capacity variable. Therefore, differences in the coefficients across the two subsetted regressions may also reflect the effect of excess demand through other variables. For example, the Medicaid reimbursement variable coefficients show that, for every additional dollar increase in the rate, homes in underbedded markets spend about S.23 less ($ $.81) on patient care than homes located where the average excess capacity is higher. Also, homes with 100 percent Medicaid patients spend about $3.80 less ($ $4.12) on patient care in underbedded markets than do 100 percent Medicaid homes in markets with more empty beds. The differences in the coefficients of these two variables are consistent with the theoretical expectations about the effect of excess demand on their relationship with costs. Consequently, these differences would seem to indicate that the effect of excess demand on costs is considerably greater than the effect captured in the average excess-capacity variable's coefficient alone. Therefore, even the magnitude of the effect of excess demand on costs may be significantly greater than the effect of the home's charter status on costs. One way to estimate the entire effect of excess demand on expenditures is to compare average costs across the two groups. The mean average cost for the underbedded observations is $41.25, while the nursing homes located where more empty beds are available had expenditures of $43.62 per patient day. This comparison, however, does not control for the differences in the characteristics of nursing homes across the two subsamples. To hold these constant, the characteristics of the average home in the underbedded sample (i.e., the mean levels of the regression variables for these homes) were multiplied by the regression coefficients of the homes in the surplus bed sample. The sum of these products is $46.37; it represents the costs per patient day of the average nursing home in the sample where the mean average excess capacity is 4.4 beds if this home were located in a market where the mean average excess capacity is 10.3 beds. In other words, by relocating the average home in a market with excess demand to a market where there is more competition for patients, this nursing home would be forced to spend $5.12 (or about 12.4 percent) more per day

15 Competition and Nursing Home Expenditures 569 for every patient in the home. Since the average number of patients in a home is about 128, this amounts to an average annual increase in total expenditures of almost $240,000 ($5.12 x 128 patients x 365 days, or $239,206 in 1983 dollars) per nursing home. EXPENDITURES AND QUALITY This article has implied throughout that a relationship exists between the need to compete as measured by excess demand and the quality of care provided by a home. Clearly, the empirical findings of an association between average excess capacity and expenditures found are consistent with this relationship. Other studies (Nyman 1985, 1988) have found evidence of a connection between the excess demand and a direct measure of quality. In those studies, the weighted number of violations was shown to be greater (and sometimes significantly so) for homes located in counties with smaller average excess capacities. These earlier studies have an advantage over this work in that they use Wisconsin data from 1979, the first year that Wisconsin collected violation data. It is likely that, because of unfamiliarity with the new quality assurance system, violations data from 1979 were more representative of all possible quality-defining characteristics of the homes. Therefore, the 1979 violation variable may represent the true "quality" of care provided by the nursing homes better than the 1983 variable, since homes may have learned by 1983 to pay closer attention to those quality-defining characteristics for which they could be cited and fined. The disadvantage of the earlier studies, however, is their reliance on a direct measure of quality, a construct that is difficult to measure. Clearly, no consensus has yet been achieved within the health services research community regarding the identity of those characteristics that constitute "quality" nursing home care. Consequently, any measure of quality is open to criticism. This is not a problem here because the meaning of the dependent variable is not in doubt. Yet average total costs may reflect some expenditures that do not contribute to the quality of patient care. It is interesting that, when the same analyses were performed with direct patient care costs as the dependent variable, the same general pattern of results prevailed. For every additional empty bed on average in the market, direct patient care expenditures rose by $.16. If located in a relatively underbedded market, expenditures would rise by $.38; if located where average excess capacity is relatively large, expenditures would rise by only $.02, and insignificantly. Likewise, an additional dollar of reimbursement would result in $.49 of direct patient care expenditures for homes

