| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
|
|
|
|||||||
|
|||||||||
Correspondence: Address correspondence to Orna Intrator, PhD, Center for Gerontology and Health Care Research, Brown University, Box G-ST, Providence, RI 02912. E-mail: Orna_Intrator{at}brown.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Key Words: Medicaid nursing home reimbursement • Nursing home competition • Medical infrastructure • Managed care organizations
Medicaid is the dominant purchaser of nursing home services in the United States (accounting for roughly 50% of all nursing home expenditures and 70% of all bed days). Reimbursement rates may have a major impact on the ability of nursing homes to invest in the medical infrastructure necessary for the provision of quality. In addition, competition, particularly for private-pay admissions, may provide an incentive for investment. Our purpose in this article is to determine whether Medicaid reimbursement rates and market competition bear on a specific investment in medical infrastructure: the employment of nurse practitioners (NPs) and physician assistants (PAs) in nursing homes.
The Impact of NPs and PAs on the Quality of Nursing Home Care
Strategies to improve the primary care delivered in nursing homes and to reduce emergency room visits and hospital admissions frequently involve the use of geriatric teams that include NPs (Bourbonniere & Evans, 2002; Fama & Fox, 1997; Ouslander, 1989; Wieland, Rubenstein, Ouslander, & Martin, 1986) and PAs (Ackermann & Kemle, 1998; Tideiksaar, 1984). Many of these studies suggest that NPs and PAs can improve the quality of care provided to nursing home residents, or provide a level of care comparable with that of physicians (Garrard et al., 1990; Kane et al., 1989, 1991; Wieland et al.). NPs and PAs improve communication and morale between nursing home staff and physicians (Fama & Fox; Ouslander; Wieland et al.). They also appear to be cost effective (Ackermann & Kemle; Kane et al., 1991; Wieland et al.), because they use fewer laboratory tests and medications and lower facilities' hospitalization rates (Ackermann & Kemle; Garrard et al.; Intrator, Castle, & Mor, 1999; Joseph & Boult, 1998; Kane et al., 1989, 1991).
NPs and PAs have been used in nursing homes to provide mental health services through staff education and counseling (Kennedy, Covington, Evans, & Williams, 2000; Miller, 1997), to influence general care (Shaughnessy, Kramer, Hittle, & Steiner, 1995; Smith, Mitchell, Buckwalter, & Garand, 1995), and to improve specific outcomes for newly admitted nursing home residents (Kane et al., 1988). Kane, Flood, Keckhafer, and Rockwood (2001) described the practice of nurse practitioners in the EverCare model. Several studies have demonstrated the effectiveness of NPs and PAs and physician teams managed by HMOs (Fama & Fox, 1997; Farley, Zellman, Ouslander, & Reuben, 1999; Joseph & Boult, 1998; Kane, Flood, Bershadsky, & Keckhafer, 2004; Kennedy et al.; Miller). Thus, it appears that NP–PA teams are effective in providing quality care.
The availability of NPs and PAs in nursing homes has been discussed within the HMO setting. Farley and colleagues (1999) reported that 38% of HMOs studied used formal primary care programs incorporating NPs or PAs in the care of their nursing home residents. However, both NPs and PAs remain scarce commodities. The American Academy of Nurse Practitioners (AANP), the certification authority for these professionals, reported that, in 2001, there were about 65,000 NPs in the United States. Only 8% of them were employed in geriatrics, and 10% were employed in institutions (hospitals and nursing homes; AANP, 2001). In 2002, there were about 57,000 PAs practicing in the United States. According to a survey of PAs, almost 90% of them practice with patients 65 years old or older, but less than 0.4% reported working in a nursing home (American Association of Physician Assistants [APA], 2002).
Facilities may choose to have NPs or PAs available by employing them or contracting for them. NPs or PAs also may be available on site to patients if they are allowed to practice within the facility but are not paid by the facility. If a facility chooses to hire or to contract, it commits to an investment in medical infrastructure, and the NPs or PAs are likely to be available to all residents. When NPs or PAs are only available on site through outside sources (such as for managed care residents), the facility may invest managerial time in establishing a collaborative arrangement with an NP, PA or physician practice employing an NP or PA, but the NPs or PAs are likely to be available only to residents of the employing physicians, health plans, or agencies such as in the EverCare model (Kane & Huck, 2000). Therefore, it is likely that those nursing homes that employ NPs or PAs, either on staff or by contract, make a significant investment in quality. In this article we investigate the facility and market characteristics associated with this investment.
