| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
a Institute for Health, Health Care Policy and Aging Research, New Brunswick, NJ
Correspondence: Nicholas G. Castle, PhD, Institute for Health, Health Care Policy and Aging Research, 30 College Avenue, New Brunswick, NJ 08901. E-mail: Castle_Nick{at}Hotmail.com.
Laurence G. Branch, PhD
 |
Abstract |
|---|
 TOP Abstract BackgroundHypothesesMethodsResultsDiscussionReferences |
|---|
Key Words: Market factors • On-line Survey and Certification of Automated Records (OSCAR) • Organizational factors
In comparison to the rather substantial literature in the hospital industry, there is a paucity of research examining innovation in the nursing home industry; therefore, there is an urgent need to facilitate and understand innovation in this setting. Nursing homes are often criticized for their poor quality of care. Press reports (Consumer Reports 1995
), empirical research (e.g., Davis 1991; Ray, Federspiel, and Schaffner 1980), and the federal government's own assessment of nursing homes generally note endemic quality problems (Institute of Medicine 1986; General Accounting Office 1998, General Accounting Office 1999a, General Accounting Office 1999b, General Accounting Office 1999c). Clearly, quality of care is influenced by many factors, but as evidenced by the experience of the hospital industry, innovation has the potential to improve quality in some areas of care (Shortell et al. 1995). With little research examining innovation in the nursing home industry, the link between innovation and improved quality in these facilities is less developed but is often assumed to exist. For example, innovations such as total quality management (TQM), computerization of medical records, and use of specialized care settings may improve quality (Banaszak-Holl, Zinn, and Mor 1996; Castle and Banaszak-Holl 1997). This study examined organizational and market factors associated with the early adoption of innovations in nursing homes.
Early adopter institutions are defined as the first 20% of facilities to adopt an innovation (Rogers 1983). Although the initial adoption of an innovation within most industries is characteristically sluggish, these so-called early adopters may subsequently facilitate its diffusion throughout the industry (Rogers 1983). Thus, identifying characteristics associated with this early adoption process could be useful in further facilitating the diffusion of innovations in the nursing home setting. However, innovation diffusion is influenced by multiple factors. These factors include changes in the competitive environment, profit potential, and regulation. Diffusion may also simply be a copycat, symbolic, or emotional phenomenon (Abrahamson 1991; Scott 1990). In this investigation organizational and market characteristics are examined: first, because they are often under the purview of legislators; second, because an organization's adoption of an innovation is highly dependent upon its own characteristics and the nature of the market (Mansfield 1968). As some authors have pointed out, organizational and market factors are of primary importance as "determinants of innovation" (Attewell 1992; Damanpour 1991).
A variety of operational definitions of innovation exist. For example, one definition is whether the innovation was created within the organization (Aiken and Hage 1971). Another common way to operationalize innovation is to focus on practices that are new to the organization adopting them (Hage and Dewar 1973). As Daft 1982(p. 131) describes, "the idea can be either new or old in comparison to other organizations so long as the idea has not been previously used by the adopting organization"; this is the definition of innovation used in this investigation.
Results of analyses using single innovations may be idiosyncratic and raise questions of generalizability (Kimberly and Evanisko 1981). Studying more than one innovation increases the chance that the role of the organizational and market characteristics will be robust (Damanpour 1991). I examined the startup of two groups of innovations, special care units and subacute care services, which together consist of 13 possible innovations.
Special care units are beds identified by a facility for residents with specific needs or diagnoses. This group of innovations includes beds for patients with Alzheimer's disease, AIDS, dialysis, head trauma, Huntington's disease, ventilators, hospice, and special rehabilitation (Health Care Financing Administration [HCFA], 1992; Banaszak-Holl et al. 1996), all of which are included in my analyses. Special care units in nursing homes are an emergent trend (Freiman and Brown 1999); for example, the number of special care hospice units in nursing homes increased by over 100% from 1992 to 1997 (On-line Survey and Certification of Automated Records [OSCAR], 1992–1997), making these units suitable for early innovation analyses.
Innovations in subacute services commonly include physical rehabilitation, intravenous (IV) therapy, wound management, cardiac treatment, and dialysis, all of which are included in my analyses. Initially, these services were offered primarily by nursing facilities that provided care to Medicare recipients because these residents were more likely to be recovering from an acute illness necessitating these modalities of care (Intrator, Castle, and Mor 1999). However, changes in technology have created a larger market for these services. Also, managed care plans are increasingly forging links with nursing homes that provide subacute care. Thus, the provision of these services has recently become more common, making subacute services suitable for early innovation analyses with the 1992–1997 data used in this investigation.
|
Background |
|---|
|
TOPAbstractBackgroundHypothesesMethodsResultsDiscussionReferences |
|---|
). It is also interesting to note that, in its first wave of deliberations, the Delphi panel identified more than 40 innovations in nursing homes. This is in stark contrast to the custodial care usually associated with these institutions.
