| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
Correspondence: Address correspondence to Mary Whelan Carter, PhD, Assistant Professor, Center on Aging, School of Medicine, West Virginia University, HSC Annex, PO Box 9127, Morgantown, WV 26506-9127. E-mail: mcarter{at}hsc.wvu.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Key Words: Hospitalizations • Hospital use • Hospital transfer • Quality of care practices
National pressure to limit rising health care costs, especially rapidly increasing Medicaid expenditures, has sparked considerable interest in the health services utilization patterns of aged beneficiaries who qualify for both Medicare and Medicaid programs. Given their relatively high utilization rates of both acute hospital care and long-term care, dually eligible nursing home residents have been seen as a logical target for cost control efforts. At the same time, however, efforts to reduce health care spending via targeting dually eligible beneficiaries have raised concerns regarding potentially adverse effects in terms of health care quality and accessibility (Stuart & Weinrich, 1998). Thus, current goals to balance cost containment strategies versus access to and quality of care pose a significant policy challenge.
Repeated calls have been made to reduce hospitalization rates among nursing home residents (Gillen, Spore, Mor, & Freiberger, 1996). Hospitalization of nursing home residents may involve considerable disruption and relocation stress (Ouslander, 2000), may complicate existing and/or trigger new illnesses (Spector & Takada, 1991), may result in new or worsening pressure sores (Berlowitz, Brandeis, Anderson, Du, & Brand, 1997), and may be followed by irreversible, functional decline (Creditor, 1993). At the same time, however, many of the hospitalizations of nursing home residents are undoubtedly necessary on medical grounds. Yet, currently, benchmarks are lacking for determining which hospital transfer rates are acceptable for nursing home residents. As a consequence, uninformed policies run the risk of unilaterally limiting access among medically needy nursing home residents.
Relatively few studies have examined facility and/or market area correlates of nursing home resident transfer rates to hospitals. Instead, most studies have tended to concentrate solely on resident-level factors related to resident outcomes and have reported inconsistent findings. In a review of the literature, Castle and Mor (1996) reported that transfer rates varied by as much as two to five times across studies, with some studies reporting annual nursing home hospitalization rates nearing 50%. To what extent do these variations reflect actual differences in resident case-mix versus variations in propensity to hospitalize related to differences in facility management and operation styles, or even broader geographic market influences?
This article reports on an effort to examine empirically the extent to which individual resident risk factors, facility-level structural factors, and geographic market factors contribute to variations in hospital admission rates among nursing home residents. The current study's main focus involves how nonclinical factors (e.g., facility proprietary status, facility staffing patterns, etc.) affect the risk of hospitalization among nursing home residents after medical need factors have been controlled. Greater understanding of the role of these contextual factors should provide additional insight regarding the salient connections between nursing home and hospital care, as well as informing future research addressing appropriate hospitalization practices for nursing home residents.
Previous Research
Many earlier hospital transfer studies of nursing home residents used medical chart review methodologies on small study samples to determine whether nursing home residents were being unnecessarily transferred to hospitals, and whether adverse consequences were observed upon return to the facility. Teresi, Holmes, Bloom, Monaco, and Rosen's study (1991) of 229 residents in one facility found infections, gastrointestinal disorders, and peripheral arterial insufficiency to be the leading medical conditions requiring hospitalization, as well as high rates for 30-day hospital readmissions (50%) and incident pressure ulcers (30%) upon a resident's return to the nursing home. In a similarly conducted study of three nursing homes, Kayser-Jones, Wiener, and Barbaccia (1989) found that a lack of on-site medical support services, such as radiology and laboratory equipment, in two of the facilities resulted in transfers that could have been avoided. They concluded that 70% of the transfers were potentially avoidable had trained staff been available to administer intravenous fluids. This finding sharply contrasts with the 7% rate of avoidable hospital transfers reported by Bergman and Clarfield (1991) using similar study methods.
Likewise, studies using large secondary databases have produced varied results as well. Baker and colleagues (1994) merged population-based data from the Monroe County (New York) long-term care program with administrative hospital claims data to estimate a hospital transfer rate of nearly 40% over a 2-year period. This rate was much higher than that found in other studies of its kind, including the estimate of 25% obtained by Weissert and Scanlon (1985) with data from the 1977 National Nursing Home Survey.