Theory and Hypotheses
Here we view the use of NPs or PAs as a strategic investment in medical infrastructure. Resource dependency theory can explain factors that influence organizational investment decisions. This perspective argues that the scarcity of resources and the uncertainty of the environment are constraints on organizational decision making (Scott, 1998). It has been used as a framework in several other studies of nursing home resource allocation decisions. In particular, resource dependency theory was used to study the provision of postacute and dementia care in special care units (Banaszak-Holl, Zinn, & Mor, 1996; Zinn & Mor, 1998; Zinn, Mor, Castle, Intrator, & Brannon, 1999). By extension, it is an appropriate perspective for the study of resource investment decisions, such as the use of NPs or PAs. In this article we use resource dependency theory to generate hypotheses regarding the behavior of freestanding facilities in urban markets. We exclude hospital-based and government-affiliated facilities because the organizational mechanisms are not independent in these facilities, specifically regarding NPs or PAs. The hypotheses we develop here also do not pertain to facilities in rural markets because of the difference in market structures in terms of availability of resources and alternative care settings.
In making investment decisions, nursing homes must take into account the resources available from the environment. As previously argued, Medicaid is the dominant purchaser of nursing home services in the United States. Moreover, Medicaid reimbursement rates are not set based on nursing home performance, and thus they are exogenous to the study of NP–PA availability in nursing homes. Therefore, our first hypothesis (H1) is as follows: Facilities in states with higher Medicaid reimbursement rates are more likely to hire or contract for NPs or PAs.
Among the critical resources needed by nursing homes are patient referrals, particularly private payers. Although managed care is not yet a major force in long-term-care reimbursement, managed care organizations are increasing their direct contracting with nursing homes (Zinn, Mor, & Gozalo, 2000b). In order to secure contracts, facilities must demonstrate that resources, such as NPs or PAs, are available to provide subacute care. Facilities that operate in a more competitive market for referrals may employ NPs or PAs in order to signal quality to potential referral sources. Therefore, our second hypothesis (H2) is that nursing facilities will be more likely to employ NPs or PAs in markets with higher managed care market penetration, and our third (H3) is that facilities operating in more competitive markets will be more likely to employ NPs or PAs.
Other factors also may influence resource investment decisions. The availability of resources in the nursing home should relate to the number (or proportion) of private-pay residents, because private-pay rates are considerably higher. Furthermore, facilities that have a resident case mix of higher acuity, especially those caring for more Medicare recipients who usually require subacute or postacute care (Zinn, Mor, & Gozalo, 2000a), will require more proficient staff such as NPs or PAs to care for patients requiring more medical care.
If NPs or PAs are a scarce resource, then they are likely to be more selective in the type of environment in which they choose to work. In markets where more NPs or PAs are available, it is likely that more of them will be employed in nursing homes, even though nursing homes are considerably less favorable working environments. Facilities that have made previous investments in medical infrastructure such as employing more nurses (Harrington et al., 2000) or opening special care units (Zinn & Mor, 1994) may be more likely to invest in the employment of NPs or PAs.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
The sample for this analysis includes facilities in OSCAR that were surveyed during the period from 1993 to 2002. We excluded Alaska, the District of Columbia, Hawaii, and Puerto Rico because of the small number of facility surveys (405 surveys from 91 facilities). We also excluded 537 facilities that changed either from or to hospital-based status during the study period, and 253 facilities with a single survey. Each observation in the final dataset links the current outcome with information occurring between 8 and 15 months before the current survey (the interval at which OSCAR surveys were normally conducted). The final sample contains 137,190 surveys from 17,635 distinct nursing facilities: 11,665 (66%) urban and 5,970 (34%) rural; 2,138 (12%) hospital based, and 15,497 (88%) freestanding.
Variable Specification
Although NPs and PAs receive different training and have different practice regulations (which also vary across states), we treat them together here because the OSCAR data do not separate between them. We did not distinguish NPs or PAs employed on staff directly by the nursing home or by contract through an agency because in both cases the facility had made the decision to pay for those services, and it is this decision that we are studying.