Banaszak-Holl and colleagues 1996 examined the impact of six market characteristics (HMO penetration, number of hospital beds, Medicare hospital discharges, market competition, regulatory policies, and Medicaid reimbursement rate) and three organizational characteristics (Medicare census, bed size, and profit status) on the provision of Alzheimer's and subacute special care units. The authors found significant results for five of these nine factors, leading them to state that special care units are partly a response to a growing demand by resource providers' and facilities' attempts to maintain a competitive edge in their markets.
More recently, Castle and Banaszak-Holl 1997 examined how the demographic characteristics of the top management team in 236 nursing homes can affect the adoption of innovations. The innovation examined was the computerization of the Minimum Data Set (MDS), and characteristics investigated were tenure, education, and involvement in a professional society. The results were generally significant for each of these factors. However, the results for top managers of nonchain nursing homes showed a greater association between demographic factors and innovation than the results for top managers of nursing homes belonging to a chain.
The lack of studies addressing innovation in nursing homes may not be problematic if research from other areas within the health care arena can be applied to this setting. Although there is a substantial body of innovation research set in hospitals (e.g., Kimberly and Evanisko 1981; Meyer and Goes 1988), the nursing home industry is dissimilar from the hospital industry in several ways that may have a significant influence on the adoption of innovations (Kimberly and Evanisko 1981). These factors include: staffing characteristics such as top management turnover and education; numbers and types of clinical staff; organizational characteristics such as size, profit status, and patient mix; and market characteristics such as levels of competition and regulatory stringency.
|
Hypotheses |
|---|
|
TOPAbstractBackgroundHypothesesMethodsResultsDiscussionReferences |
|---|
). This has been referred to as the "liability of smallness" (Gifford and Mullner 1988) and may be why facility size is an important correlate of innovation. For example, hospitals with fewer beds are less likely to innovate (Kimberly and Evanisko 1981). I hypothesize that small nursing homes will also innovate conservatively, and propose: Hypothesis 1: Small nursing homes are less likely to be early adopters of innovations.
The organizational goals and resultant behavior of for-profit and not-for-profit providers may be dissimilar (Banaszak-Holl et al. 1996; Holmes 1996; Koetting 1980). For-profit nursing homes are often seen as profit oriented, and as a result may be less aggressive in implementing costly resident care services (Davis 1991; Greene and Monahan 1981; Koetting 1980; Spector and Takada 1991). Not-for-profit nursing homes are often seen as more altruistic and as a result may be more aggressive in implementing resident care services, irrespective of costs (Davis 1991; Greene and Monahan 1981; Koetting 1980; Spector and Takada 1991). These differences in mission are likely to be reflected in the types of innovations adopted by the two organizational forms. Not-for-profit facilities are likely to adopt innovations that facilitate better resident care, such as special care units and subacute care. Thus, I propose:
Hypothesis 2: Not-for-profit nursing homes are more likely to be early adopters of innovations.
Chain membership may promote innovation. For example, chain membership was associated with a greater likelihood that a nursing home would adopt a computerized information system (Castle and Banaszak-Holl 1997). Facilities that are part of a chain may have access to capital, along with other resources that may help facilitate innovation. Corporate management may also disseminate successful innovations among member facilities once they are seen to be viable in pilot facilities. Therefore, I hypothesize:
Hypothesis 3: Nursing homes that are members of chains are more likely to be early adopters of innovations.
The majority of persons in nursing homes are Medicaid recipients. The Medicaid program, due to budget constraints, provides lower reimbursement to nursing homes compared to private-pay residents. As a consequence, it may be difficult for facilities to provide adequate services (Kim 1990; Wagner 1987, Wagner 1988). Indeed, there is some indication that the Medicaid program may be paying less than costs in some states (Kim 1990; Wagner 1987, Wagner 1988). To avoid dependence on this public program, many nursing homes have focused on the private-pay market because they have greater latitude in establishing the price of services they provide to these residents. Nursing homes may be more able to innovate when they cater to private-pay residents. I hypothesize:
Hypothesis 4: Nursing homes with more private-pay residents are more likely to be early adopters of innovations.
Market Factors
States have some degree of flexibility in establishing their methodologies for Medicaid payment (Buchanan, Madel, and Persons 1991). Some of these are more restrictive than others. Facilities operating under retrospective methodologies are reimbursed for actual costs incurred, whereas prospective payment is more likely to pay a pre-set flat rate. From a nursing home's perspective, prospective reimbursement is more stringent than retrospective reimbursement (Banaszak-Holl et al. 1996). Facilities operating under retrospective reimbursement may be more able to innovate than those operating under prospective reimbursement because of these higher reimbursement levels. I hypothesize:
Hypothesis 5: Nursing homes operating in markets with retrospective Medicaid reimbursement are more likely to be early adopters of innovations.