However, over the last decade, a broader picture of what contextual factors influence nursing home hospitalization rates has emerged. Freiman and Murtaugh (1993) estimated an econometric model of resident hospitalizations with resident, nursing home, and economic market variables on a large secondary database. In addition to several significant resident risk factors, the risk of hospitalization was found to be lower in nonprofit nursing homes and facilities receiving higher reimbursement rates. Similarly, Murtaugh and Freiman (1995) also found nonprofit status to be negatively associated with increased risk of hospitalization. Additionally, the work of Fried and Mor (1997) identified several variables suggestive of nursing home care practices and/or practice styles, such as percentage of residents: receiving new medications, with feeding tubes present, and with pressure ulcers (Stage 2+), to be associated with increased risk of hospitalization.
Most recently, Intrator, Castle, and Mor (1999) used cross-sectional data from the 1993 Minimum Data Set and the 1994 On-line Survey Certification of Automated Records to specify a broad set of resident and facility attributes in a multinomial logit model of the competing risks of hospital transfer and mortality among nursing home residents. Their findings revealed that increased nursing home use of medical professionals (physicians, nurse practitioners, and physician assistants) decreased resident risk of hospitalization. Special care units in the nursing home also appeared to lessen the risk of hospital transfer, whereas nursing homes with a greater proportion of residents using respirators had greater odds of being hospitalized, suggesting that practice styles and organizational attributes affect hospital use by nursing home residents.
Overall, however, few facility-level factors have consistently been found in past research efforts to affect hospitalization. Table 1 contrasts four key studies exploring the link between facility attributes and hospitalization risks of nursing home residents. Included in the table are study size, data source(s), facility measures, and magnitude of effects found.
|
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Because facility-level contextual factors are the central focus of this study, 6,205 records (8%) with missing facility-level data in two or more data fields were dropped, leaving 72,319 person-quarters available for analysis. Nursing homes with missing data tended to be slightly newer, served a lighter case-mix of residents, had fewer Medicare reimbursed days as a percentage of all paid resident days, and were more likely to have recently changed ownership. Where only one field of data was missing, standard imputation techniques using multivariate regression analysis of facilities with complete data were employed (Harrell, 2001).
Because Medicaid uses the MMQ data for case-mix reimbursement in Massachusetts, nursing homes have a financial incentive to overstate the nursing needs of residents. Porell, Caro, and Silva (1993) found that, although total MMQ scores reported by facilities were about 6% higher than those of state auditors, agreement rates for all of the individual health and functional status items exceeded the conventional reliability threshold of 80%. Additional details about the reliability analysis and analytic file construction can be found elsewhere (Porell et al., 1993, 1998; Porell & Caro, 1998).
The analytic MMQ file described previously was augmented by merging information from four calendar years (1990–1993) of individual Medicare Provider Analysis and Review hospital claims obtained for all Medicaid residents with at least one quarter of MMQ data over the 3-year study time period spanning March 1991–1994. Hospital records were first assigned to quarter-years on the basis of admission dates, aggregated by quarter, and subsequently merged to quarterly MMQ records of residents. Additionally, ICD-9-CM (International Classification of Diseases, 9th Revision, Clinical Modification) codes of principal diagnoses from past hospitalizations were used to assign values to residents for a prior use diagnostic risk adjustment variable (Ellis & Ash, 1989).
Measures
Table 2
contains definitions and data sources for all variables by conceptual groupings.
|
|
Independent Variables
Resident Characteristics
Demographic information for nursing home residents from MMQ data were used to specify variables for age, gender, and newly admitted status—factors shown in past studies to influence hospitalization risk (Castle & Mor, 1996). Racial status was not specified because of insufficient variability, as nearly 97% of all race values were classified as White. Nursing home length of stay was also not specified; it is likely to be a function of facility care practices in the nursing home and, therefore, potentially an endogenous variable.