First we plotted time trends in the employment of NPs or PAs in four prespecified markets—freestanding facilities in urban and rural markets, and hospital-based facilities in urban and rural markets. Then, we modeled factors associated with investment in NP–PA services, measured by a dichotomous variable identifying facilities that chose to employ NPs or PAs, versus all others. We considered facilities that were recorded as having NP–PA services available to residents on site but that did not report any employed NPs or PAs with those that did not have any NP–PA services, because they were not considered to have made a significant investment in medical infrastructure through the employment of NPs or PAs.
Many independent variables have a continuous form, but we included them in the multivariate model by using a threshold to reduce noise, and to ease interpretation. In particular, we used an indicator for facilities in states with reimbursement rates in the upper quartile, that is, over $128.67 in 2002 dollars adjusted by the Consumer Price Index (CPI), to test the hypothesis regarding the effect of Medicaid reimbursement rate on NP–PA employment (H1). To test the hypothesis regarding the effect of managed care penetration (H2), we included an indicator variable for markets (counties) in which more than 15% of the Medicare population was receiving care from a managed care organization (upper quartile). To test the hypothesis regarding the effect of competition (H3), we included two variables: an indicator for low market concentration using the Hirschman–Herfindahl index (Hirschman, 1945; indicated in the lower quartile, less than.09 on a 0–1 scale), and an indicator of excess capacity (Nyman, 1988) of nursing home beds in the market (indicated in the upper quartile, more than 20 empty beds per facility, on average, in the market). We computed both indices by aggregating facility data from OSCAR to the county level.
With respect to facility payer mix and facility case-mix acuity, the proportion of residents who were not paid by Medicare or Medicaid (mostly self-paying) was centered at its mean of 26% with steps of 5%, and facilities with more than 10% Medicare residents (upper quartile) in each survey were indicated. We measured case-mix complexity by means of three indicator variables: (a) facilities providing rehabilitation services to a higher volume of residents, that is, having 35 or more rehabilitation residents and 30% or more residents receiving rehabilitation services, or having 20 or more rehabilitation residents and 50% or more residents receiving rehabilitation services (Berg, Intrator, & Lemon, 2001); (b) facilities with any residents receiving intravenous therapy; and (c) facilities with any residents receiving tracheotomy care.
Other investments in medical infrastructure are represented by three indicator variables: (a) whether a facility has more than one half full-time equivalent additional physician other than the medical director; (b) whether the total number of hours of employed nurses (RN, LPN, and CNA) per resident day is greater than 4.55, the standard recommended (Harrington et al., 2000) to provide adequate levels of care to residents; and (c) the existence of any special care unit in the facility (Zinn & Mor, 1994).
We included several control variables in the model. For-profit facilities may be less inclined to invest in costly medical infrastructure, particularly if these costs are not recoverable. Facilities that were part of chains may benefit from shared experience and hiring practices, and may have been better able to contract NPs or PAs. We also included an indicator for stand-alone (nonchain) for-profit facilities, as those are different from for-profit chain facilities. Larger facilities command greater internal resources and may be more capable of accommodating environmental demands through internal restructuring than smaller facilities would be; therefore, the model included the number of nursing home beds centered at its mean of 118 beds with steps of 10 beds. We included nursing home occupancy rate on the date of survey as an indicator variable, where low occupancy was determined as below 85% of the facilities' total capacity (lower quartile).
Nursing home markets have traditionally been defined as counties because of patterns of funding and patient origin (Banaszak-Holl et al., 1996; Nyman, 1985, 1989; Zinn, 1994). Although there have been reports that this may be less than an ideal approximation (Zwanziger, Mukamel, & Indridason, 2002), aggregate data based on geographic market definitions other than the county are not readily available. The ARF is a county-level database containing information on socioeconomic data as well as the availability of medical professionals and services, which we used to control for market heterogeneity (Stambler, 1988). Market characteristics that may influence the strategy to employ NPs or PAs included a proxy measure for their availability. Because NPs and PAs work directly with physicians, it is reasonable that there would be more NPs and PAs in markets where there are more physicians. Therefore, we used an indicator for markets in which the number of physicians per 1,000 population was in the upper quartile (over 2.4 physicians). Other market controls included a measure of resource availability—the number of hospital beds per 1,000 population (centered at its mean, 4); a measure of need for nursing home services—the proportion of population over age 75, based on the 1990 figures available in the ARF (centered at 5%); and a measure of attraction for more competitive care—per capita income as recorded annually in the ARF (centered at $23,000 with steps of $1,000). We used the Metropolitan Statistical Area average wage index (hospital sector) to adjust for differences in the buying power of the dollar when comparing the effect of Medicaid reimbursement rates across geographic areas. We standardized the wage index at its mean, 0.96, with steps of 1 SD, 0.13. We included calendar time, measured by the number of years from January 1, 1993 up to the current survey date, to control for unmeasured changes that occurred over time.