Some states have focused on containing the supply of nursing home beds as a strategy aimed at controlling their Medicaid costs. Common methods of constricting the bed supply include Certificate of Need (CON) legislation and new construction moratoria. For potential providers these limitations are a barrier to entry into these markets (Banaszak-Holl et al. 1996); but, the reduced bed supply increases the demand for existing nursing homes' services (Cohen and Dubay 1990). As a result, facilities that already provide care in these markets may be less inclined to innovate because of a greater proportion of residents with need for traditional nursing home care. Therefore, I hypothesize:
Hypothesis 6: Nursing homes operating in markets with CON legislation or new construction moratoria are less likely to be early adopters of innovations.
Environments vary in their degree of competitiveness. Pfeffer and Salancik 1978 argue that organizations share a limited pool of resources in more competitive environments and that survival depends on organizational effectiveness. Furthermore, in more competitive environments, nursing homes have the incentive to differentiate their services in order to improve their image in the marketplace. The use of innovations can help to increase both organizational effectiveness and differentiate services. One source of competition is from other nursing homes. Therefore, I propose:
Hypothesis 7: Nursing homes operating in competitive nursing home markets are more likely to be early adopters of innovations.
Nursing homes also compete with providers in other sectors of the health care industry. For example, hospitals provide a variety of long-term care services to the elderly population, on both an inpatient and outpatient basis (Muramatsu, Lee, and Alexander 2000). Services such as hospital-based geriatric care units increased from 666 in 1990 to 709 in 1993. Likewise, hospital hospice services increased from 868 to 1,082, and geriatric clinics from 461 to 537 in the same period (American Hospital Association 1991, American Hospital Association 1994). Thus, following Hypothesis 7, I propose:
Hypothesis 8: Nursing homes operating in markets with hospital-sponsored outpatient long-term care services are more likely to be early adopters of innovations.
Hypothesis 9: Nursing homes operating in markets with hospital-sponsored inpatient long-term care services are more likely to be early adopters of innovations.
In each market, the abundance of resources available to organizations can vary. In some markets, resources are relatively abundant, or munificent (Staw and Szwajkowski 1975); as a result, nursing homes may have greater access to resources that would allow them to innovate. Among the more important resources for long-term care providers are the number of elderly persons, average income, and the number of hospital beds in the market.
Elderly people (over the age of 65) are the primary recipients of nursing home services, so an area with a high proportion of such individuals is munificent from the perspective of market growth potential. In markets with higher per capita income, the private-pay segment is likely to be larger. Finally, nursing homes are also dependent on discharges from hospitals, suggesting that in areas with more hospital beds the environment is likely to be more munificent. Therefore, it follows:
Hypothesis 10: Nursing homes operating in markets with many elderly people are more likely to be early adopters of innovations.
Hypothesis 11: Nursing homes operating in markets with higher incomes are more likely to be early adopters of innovations.
Hypothesis 12: Nursing homes operating in markets with many hospital beds are more likely to be early adopters of innovations.
|
Methods |
|---|
|
TOPAbstractBackgroundHypothesesMethodsResultsDiscussionReferences |
|---|
Those facilities that are neither Medicare nor Medicaid certified (approximately 800 in 1992 and 1,000 in 1997) are not included in the OSCAR data. This may have some impact on the representativeness of my results; but clearly, in the absence of data and previous research on these nursing homes, I am unable to determine whether these facilities are early adopters of innovations.
There are approximately 300 data elements in the OSCAR, the majority of which are either organizational or aggregate resident data. Organizational data relevant to this study include bed size, chain membership, ownership, and the number of nursing personnel (by job category and full-time equivalent [FTE] status). Resident data relevant to this study include dependence/independence in activities of daily living (ADLs) and the number of residents by payer category.
Much of the OSCAR data are self-reported by the nursing home administrator and director of nursing. Interrater reliability testing has not been performed for the data as a whole, and such biases are generally unknown, although in prior analyses I compared primary data with many items overlapping with those in the OSCAR collected from more than 400 nursing homes and found very high correlation. This was not unexpected given that it is doubtful whether often stable facility factors such as ownership, chain membership, or bed size are subject to reporting bias because these factors are often found on business records (e.g., purchase orders, letterhead, advertisements) readily available to surveyors. In addition, most data elements pertaining to resident characteristics are verified by the surveyors. Thus, it is not surprising that these data are widely used as a secondary source of nursing home characteristics.