Resident Clinical Characteristics
The most compelling reason for using person-quarter observations rather than person-years stems from the extensive clinical data in the MMQ, which permit quarterly updates of resident diagnoses and functional status, allowing for dynamic case-mix adjustment for population differences over time. Using ICD-9-CM codes, dummy variables were first specified for the 15 most prevalent diagnoses responsible for nursing home care in the state. Remaining clinical conditions were then coded into a set of residual diagnostic categories organized by body system.
Other Health/Risk Indicators
An indicator variable indicating the presence of decubitus ulcers (defined as Stage 2 or higher) or having sustained a reported accident within the past 90 days were specified as indicators of potentially serious health complications (Berlowitz et al., 1997). Reported accidents include documentation of falls or other, nonroutine events in the nursing home leading to suspected injury among residents. Measures of nutritional well-being, including calorie malnutrition and vitamin deficiencies have been used as quality of care indicators because of their relationship to overall health and well-being (Flanagan, Monane, Chawla, Schroeder, & Tillisch, 1997). A dummy indicator of potential malnutrition was included. Use of physical restraints has long been used as an indicator of poor quality of care practices in the nursing home and associated with poorer resident outcomes (Tinetti, Liu, Marottoli, & Ginter, 1991). A dummy variable was included to indicate the daily or as needed (PRN order) use of physical restraints in the past 90 days. Also, dummy indicators for hypnotic, tranquilizer, antipsychotic, and antidepressant drug use were also specified given their noted relationship to poor health outcomes (Castle, 1999; Flanagan et al., 1997). The total MMQ score, an aggregate measure reflecting overall nursing care needs, was specified as a proxy measure of a resident's frailty. Finally, three diagnostic cost group (DCG) dummy variables were specified to distinguish residents with certain medical conditions, placing them at much higher risk of subsequent hospitalization because of the nature of their illness and its treatment by physicians.
Ash, Porell, Gruenberg, Sawitz, and Beiser (1989) and Ellis and Ash (1995) used diagnostic information from prior hospitalizations to develop a set of health status-based DCG risk classifications intended for risk adjustment of Medicare health maintenance organization capitation payments. The DCG risk classifications are based on the principal diagnosis of inpatient hospitalizations judged to be nondiscretionary (i.e., physicians have little discretion but to hospitalize for treatment). The DCG risk categories are intended to reflect health status differentials, in the sense that persons assigned to higher risk DCG categories have illnesses that are associated with much higher than average expected future Medicare (Parts A and B) costs. For example, whereas Medicare beneficiaries hospitalized with a principal diagnosis of malignant neoplasm of the liver (ICD-9-CM code 155) have much higher than average expected annual Medicare costs, those beneficiaries who were not recently hospitalized, or who were hospitalized with a principal diagnosis such as genital prolapse (ICD-9-CM code 618), have lower than average expected annual Medicare costs. Although more complex hierarchical coexisting condition diagnostic risk classification models incorporating diagnostic information from both outpatient claims and inpatient hospital claims have been developed (Ellis et al., 1996), outpatient claims data were not available for the study population. Additionally, Porell and Gruenberg (2000) report that changes in the interpretation of ICD-9-CM codes over time have had no remarkable impact on DCG risk classifications.
The seven higher risk DCG risk classes of Ellis and Ash (1995), ordinally ranked on the basis of higher expected annual Medicare costs, were first collapsed into three classes: DCG 1–2, DCG 3–4, and DCG 5–7. In each quarter, the principal diagnoses from all hospitalizations in the previous four quarters (1 year) were compared for each resident. Each resident was then assigned to the highest corresponding DCG risk class observed across the three classes over 1 year, or to an omitted reference risk class. Residents who were assigned to the omitted reference risk class were either: (1) not hospitalized at all in the past year or (2) hospitalized for conditions that were rated as highly discretionary and/or not empirically associated with high subsequent year Medicare costs (Ellis & Ash, 1995).