We measured most covariates at the previous survey to predict the outcome, except calendar time, which was measured concurrently at the time of the outcome. A description of all the independent and control variables included in the multivariate model (for urban freestanding facilities aggregated over all surveys between 1993 and 2002), including their means and standard deviations (for continuous variables), is provided in Table 1. Medicaid reimbursement rate averaged $115.22 and ranged between $70.19 and $214.52 in 2002 CPI-adjusted dollars over the study period. Average managed care penetration rate was 9.2% in these markets over this time period. We standardized the Herfindahl index of concentration to the 0–1 range, and it averaged.25, with the lower quartile (highest competition) at.09. The county average number of empty beds per nursing home had a mean of 12, and an upper quartile (highest competition) of 20 empty beds.
|
Specifically, we estimated a cross-sectional time-series generalized estimating equation (GEE) model with a logit link function for Pit , the probability that facility i employs an NP or PA at time t. The model controlled for unobserved state effects by including state dummies (with Wyoming being the reference state), and a linear calendar time effect to control for time trend. The form of the model is log[Pit/(1 – Pit)] = ß0 + ßXit, where ß0 is the intercept, Xit is a vector of facility and market characteristics, and ß is a vector of parameter estimates for the effects of the covariates. GEE models properly account for within-facility correlations and are suitable for analyses of cross-sectional time-series data like OSCAR. We used the XTGEE procedure available in Stata (2003), along with the Huber-White sandwich estimator (Huber, 1967; White, 1980) to correct for the clustered nature of the data. The estimation requires specification of a working correlation matrix that tries to follow the within-facility correlations. The final estimates from these methods are unbiased in both the parameter estimates and the standard errors, even if the within-facility working correlations are misspecified. In this analysis, we assumed an exchangeable working correlation structure.
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
|
|
Facilities in markets with more physicians and with higher per capita income were more likely to employ NPs or PAs (AOR = 1.09 and 1.01, respectively). Facilities in markets with a higher proportion of older adults were less likely to employ NPs or PAs (AOR = 0.96).
State fixed effects (not presented) were very substantial with an average log odds of –0.53 (SD = 0.53). This ranged between AOR = 0.19 in Illinois and AOR = 1.69 in Montana, compared to Wyoming. Although we controlled for many time-varying facility and market characteristics, a linear calendar time trend was still apparent with an increase of 14% in the odds of employing an NP or PA for every year.
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Facilities with special care units, those able to meet the desired level of skilled nurses on staff, and those that have other physicians in addition to the medical director on staff or contract are clearly of a type that are making a significant investment in quality, in the comprehensiveness of their care patterns, and in the overall skill level of the environment. In contrast, nonaffiliated for-profit facilities, which have traditionally been mom-and-pop operations, mostly handling lower acuity residents, were significantly less likely to employ an NP or PA (AOR = 0.76).
The employment of NPs or PAs also reflects potential market demand. In markets with higher managed care penetration and in markets with more physicians, there appeared to be greater NP–PA availability. This may reflect the service capabilities that managed care organizations enrolling Medicare beneficiaries expect from contracting nursing facilities. These facilities are used to further reduce hospital days and to provide ongoing postacute care in the nursing facility even in the absence of physician observation and care.Furthermore, the findings that facilities with a higher proportion of residents with complex clinical problems and those providing intravenous care also indicate that the investments in medical infrastructure are either in response to, or in preparation for, serving patients with complex medical needs.