The OSCAR data can be limited because the information in the surveyors' report is pertinent only for the time they make rounds in the facility, usually occurring during the day shift. Twenty-four-hour observation by the surveyors in each facility is not possible. Care practices, such as physical restraint use, may be biased in analyses because other shifts may not follow day shift practices (Castle 2000). I have no reason to believe that the variables used in this analysis (e.g., census, bed size, or facility ownership) are biased in this way.
The 1999 ARF data are compiled from a number of sources, including the American Hospital Association (AHA) annual hospital survey, the U.S. Census of Population and Housing, the Centers for Disease Control and Prevention (CDC), and the National Center for Health Statistics (NCHS; Stambler 1988). These data are at the county level and are commonly used in health services research (e.g., Banaszak-Holl et al. 1996; Castle and Banaszak-Holl 1997; Nyman 1987). In this investigation the ARF was used to measure the number of outpatient long-term care facilities, number of hospital-based long-term care services, number of elderly subjects, and average income in the county in which the nursing home is located. The 1999 data include these figures from the 1980s through 1996. The OSCAR data generally contain information from the prior year. Therefore, I was able to match the OSCAR and ARF with presumably little measurement error.
Variables
I used a single binary variable created from two groups of innovations—special care units and subacute care services. As shown in Table 1 , together these consist of 13 possible innovations.
|
); conversely, in facilities with high average ADL scores, less caregiver time may be available to implement innovations. Facilities with high average occupancy rates may have more available resources to innovate (Nyman 1987).
Table 1 shows how the variables are operationalized and, where applicable, the coding for the analysis is included. With the exception of the Herfindahl index and Medicaid reimbursement, these variables are self-evident. The Herfindahl index is a measure of how competitive a market is in which a facility is located. Following a number of studies, the county was considered to be the market (Castle and Banaszak-Holl 1997; Banaszak-Holl et al. 1996). The index ranges from 0 to 1, with 1 representing a monopoly market, and lower values in cases where there are many homes each with a small share of the market (White and Chirikos 1988). A dummy variable identifying whether a facility receives either retrospective or prospective reimbursement under the Medicaid program is included. Facilities are reimbursed in varying degrees for actual costs under retrospective policies, but prospective Medicaid reimbursement methodologies provide lower payments to nursing homes than retrospective methodologies; therefore, I represent these payment methodologies as a dichotomous variable.
Procedures and Limitations
This analysis excludes hospital-based facilities and facilities that are part of a retirement center (n = 1,138) because they tend to be unrepresentative of other nursing homes in terms of staff, residents (Burns and Taube 1984), and organizational factors (Singh and Schwab 1998). For example, they are predominantly not-for-profit. These nursing homes may also have greater economies of scope and/or greater access to capital because of their partnership. Greater access to capital, along with other resources, may help facilitate innovation adoption in these facilities.
Starting with the April 1992 OSCAR data, facilities identified as having one of the innovations were assigned a value of 1, and those identified as not having any of the innovations were assigned a value of 0. This process was repeated at 6-month intervals through to the April 1997 OSCAR data, giving an approximate time period during which each facility may have adopted the innovation. Six-month intervals were used because of the data available; thus, this is a time period of convenience and has no conceptual or theoretical importance. Other analyses using more frequent intervals would surely be a further refinement to my analyses.
This approach has several disadvantages. First, data prior to 1992 were not available to me; thus, some of the early adopters are not truly matched to the facility and market factors at the time of innovation initiation. However, the development of the innovations investigated in this study has been rapid, so censoring (i.e., incomplete/unavailable data prior to 1992) represented only 2% of the early adopter sample. Indeed, in sensitivity analyses (not reported) excluding these facilities, the findings were robust. But clearly, I am still restricted to a sample of convenience, and these 2% of earliest adopters may be influenced by factors other than those I include in the analyses. For example, some innovation may be triggered by the Prospective Payment System (PPS) in hospitals, because this led to patients being discharged to nursing homes with greater acuity levels (Harrington and Carrillo 1999). However, the concept of "early" in adoption studies is generally regarded to encompass the initial 20% of adopters (Rogers 1983). Based on this, I believe that these earliest adopters do not compromise this study.
A second limitation of this study is that some facilities are excluded from the sample because of difficulties matching nursing homes from year to year. This problem arises because nursing homes do not necessarily keep the same identification number, the primary source of facility identification. Nursing homes that change management, ownership, or location are often given new numbers. Other facilities may have closed or opened during the 5-year study period, leading to the loss of identification numbers and the creation of new identification numbers, respectively. For those facilities with unmatched identification numbers, matching algorithms by facility name, address, and ZIP code were used. Missing cases still occurred, and from the 1992 data approximately 8% (n = 1,107) of nursing homes were not represented in the analyses because they could not be identified in one of the subsequent data sources. Clearly, I am unable to determine the potential impact these missing data points have on the analyses. Some facilities may change management, ownership, or location because they are successful innovators, yet other facilities may undergo these same changes, or close, because they do not innovate.