Facility Characteristics
Facility-level contextual variables were specified for 527 Medicaid eligible nursing homes throughout Massachusetts. Intermediate care facilities (ICFs) were distinguished from skilled nursing facilities by use of a dummy variable. Some research has suggested that higher Medicare skilled nursing facility reimbursement rates encourage nursing homes to hospitalize for the purpose of converting a resident's Medicaid payor status to Medicare upon return to the facility (Weissert & Musliner, 1992). Because many ICFs in Massachusetts do not have any Medicare certified beds, and most of these facilities have a lighter resident case-mix, the distinction between ICFs from nursing facilities with Medicare-certified beds will likely reflect the net effect of several factors, but are expected to be associated with a lesser risk of hospitalization.
Much interest has surrounded the question of whether quality of care differences can be attributable to the profit status of the facility (Spector, Selden, & Cohen, 1998). To control for this possibility, a dummy variable indicating nonprofit status was included, as was a dummy variable indicating management by an operating chain versus otherwise. Because nonprofit facilities have been linked to better care processes, residents residing in nonprofits should exhibit lesser risk of hospitalization. A continuous variable measuring years of operating tenure by the current facility owner was included, as was a dummy variable indicating recent change in ownership. Operating tenure should capture, at least in part, differences relating to experience in providing care, whereas change in ownership should provide a proxy measure for potential instability associated with ownership turnover. Several studies have indicated that Medicare and Medicaid status affect outcomes of nursing home residents (Freiman & Murtaugh, 1993; Shaughnessy, Kramer, Schlenker, & Polesovsky, 1985). To investigate whether care styles associated with a facility's payor mix influence decisions to hospitalize, two facility-level measures were specified as the number of paid Medicare and Medicaid days, respectively, each as a percentage of total annual days of care expressed in 10-unit increments. Because financial incentives may affect facility decisions to hospitalize, a measure of the availability of cash resources was also specified as an indicator of financial operating health. To the extent that poor financial health curtails available resources for resident care and/or to the extent that financial incentives, such as Medicaid hold days, may exist to hospitalize residents, hospitalization risks should be lower in facilities with greater cash flow. Additionally, a summary count of all OBRA (Omnibus Budget Reconciliation Act) deficiency citations received during a facility's last state inspection was specified to tap into variations in quality-of-care practices across study facilities, with the expectation that residents residing in homes with fewer deficiencies will experience decreased risk of hospital transfer. Although the recent Institute of Medicine's (2001) report on improving quality in the nursing home encourages the continued use of deficiency citations for quality studies, it should be noted that some limitations with the measures have been cited, including variation in regulatory practices across states and a national trend toward downward reporting of deficiencies (Harrington & Carrillo, 1999).
Nursing staff expenses account for the bulk of facility operating costs, and nursing staff levels have been advocated as quality-of-care indicators (Davis, 1991). Measures of registered nurse (RN) and licensed practical nurse (LPN) expenses, each expressed as a proportion of total annual facility nursing expenses, were specified to reflect potential differences in care practices associated with the relative mix of nursing staff. Residents residing in facilities that rely more heavily on non-RN nursing staff are expected to have greater risk of hospitalization. Lastly, a facility average MMQ score was included as an indicator of the overall resident case-mix with respect to nursing care needs. To the extent that greater care demands place greater requirements on existing nursing personal, residents in homes with a higher case-mix of residents are expected to experience increased risk of hospitalization.
Market Characteristics
Community-based population studies have long revealed wide variations in regional hospital utilization rates that are much more highly correlated with measures of physician and hospital bed supply than population health status differences (Wennberg & Gittelsohn, 1982). Although relatively little is known about the influence of such broad market factors on nursing home residents' hospital use, similar influences are expected because most nursing home residents will be cared for by physicians from the local community. Thirteen county-level market dummy variables were specified to capture geographic market variations as "fixed effects," potentially stemming from regional differences in such attributes as available number of hospital beds and nursing home beds, area income levels, and area population density (Phillips, Morrison, Anderson, & Aday, 1998; Stump, Johnson, & Wolinsky, 1995). Although the effects of specific market area attributes such as hospital bed supply are not known, the estimated coefficients for the county dummy variables should nevertheless reflect the net effect of geographic factors, because there is considerable intra- and intercounty variation in nursing facility attributes.