It is noteworthy that NP–PA availability in nursing homes is related to the number of physicians available in the counties in which facilities are located. Presumably this suggests that the availability of NPs or PAs increases as the number of physicians increases, and the more NPs or PAs the more likely that nursing homes will be able to employ or contract with them. The relationship between physician availability in a market and NP–PA availability in nursing homes raises the issue of whether they represent substitutes or complements for physicians in nursing homes. The argument for substitution derives from profit maximization. Profit maximization assumes that rational facilities choose the least-cost combination of inputs to produce a given quantity of output. Profit maximization is a valid assumption for for-profit facilities. In addition, in more competitive markets, nonprofit nursing facility behavior is not distinguishable from for-profit behavior (Grabowski & Hirth, 2003). However, the question may lack relevancy in the nursing home setting. First, physicians are a very limited input into the nursing facility production process. Medicare (arguably the payer for residents requiring the most complex care) only requires one physician visit per month. Second, nursing facility output has not remained homogeneous over time. There is evidence that nursing home acuity has increased since the implementation of the prospective payment system in hospitals (Keeler et al., 1990). Increased output acuity may require either additional or a new mix of inputs. However, nursing homes are not employing physicians to produce a higher level of acuity. Rather, NPs or PAs are a new input into the production process. That higher levels of professional nurse staffing are associated with NP–PA availability suggests that NPs or PAs are complementing skilled nursing as opposed to substituting for physician services.
Although this study is one of the first to examine factors associated with the availability of both NPs and PAs in nursing homes, there are some limitations. The reliability of the OSCAR data has been reported to be questionable (Centers for Medicare and Medicaid Services, 2001; General Accounting Office, 2002; Health Care Financing Administration, 2000; Straker, 1999), especially for staffing data (Harrington, Carrillo, Mullan, & Swan, 1998). In terms of availability of NPs or PAs, there may be two problems: First, the report of NPs or PAs available to residents on site who were not employed by the facility may not be uniformly reliable across facilities; second, the reported number of hours worked by NPs or PAs appeared to have many improbable outliers. Thus, we used the employment of NPs or PAs, and not its reported intensity, to minimize the potential error in the outcome.
Some of the variables used in the analyses are proxies for particular concepts. For example, the constructs measuring case-mix complexity are the best available proxies for resident acuity. Another limitation is the data available to control for other market factors such as state Medicaid program factors and market share.
This study is based on an administrative database of nursing homes surveys, the OSCAR data. This study is observational in nature in that Medicaid payment rates are observed and not randomly assigned. However, the analyses adjusted for other unobserved differences between states by including state fixed effects. Thus, idiosyncratic forces such as wealth and education and other population characteristics that might predispose a state to pay a higher Medicaid rate are likely controlled. Staffing variables, as well as facility variables describing resident mix, are endogenous to facilities' decision to employ NPs or PAs. Because the inclusion of endogenous covariates may bias other estimated effects, we compared the results of the full model with those of a partial model in which all these variables were eliminated. We found no differences between the estimates, suggesting that endogeneity is not a serious concern in this analysis.
The positions of NP and PA are not completely interchangeable. The education and training of these individuals differ, as does the regulation of their practice. Because these differences may reflect on the different capacities of these professionals to function in the nursing home, and on their potential impact, further research is needed to understand the implications of state policies on the employment of NPs and PAs in nursing homes, and to examine whether there are different clinical practice patterns and responsibilities.
Many factors appear to affect the employment of NPs and PAs in nursing homes. Because states rely on nursing homes to care for their most frail population through their Medicaid programs, it is important that they provide the appropriate resources to the nursing homes to care for them. This article demonstrated that higher Medicaid reimbursement rates advance the investment in medical infrastructure. With the current economic crisis, when states are considering cutting back on nursing home spending, it is important that states reflect on these findings for the welfare of dependent frail populations.
|
Footnotes |
|---|
1 Center for Gerontology and Health Care Research, Brown University, Providence, RI.
2 Department of Community Health, Brown University, Providence, RI.
3 Rhode Island Quality Partners, Providence, RI.
4 Yale University School of Nursing, New Haven, CT.
5 Fox School of Business and Management, Temple University, Philadelphia, PA.
Decision Editor: Linda S. Noelker, PhD
Received for publication March 25, 2004. Accepted for publication January 10, 2005.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  A. Gruneir, V. Mor, S. Weitzen, R. Truchil, J. Teno, and J. Roy Where People Die: A Multilevel Approach to Understanding Influences on Site of Death in America Med Care Res Rev, August 1, 2007; 64(4): 351 - 378. [Abstract] [PDF]
|
|
|
|
 |
|
 A. Gruneir, S. C. Miller, O. Intrator, and V. Mor Hospitalization of Nursing Home Residents With Cognitive Impairments: The Influence of Organizational Features and State Policies Gerontologist, August 1, 2007; 47(4): 447 - 456. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| All GSA journals | Journals of Gerontology Series A: Biological Sciences and Medical Sciences | Journals of Gerontology Series B: Psychological Sciences and Social Sciences |