The analyses are less representative of facilities in the 1997 OSCAR. This is because the number of facilities included in the OSCAR has grown since 1992. Approximately 17% (n = 2,231) of nursing homes in the 1997 OSCAR were not represented in the analyses because they were not included in one of the prior data sources.
There were very little missing data on any of the dependent or independent variables. Insufficient data were present in 48 cases for the dependent variables, resulting in an analytic sample of 13,162 facilities. In most cases, information for the independent variables in this analytic sample was available. Missing cases represented between 0%–2% for all of these variables, and less than 1% of facilities had any missing data. All missing values for continuous or ordinal variables were imputed using mean substitution. Dichotomous variables were randomly assigned 0 or 1 values according to the binomial distribution with a probability as observed for the complete cases (Maddala 1977). Other methods for dealing with missing data are available (see Little and Rubin 1987); because of the small number of facilities with any missing data, however, these methods are unlikely to be advantageous in my analyses. The results reported are robust in that imputation did not produce any significant change in my results compared to analyses performed prior to imputing missing data values.
Common errors in the OSCAR data include approximately 2% of duplicate facilities and between 0%–4% of data with entry errors. Duplicate facilities were eliminated using the federal identification number and the survey date. When an identification number appeared more than once in the data, the information associated with the most recent survey date was used. If the survey dates were identical, one of the duplicate facility records was chosen randomly to be used in the analysis. Following the approach outlined by other researchers using these data, frequency distribution plots were used to identify obvious outliers (Castle and Fogel 1998). Imputation, as described above, was used to replace these data entry errors.
To recap, 15,455 unduplicated nursing facilities are included at baseline in the 1992 OSCAR. I excluded hospital-based facilities and facilities that are part of a retirement center (n = 1,138). A further 1,107 facilities are lost to follow-up in the 1997 data and 48 have missing data, resulting in an analytic sample of 13,162 facilities.
Analysis
I analyzed the effect of these variables on nursing homes' adoption of the innovations of interest using discrete-time logit modeling. This method is appropriate for examining dichotomous dependent variables with longitudinal data (Yamaguchi 1992). Cross-sectional logit and probit models more simply characterize organizations as adopters and nonadopters, thereby suppressing information on the timing of innovation adoption (Lee and Waldman 1985). In my case, the advantage this regression technique has over these more commonly used cross-sectional models is the robust ability to account for right censoring and the large time intervals used. Right censoring occurs when the time of an event is unknown. For example, in this investigation the time of innovation adoption in unknown. A facility may adopt an innovation of interest at any time from 1992 to 1997, or it may not adopt any innovations of interest during this interval. The unit of analysis in the discrete-time logit modeling used in this analysis is the facility-interval, rather than the individual facility (or person) as would be used in ordinary regression models. Specifically, this is a nursing home 6-month interval. Because the data are only in 6-month increments, I cannot provide an unequivocal determination of the precise timing of an innovation; in cases such as these, logit modeling is preferred over continuous time models (Cox 1972) such as Cox proportional hazards (Ingram and Kleinman 1989; Yamaguchi 1992).
The dependent variable of moving from the noninnovation state to the innovation state is a dichotomous variable (0,1). At every nursing home 6-month interval, this variable will be 1 if an innovation was observed and 0 if not. Logit models are recommended for use with such dichotomous dependent variables. Controlling for all of the variables in a model, they provide a robust test of significance. In evaluating the effects of these independent variables, odds ratios were calculated by taking the exponent of the parameter estimates on all variables.
I also used generalized estimating equations (GEE; Zeger and Liang 1992). This was because biases can occur in data consisting of repeat observations. The biases are due to the potential correlation among the repeat observations and they can lead to elevated significance levels. GEE controls for the correlation due to repeat observations and provides more robust significance levels (Zeger and Liang 1992). It does so by separating the within-subject correlation from the regression coefficient estimations (Karim and Zeger 1988; Succi, Lee, and Alexander 1997).
|
Results |
|---|
|
TOPAbstractBackgroundHypothesesMethodsResultsDiscussionReferences |
|---|
|
). However, before discussing the effects of specific variables, it should be noted that with a chi square of 71 (p < .001) the overall goodness-of-fit of the model is statistically significant. It should also be noted that because of the large number of cases included in this study, statistically significant results are achieved with relatively small effect size.
|
With regard to market characteristics, prospective Medicaid reimbursement (AOR 0.67; p < .01) decreased the likelihood of early adoption of innovations. Facilities in areas with prospective Medicaid reimbursement have 0.33 times the likelihood of early innovation adoption as do other facilities. Facilities in areas with a higher Herfindahl index score (AOR 0.89; p < .01) decrease the likelihood of early adoption of innovations. Higher average income (AOR 1.09; p < .05) and higher numbers of hospital beds (AOR 1.11; p < .01) increase the likelihood of early adoption of innovations.