Analytic Methods
Because a resident is no longer at risk of hospitalization after death, mortality is a competing risk with inpatient hospitalization. A few researchers have addressed the competing risk posed by death by estimating a multinomial logit model in which a death that is not preceded by hospitalization is specified as a distinct outcome from hospitalization (inclusive of an admission followed by death), and no hospitalization, respectively (Freiman & Murtaugh, 1993; Intrator et al., 1999). Although there are merits in this modeling approach, specifying death without hospitalization as an additional outcome does not take into account the briefer exposure of decedents to the risk of hospitalization. We followed an alternative approach to the problem by assigning fractional case-weights to all quarterly observations in which a resident death occurred (Gruenberg, Kaganova, & Hornbrook, 1996; Weiner et al., 1996). Regardless of whether death occurred before or after at least one hospitalization, case-weights were set equal to the proportion of days a decedent was alive and at risk for hospitalization in the quarter. Hence, quarterly observations with deaths are essentially treated as partially censored observations. Because such censoring may be selective, we performed sensitivity analyses typically used for estimation of survival/hazard models in which there is concern about selective censoring from a competing risk such as death (Allison, 1984). The empirical results were found to be extremely robust with respect to how deaths were treated.
Because most nursing home residents had multiple quarterly observations resulting from the longitudinal nature of the study data, the standard errors of all parameter estimates were adjusted to account for the expected nonzero covariance among the errors arising from repeated observations for residents over time, with a maximum likelihood procedure developed independently by Huber (1981) and White (1980). The Huber-White procedure is available as an option for logistic regression in the statistical package STATA.
Given the large number of independent variables specified in the model, reverse stepwise logistic regression was used to test the stability of the empirical results. If a fully specified model is fragile, notable parameter shifts will be observed among significant variables when insignificant variables are dropped. Using three different threshold p values (.30,.20, and.10), no significant variables became insignificant, while the significance levels for most of the remaining variables were enhanced by the removal of weaker variables. These analyses indicate a robust fully specified model.
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
contains descriptive statistics for all variables used in the multivariate analysis. The study sample contained 72,319 person-quarter observations for dually eligible residents of 527 nursing homes in Massachusetts between April 1991 and May 1994. A resident was hospitalized at least once in about 11% of all person-quarters. The majority of residents were female (79%). The most prevalent primary diagnoses listed as responsible for resident nursing care needs included: hypertension (30%), dementia (21%), diabetes (15%), congestive heart failure (12%), osteoarthritis (11.5%), and Alzheimer's disease (11%). The frailty level of the study population is reflected in the high prevalence (78%) of person-quarters for residents with at least four activity of daily living limitations.
|
|
Multivariate Analysis Results
Logistic regression analysis was used to estimate a model in which the probability of experiencing at least one hospitalization in the subsequent quarter t +1 is specified to be a function of resident, facility, and market area effects (as listed in Table 2
) measured in quarter t. Table 4 presents the empirical results, including estimated coefficients, z-statistics, and the corresponding odds ratios (ORs) for statistically significant coefficient estimates (p <.05). The pseudo-R2 square measure of model fit of.09 was statistically significant (p <.001). Because idiosyncratic factors should have greater influence on whether a hospital admission occurs with shorter time intervals of observation, the model fit is quite good. Although hospital use is measured over 3-month periods, the pseudo-R2 is comparable with values typically reported for binary logit models of annual hospital use.
|
|
strongly support the main study hypothesis that facility-level contextual factors are associated with observed variability in risk of hospital transfer among nursing home residents. Specifically, 11 of the 14 facility-level variables met conventional thresholds for statistical significance. Generally speaking, the findings indicate that residents residing in nonprofit facilities, facilities with higher overall MMQ case-mix, facilities classified as ICFs in Massachusetts, facilities with more operating beds, and those with a greater proportion of resident days reimbursed by Medicare exhibit a significantly lower risk of hospital transfer in the next quarter than otherwise similar residents in other nursing facilities. In contrast, residents residing in facilities operated by management chains, facilities with a greater percentage of Medicaid reimbursed resident days, and facilities that spent a greater proportion of total nursing expenses for LPNs appear to be at greater risk of hospitalization in the next quarter, holding all other factors constant. The estimated parameter for proprietary status suggests that, for residents living in nonprofit nursing homes, the odds of being hospitalized in the next quarter are about 9% less in comparison with otherwise similar residents living in for-profit facilities. The risk of hospitalization appears to be greater among residents of nursing homes operated by a management company. The estimated OR of 1.07 suggests that residents who reside in a nursing home operated by a management company have 7% greater odds of experiencing a hospital transfer than residents residing in independently managed facilities, when all other factors are held constant. Residents of facilities that spend a greater proportion of total nursing expenses for LPNs also experience substantially greater risk (OR = 1.30) of hospitalization than otherwise comparable residents in other nursing homes. This increased risk of hospital transfer may be the result of a greater reliance on LPNs for some medical decision-making. Such facilities may be exchanging professional training for a less costly labor force, which may, however, potentially threaten resident quality of care.