16 570 HSR: Health Services Research 23:4 (October 1988) located in underbedded markets and S. 70 for homes located where there are more empty beds. Overall, if the average nursing home from the underbedded sample was located where more beds were available, that home would incur $2.32 more in direct patient care costs per patient day or about $110,000 more per year. Of the variables that can be linked to the need to compete for patients, only the percentage of Medicaid patients showed a coefficient with a magnitude inconsistent with earlier results. Still, this coefficient was insignificant. Despite this one failing, the results from the direct patient care regressions generally reinforce the earlier findings and are consistent with a link between the lack of need to compete for patients and poor-quality care. IMPLICATIONS FOR POLICY Government policies are at least partially responsible for the creation of excess demand. Almost all states have enacted certificate-of-need legislation and at least 12 states have imposed construction moratoria on nursing home beds in recent years (Swann and Harrington 1986). These policies have constrained the supply of beds in the face of demand increases caused by an aging population and by the incentive embodied in prospective reimbursement of hospitals to discharge patients earlier and in more dependent conditions. The most direct policy response to this problem, and probably the most appropriate one, is to eliminate excess demand. This could be accomplished by either increasing the supply of beds or constraining demand. Increasing the bed supply does not necessarily mean eliminating certificate-of-need completely, although this measure would no doubt help. Governments have competing goals, and one of these goals is the containment of costs, especially the costs of the Medicaid program. In light of cost considerations, perhaps a better plan would be to determine first an optimal level of excess supply - that is, one that ensures that nursing homes will be forced to compete for patients - and to direct certificate-of-need policy toward achieving that goal. Alternatively, excess demand could be eliminated by restricting access to nursing home care to those Medicaid patients who can meet a more rigorous set of admission standards. If about the same number of patients are served or fewer, this may mean no substantial increase in total costs. Additionally, such a restriction would ensure that nursing homes could not choose to admit only the lighter-care patients, since they presumably would be ineligible for care. The disadvantage of restricting demand is that some people who are currently deemed deserving of Medicaid benefits would (still) be disenfranchised of those

17 Competition and Nursing Home Expenditures benefits. Nevertheless, the government would choose which prospective patients would be disenfranchised under this policy, as opposed to the nursing homes, which are able to choose which patients to exclude under excess demand. Clearly, alternative modes of care would need to be found for the disenfranchised, and this would mean an increase in total costs. In recent years, a number of states (such as Wisconsin) have converted from retrospective to prospective systems. Currently, caseadjusted prospective systems are gaining popularity. Under all prospective or flat-rate systems, the increase in nursing home expenditures that would accompany an increase in competition for patients would not be borne by a direct increase in the costs of Medicaid programs, as it would have under a retrospective system. These increases in costs would simply translate into decreased profit margins per patient day in the for-profit homes. In the nonprofit homes, the increased competition may mean either smaller revenue surpluses in the short run or greater productivity in the long run as the same workers are asked to work harder in order to provide competitive-quality care with existing financial resources. Perhaps a better interpretation of this shrinking profit (surplus) margin and increased productivity would be that a larger percentage of the Medicaid reimbursement payment is being used to produce adequate-quality patient care. This form of efficiency in the use of government funds should be seen as desirable - a standard against which the success of state Medicaid programs can be evaluated. ACKNOWLEDGMENTS Helpful comments on an earlier draft of this article were received from James Rohrer and two anonymous referees. Michael McKillip and Debra Magrosky provided able assistance in managing the data set. The author, of course, is responsible for any errors or oversights. NOTES A number of these studies have included a variable measuring the number or percentage of empty beds in the individual homes. This variable is generally included to test whether an increase in the number of empty beds increases fixed costs per patient day, and thus total costs per patient day. Nowhere is this variable included as a measure of the need to compete for patients. Indeed, information on the occupancy of an individual firm pro-

18 572 HSR: Health Services Research 23:4 (October 1988) vides little information on the need to compete for patients in the market. For example, a completely occupied nursing home might reflect either the firm's having successfully competed (if there are a great number of empty beds in other nursing homes in the market) or not having had to compete at all (if all the firms in the market are full and had waiting lists). 2. If this were a retrospectively reimbursed state, we would have the opposite expectation. Under retrospective reimbursement, all costs incurred are reimbursed. Therefore, the incentive under that system is to increase expenditures, but not necessarily by purchasing inputs that improve patient care. Managers may benefit from increased expenditures if their job is made easier, more prestigious, or more enjoyable. Managers will act on these incentives only to the extent that they have Medicaid patients. With private patients, they can still maximize profits and thus are interested in minimizing costs. Therefore, under retrospective reimbursement, average cost increases as the percentage of Medicaid patients increases. 3. Note that this relationship would be difficult to attribute to lower reimbursement rates constraining the level of quality that can be provided by the home, since the reimbursement rate is controlled for in both equations and since the difference between the equations is based on the average excess capacity variable and not on the reimbursement rate. Attributing this relationship to lack of competition and not inadequate reimbursement rates is consistent with other research. Nyman (1988) shows evidence that, when located in underbedded markets, 100 percent Medicaid homes have over four times as many weighted violations as when they are located where beds are relatively available. This is true, again controlling for the level of the reimbursement rate. 4. Marginal spending that exceeds marginal Medicaid revenue suggests that nursing homes cross-subsidize only when they are forced to by competition. Rather than viewing cross subsidization as a payment by private patients of part of the Medicaid patients' bill because the reimbursement rate is too low, this finding suggests that cross subsidization is a natural consequence of competition that might occur even if reimbursement rates were raised significantly. 5. Even at this level, excess demand may not be eliminated. REFERENCES Birnbaum, Howard, Christine Bishop, A. James Lee, and Gail Jensen. "Why Do Nursing Home Costs Vary? The Determinants of Nursing Home Costs." Medical Care 19 (November 1981): Bishop, Christine E. "Nursing Home Behavior under Cost-Related Reimbursement." Discussion Paper No. 9. Waltham, MA: University Health Policy Consortium, "Nursing Home Cost Studies and Reimbursement Issues." Health Care Financing Review 5 (Spring 1980): Caswell, Robert J., and William 0. Cleverly. "Cost Analysis of the Ohio Nursing Home Industry." Health Services Research 18 (Fall 1983):