The control factors FTE RNs/bed (AOR 1.74; p < .001), FTE LPNs/bed (AOR 1.11; p < .01), FTE nurse aides/bed (AOR 0.76; p > .01), and average occupancy (AOR 0.89; p < .01) are also significantly associated with the likelihood of early innovation adoption.
|
Discussion |
|---|
|
TOPAbstractBackgroundHypothesesMethodsResultsDiscussionReferences |
|---|
|
). This analysis shows that this may not always be the case. To avoid dependence on low Medicaid payments, many nursing homes have focused on private-pay residents because they have greater latitude in establishing the price of the services they provide to these residents. Consequently, competition for private-pay residents has increased. As nursing homes may be more able to innovate with a high census of these residents, it is not surprising that early adopter nursing homes also have greater numbers of private-pay residents (supporting Hypothesis 4). Being that Medicaid residents are the reference category used in the analysis, the opposite, of course, is the case for these residents. Early adopter nursing homes have lower numbers of Medicaid residents. Although not examined in this analysis, it may also be interesting to determine whether early adopter nursing homes attract private-pay residents.
The results also suggest that future policies that attempt to reduce Medicaid costs in nursing homes should be more sensitive to the externalities of these decisions—in this case, less innovation. Because the effect of lower Medicaid reimbursement is mitigated by state Medicaid reimbursement policies (Hypothesis 5), it would appear that states wanting to reduce costs while not limiting facilities' abilities to be early adopters of innovations should use retrospective reimbursement.
Competition in the nursing home industry is often proposed as a driving force for change in the industry (Banaszak-Holl et al. 1996). This is congruent with the usual econometric propositions regarding the effects of competition, although previous studies in nursing homes have not always shown this to be so (Castle and Fogel 1998). My result for the Herfindahl index shows that, the more facilities compete with other nursing homes, the more likely they are to be early adopters of innovations (supporting Hypothesis 7).
I proposed that, in more munificent markets, nursing homes would have greater access to resources allowing them to innovate. The resources I investigated were the size of the elderly population, average income, and the number of hospital beds in the market. I find support for the latter two resources (Hypotheses 11 and 12).
The results of the control variables for staffing are also worth noting. The provision of the innovations described in this analysis are often seen as intricate, involving training and requiring more care and attentiveness from staff. RNs and LPNs may be most able to respond to these challenges, thus explaining the results suggesting facilities that maintain higher levels of RNs and LPNs are likely to be early adopters of innovations. However, these caregivers also come at a higher cost than other staffing alternatives, such as nurse aides. Therefore, in recent years many nursing homes have substituted nurse aides for LPNs and RNs. The hidden cost of this substitution may be a lower ability to innovate.
Contrary to my expectations, early adopter nursing homes have lower overall average occupancy rates. It would seem likely that higher average census would allow a facility to achieve economies of scale, thereby freeing resources allowing it to innovate. Nyman 1987, however, suggests that facilities operating with high occupancy rates have less incentive to provide quality care; comparably, it may be that these facilities also have less incentive to innovate. Also, facilities with lower overall average occupancy rates may view themselves as less constrained with regard to time pressures in routine daily care tasks. Clearly, this is also dependent upon staffing levels, but lower overall average occupancy rates may enable facilities to innovate.
In preliminary analyses (not shown), I included other independent control variables, in addition to ADLs, to adjust for residents' acuity in each facility. These variables included: bladder incontinence, bowel incontinence, pressure ulcers, restraint use, psychotropic drug use, and the percentage of residents with mental health problems. My rationale for including these variables was the belief that resident sickness in general, and increasing acuity levels as a result of the PPS, may promote innovation. Although these variables were not significant, their direction, and that of the ADL variable used in the analyses, suggest that if anything, wellness may be an engine of change. It may be that in facilities with high resident acuity, staff may have less time available to implement innovations. This may be an area for subsequent investigation.
In one of the few studies addressing innovation in nursing homes, Banaszak-Holl and coworkers 1996 used similar data to examine market and organizational characteristics. They also used similar dependent variables, the provision of Alzheimer's and subacute special care units. This study is different in that it was longitudinal and used more recent data. The principal difference between this study and this previous study lies in the operationalization of the dependent variable; I examined early adoption of innovations, whereas Banaszak-Holl and colleagues 1996 examined any adoption of innovations. As a result of these differences, some caution is required in comparing the studies. It is worth noting, however, that some of the findings are similar. Increased facility size and higher numbers of private-pay residents are common significant organizational factors associated with innovation. Competition and Medicaid reimbursement rates are common significant market factors associated with innovation. The competition variables are especially robust, in that a total of six different competition variables are examined in the two analyses, three of which are significant. In both studies the greatest innovation response was to competition from other nursing facilities in the market.