The risk of hospitalization appears to vary with the overall payor mix and resident nursing needs case-mix as well. Residents of facilities where Medicaid paid days comprise a greater percentage of total annual nursing home days exhibit an increased hospitalization risk (OR = 1.10). On the other hand, residents of facilities with a greater percentage of Medicare paid days appear to have a lesser risk of hospitalization (OR =.89). To the extent that a higher concentration of Medicaid residents (vs. Medicare and/or private pay residents) in a nursing home indicates poor quality-of-care practices, these results suggest that poor care practices in facilities with higher concentrations of Medicaid residents may underlie the increased hospitalization risk. Finally, there was a negative association between average facility MMQ case-mix score and risk for hospitalization in the next quarter (OR =.87). This result is contrary to our expectations, which held that hospitalization rates would be higher among residents in facilities with heavier care demands overall. Most likely, this finding reflects unmeasured aspects of the type of facilities that exhibit higher average facility MMQ scores. That is, certain nursing homes may have special attributes, such as intravenous therapy or wound healing clinics, that attract a higher case-mix of residents, but that also allow for more complex treatments to be provided in the nursing home, thus reducing the odds of hospitalizations for certain types of illnesses.
The final two facility attributes with statistically significant coefficients include facilities that were classified as ICFs in Massachusetts and the count of annual care-related deficiency citations received by a facility. Residing in an ICF was associated with substantially lower odds of hospitalization (OR =.77), holding other factors constant. Surprisingly, higher rates of deficiency citations were also associated with lower odds of hospitalization (OR =.99). Although the results for the ICF classification met directional expectations given that ICFs generally house a much healthier nursing home population, the results for the deficiency citation measure were unexpected. Conceivably, to the extent that the propensity to hospitalize in a nursing home reflects, at least in part, aggressiveness of care tied to its organizational structures, this same tendency may also account for better performance on the assessment procedures now in use for inspection purposes, contributing to fewer deficiency citations.
Magnitude of Facility Contextual Effects
Because few nursing facilities tend to differ from others on only one attribute at a time, there is difficulty in substantively gauging the cumulative impact of facility-level factors in our empirical findings. The logistic regression model parameter estimates reported earlier in Table 4 were used to develop estimates of resident case-mix adjusted hospitalization rates for individual nursing facilities throughout the state. A common set of sample mean values was specified for all resident-level variables and county dummy variables for each case, whereas actual values were left specified for each of the facility-level variables. Mean predicted values were then computed for each of the 527 nursing homes throughout the state of Massachusetts. Figure 1 illustrates the expected probability of hospitalization risk for the "typical" nursing home resident moving through each facility, while holding geographic market effects constant. Because the predicted values of hospitalization risk are based on observed combinations of facility attributes, they should portray a realistic sense of the magnitude of the cumulative influence that contextual factors have on the hospitalization rates of nursing home residents in Massachusetts.