19 Competition and Nursing Home Expenditures 573 Christianson, Jon B. "Long-Term Care Standards: Enforcement and Compliance." Journal of Health Politics, Policy and Law 4 (Fall 1979): Kmenta, Jan. Ekments of Econometrics. New York: Macmillan, 1971, 373. Koetting, Michael. Nursing-Home Organization and Efficiency: Profit versus Nonprofit. Lexington, MA: Lexington Books, Lee, A. James, and Howard Birnbaum. "The Determinants of Nursing Home Operating Costs in New York State." Health Services Research 18, Part II (Summer 1983): Meiners, Mark R. "An Econometric Analysis of the Major Determinants of Nursing Home Costs in the United States." Social Science and Medicine 16 (1982): Nyman, John A. "Excess Demand, the Percentage of Medicaid Patients and the Quality of Nursing Home Care." Journal ofhuman Resources 23 (Winter 1988): "Prospective and 'Cost-Plus' Medicaid Reimbursement, Excess Medicaid Demand, and the Quality of Nursing Home Care."Journal of Health Economics 4 (September 1985): Palm, David W., and Scott Nelson. "The Determinants of Nursing Home Costs in Nebraska's Proprietary Nursing Homes." Socioeconomic Planning Sciences 18, no. 3 (1984): Ruchlin, Hirsch S., and Samuel Levey. "Nursing Home Cost Analysis: A Case Study." Inquiry 9, no. 3 (1972):3-15. Scanlon, William J. "The Market for Nursing Home Care: A Case for an Equilibrium with Excess Demand as a Result of Public Policy." Ph.D. diss., University of Wisconsin, Madison, Schlenker, Robert E. "Case Mix Reimbursement for Nursing Homes."Journal of Health Politics, Policy and Law 11 (Fall 1986): Schlenker, Robert E., and Peter W. Shaughnessy. "Case Mix, Quality, and Cost Relationships in Colorado Nursing Homes." Health Care Financing Review 6 (Winter 1984): Smith, Howard L., and Myron D. Fottler. "Costs and Cost Containment in Nursing Homes." Health Services Research 16 (Spring 1981): Smith, Howard L., Myron D. Fottler, and Borje 0. Saxberd. "Does Prospective Payment Encourage Nursing Home Efficiency?" Evaluation and the Health Professions 8 (June 1985): Swan, James H., and Charlene Harrington. "Estimating Undersupply of Nursing Home Beds in States." Health Services Research 21 (April 1986): Ullmann, Steven G. "Cost Analysis and Facility Reimbursement in the Long- Term Health Care Industry. Health Services Research 19 (April 1984): "The Impact of Quality on Cost in the Provision of Long-Term Care." Inquiry 22 (Fall 1985): U.S. Congress. Senate. Committee on Finance. The Social Security Amendments of 1972: Report to Accompany H.R.1., 92d Cong., 2d sess. S. Rept : U.S. Congress. Senate. Special Committee on Aging. Subcommittee on Long-Term Care. Nursing Home Care in the United States: Failure in Public Policy, 93rd Cong., 2d sess., 1974:

20 574 HSR: Health Services Research 23:4 (October 1988) Vladeck, Bruce C. Unloving Care: The Nursing Home Tragedy. New York: Basic Books, 1980, 12 et passim. Walsh, ThomasJ. "Patient-Related Reimbursement for Long-Term Care." In Reform and Regulation in Long-Term Care. Edited by Valerie LaPorte and Jeffery Rubin. New York: Praeger, 1979, Wisconsin, State of. Department of Health and Social Services. Methods of Implementation for the Nursing Home Reimbursement Formula for January 1 throughjune 30, 1983 as Approved by thejoint Committee on Finance on March 16, Madison, WI: Department of Health and Social Services, Methods ofimplementationfor the Nursing Home Reimbursement Formulafor July 1, 1983 through June 30, 1984 as Approved by the Joint Committee on Finance and as Approved by the Governor. Madison, WI: Department of Health and Social Services, 1984.