In comparing the findings of Banaszak-Holl and colleagues 1996 with this study, some results are also dissimilar. These differences include significant findings for CON, case-mix adjustment, and for-profit status (Banaszak-Holl et al. 1996). Some trepidation is required in comparing significant findings from one study with nonsignificant findings in another; however, for-profit status may be noteworthy. Banaszak-Holl and coworkers identified not-for-profit facilities to be strongly associated with innovation, whereas I find no significant results for facility ownership. It may be that ownership has little effect on the early adoption of innovations, but not-for-profit facilities are more likely to persist with the innovation.
The reasons for performance gaps in innovation between nursing homes remain largely unexplained, but are of interest to the industry and policy makers (Nelson and Mullins 1985). An important challenge facing nursing homes is the attempt to become more innovative. Organizational and market characteristics can be important in demonstrating which factors might be most compatible with innovation (Kim 1980). Clearly, the practical implications from some results, such as average occupancy, are limited. Few legislators or facilities are likely to advocate reducing average occupancy rates in the interests of increasing innovation. The results for competition and staffing have more utility.
The nursing home industry has some barriers to free market competition, including CON legislation, new construction moratoria, and regulated reimbursement rates. Some debate has been aimed at promoting more competition in the industry. This study shows that one potential benefit of promoting more competition would be more early innovation adoption. I do not include competition from such important substitutes for nursing home care as board-and-care homes and home care providers in my analyses. Examining whether competition from these providers affects early innovation adoption may also be relevant to policy.
Caregiver staffing provisions are included in many recent pieces of nursing home legislation. For example, the Nursing Home Reform Act (NHRA) mandated that nurse aides must complete 75 hours of training and pass a competency examination. I do not assess training, but this study supports this legislative emphasis on caregivers.
An understanding of the differences of innovation in the hospital and nursing home industries may be useful, especially for policy makers concerned with both. Although this study does not answer this question, some commonalities are evident when combining the results of this study with other nursing home studies and comparing these with hospital-based studies. Facility size and competition are especially robust.
The implications of this study, of course, must be tempered against researchers' relatively undeveloped understanding of innovation in general. The innovation literature is very clear in stating that the knowledge base is inconclusive and inconsistent (Abrahamson 1991). Research is needed in several areas, including types of innovations, stages of the innovation process, outcomes of innovation, and the attributes of the innovations examined. In a review of the literature, Wolfe 1994 expands upon each of these areas.
Limitations of the Study
Although the longitudinal study design has desirable statistical properties for this analysis, several limitations are worth noting. First, some nursing homes do not seek either Medicare or Medicaid certification and are not included in the OSCAR (these facilities represent only 7% of all nursing home beds nationally). Second, OSCAR data are collected for each nursing home at one point in time. The collection of data in this way is limited in that the prevalence of factors such as the use of staff may either be under- or overestimated. Other factors, such as ownership, may alter between inspections. To the extent that this occurs, some of the organizational factors may be misrepresented. Similarly, the market variable measuring presence of a CON or new construction moratoria does not capture the historical influence of policies to limit bed supply.
Managed care may influence early innovation adoption. Binstock and Spector 1997 identify managed care as a potential important influence on long-term care facilities. Banaszak-Holl and colleagues 1996 find HMO penetration significant for some innovations. In preliminary analyses I attempted to include a variable representing managed care penetration. Data for this variable proved to be either unreliable or unavailable for all time periods. Clearly, more research is needed to understand the influence of managed care on early innovation adoption.
The proportion of Medicare recipients in a facility may also influence early innovation adoption (Banaszak-Holl et al. 1996). For example, Medicare recipients are more likely to require subacute care, and as such a higher proportion of these residents may foster innovation in the provision of this care (Banaszak-Holl et al. 1996). In preliminary analyses I examined the proportion of Medicare recipients. My results were not significant. This may be because I examined 13 innovations, some of which may not be influenced by the proportion of Medicare recipients; alternatively, the proportion of Medicare residents may not influence early innovation adoption but may influence innovation in general. The differences between early innovation adoption by a facility and innovation in general may be an area for subsequent investigation.
Shortell and colleagues 1995 have shown that a narrow focus on a small number of outcome measures can be misleading and may lead to erroneous, or incomplete, conclusions. For similar reasons, Kimberly and Evanisko 1981 have advocated examining several innovations. By including 13 types of innovations, my approach provides a more generalizable picture of early innovation by nursing homes. Given the limited number of studies in this area, I feel this approach to be an appropriate starting point. In using this approach, however, my analyses may be underspecified in that they may not include specific variables such as cost or quality factors influencing the adoption of individual innovations. For example, some states have provided increased Medicaid reimbursement for special care units, a factor that I do not account for. As such, more specific innovation studies are also warranted. Examining costs may be especially productive. Organizational types may differ in their innovation adoption based on costs. For example, for-profit homes may be early adopters of cost-reducing innovations. I would have also liked to include other innovations examined by previous researchers, such as use of TQM and computerization of medical records. However, these variables are not included in the OSCAR data.
Some conceptual limitations of this study are also worth noting. For example, this study was undertaken with the belief that identifying characteristics associated with early adoption of innovations could be useful in further facilitating the diffusion of innovations in the nursing home setting. This may be the case, but it is also representative of the positive bias that pervades innovation research (Kimberly 1981). Certainly, not all innovations are beneficial, and organizations may fail as the result of innovation adoption. Moreover, it is possible for organizations to superficially adopt innovations—for the purposes of prestige, for example (Downs and Mohr 1975). In addition, early adopters do not always have the highest degrees of commitment to the innovation (Downs and Mohr 1975).
I also assume that the facilities I examined are early innovators. This is based on the assumption that special care units and subacute care services will continue to diffuse throughout the industry. If these innovations do not continue to diffuse, then my analyses include both early and late adopters. If this were the case, my analyses would not be representative of early adopters, but represent a comparison of any innovation adoption compared to nonadoption. However, given the trend of the 1992 to 1997 data and examining the more recently available 1999 data, these innovations are continuing to diffuse throughout nursing homes. A related point is that my analyses examine early adopters as compared to nonadopters of innovations. When the data become available, and innovation diffusion has been sufficient, more refined analyses could compare early, late, and nonadopter facilities.
Conclusion
This analysis examines the impact of market and organizational characteristics on the early adoption of innovations. Admittedly, market and organizational characteristics are but two of several factors that could influence the provision of innovations. Other important factors could include influential physicians and nursing home administrators. These are no less important and should also be examined. However, a nursing home will innovate, or will not, insofar as routines and practices embodied in the facility promote innovation. These routines and practices may reflect an amalgam of individual skills and need not correspond to any one individual's learning or skills. Moreover, routines and practices may persist because of past experience and "lessons" within the organization, regardless of its current employees (Levitt and March 1988). As Zmud 1984(p. 727) has described, organizations can have a "receptivity toward change." This analysis is important in that it shows that organizational and market characteristics of nursing homes affect their receptivity toward early adoption of innovations.
These findings add to the body of literature on innovation in nursing homes. However, as with most research, several intriguing questions follow from this study. First, consideration may be given to the question of whether organizational and market characteristics affect the rate of innovation adoption. Second, examining how innovation is diffused among early adopters and if this is different from later adopters may offer further insight into the mechanics of early adoption. The extent of innovation adoption may be influenced by organizational and market characteristics and may also be different among early adopters and later adopters. Finally, some refinement in innovation attributes may add to the consistency of research results (Damanpour 1991). For example, administrative versus technical, radical versus incremental, high-risk versus low-risk, and product versus process are refinements commonly used, and each may be associated with distinct organizational and market characteristics.
Received for publication March 6, 2000. Accepted for publication October 11, 2000.
|
References |
|---|
|
TOPAbstractBackgroundHypothesesMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  C. L. Coogle, I. A. Parham, and K. A. Young Job Satisfaction and Career Commitment Among Nursing Assistants Providing Alzheimer's Care American Journal of Alzheimer's Disease and Other Dementias, September 1, 2007; 22(4): 251 - 260. [Abstract] [PDF]
|
|
|
|
 |
|
 D. Segaar, C Bolman, M. Willemsen, and H De Vries Identifying determinants of protocol adoption by midwives: a comprehensive approach Health Educ. Res., February 1, 2007; 22(1): 14 - 26. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 O. Intrator, Z. Feng, V. Mor, D. Gifford, M. Bourbonniere, and J. Zinn The Employment of Nurse Practitioners and Physician Assistants in U.S. Nursing Homes Gerontologist, August 1, 2005; 45(4): 486 - 495. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. A. Lucas, T. Avi-Itzhak, J. P. Robinson, C. G. Morris, M. J. Koren, and S. C. Reinhard Continuous Quality Improvement as an Innovation: Which Nursing Facilities Adopt It? Gerontologist, February 1, 2005; 45(1): 68 - 77. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Aylward, P. Stolee, N. Keat, and V. Johncox Effectiveness of Continuing Education in Long-Term Care: A Literature Review Gerontologist, April 1, 2003; 43(2): 259 - 271. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