|
Market Characteristics
The estimated coefficients for county dummy variables suggest that rates of hospital transfers for nursing home residents still exhibit considerable variation across market areas, even after resident and facility factors have been controlled. Nine of the thirteen regional areas demonstrated significantly increased risk of hospitalization in comparison with the omitted reference category, Barnstable County. For example, residents of nursing homes located in Suffolk County, home to the City of Boston, experienced 111% greater odds of hospitalization than otherwise similar residents of facilities located in Barnstable County. Rural counties, such as Berkshire and Hampshire, located in the far western portion of the state, appeared no more (less) likely to be hospitalized than residents of Barnstable, suggesting that risk of hospitalization, at least in part, reflects broader hospital bed and physician supply factors more commonly associated with urban settings. Figure 2 illustrates the magnitude of area market effects by estimating the expected probability of hospitalization risk for the "typical" resident from the "typical" facility moving through each county, revealing a 2.39-fold difference between the counties with the lowest (Nantucket) and highest (Norfolk) probability estimates.
|
Consistent with expectations, several resident-level indicators of compromised quality of care practices were positively associated with an increased risk for hospitalization, lending support for their specification as potential determinants of hospitalization. Compared with otherwise similar residents, residents with decubitus ulcers (OR = 1.23), sudden and unplanned changes in body weight (OR = 1.14), or history of accident in the past 90 days (OR = 1.16) experienced greater odds of hospitalization in the next quarter. Additionally, with the exception of antipsychotic drug use, the findings reveal positive associations between increased hospitalization risk and psychotropic drug use.
The DCG risk classifications based on diagnostic information from prior hospitalizations are the most novel health status indicator to be used in this study and produced the largest odds ratios among all model covariates, after newly admitted status. The odds of hospitalization in the next quarter are increased between 100% and 186% for residents with at least one prior low discretion hospital stay in the past year for a condition classified in one of the three DCG risk categories relative to otherwise similar residents with no prior nondiscretionary hospitalization with future cost implications. The strong findings reported here suggest that diagnostic measures from prior hospitalizations offer an effective method for empirically discriminating illness acuity levels among nursing home residents.
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
The handful of empirical studies that heretofore examined facility-level and market-level factors affecting hospitalization of nursing home residents have reached different conclusions about the influence of contextual factors. The current study differs by nature of the careful attention given to health status risk adjustment. Drawing from the capitation rate setting literature to specify more sensitive measures of a resident's clinical risks of hospitalization (in the form of DCG risk classifications), our empirical findings impart some useful observations about the relative influence of facility-level factors on the hospitalization rates of nursing home residents. Namely, structural attributes of delivering long-term care—such as type of ownership, payment sources to facilities for care provided, and staff remuneration practices, all affecting hospitalization of nursing home residents—clearly emerge as important components of variation in nursing home hospitalization outcomes. This finding takes on particular significance in light of the current policy milieu of efforts to secure cost savings. Given ongoing efforts toward containing reimbursement practices for both acute care hospitals and nursing homes, concerns regarding possible incentives for resident shifting to increase profits arise. More specifically, as more nursing homes face the prospect of reimbursement rates that are tightly tied to case-mix assessments, the financial incentives that encourage the classification of residents into higher care need categories may provide little incentive for nursing homes to attempt to decrease morbidity among residents (Weissert & Musliner, 1992).
This concern takes on added significance given the significant association found between LPN versus RN staffing patterns in a facility and increased hospitalization risk. Residents of nursing homes with nursing personnel expenses more heavily allocated to LPNs versus RNs were at greater risk of hospitalization than otherwise similar residents of other nursing facilities. Although the reasons for why nursing staffs comprised more of LPNs versus more highly trained RNs are associated with greater hospitalization risk remain uncertain, plausible explanations include that LPNs generally may be: (1) less capable of identifying medical symptoms and initiating effective interventions, (2) less capable of managing complicated care regiments, and (3) more likely to depend on hospital transfers to treat medical conditions as they arise. Moreover, our findings join a growing body of literature linking nurse staffing ratios and outcomes of care. Recent efforts to examine nurse staffing patterns and resident outcomes have revealed similar findings in terms of direction and magnitude of effects (e.g., Castle & Fogel, 1998) and have lead to the policy recommendation to increase nurse-staffing requirements in nursing homes (Harrington, Zimmerman, Karon, Robinson, & Beatel, 2000).
Conclusions to be drawn from the proceeding discussion raise concerns regarding the role of contextual factors on nursing home outcomes, particularly with respect to shaping policies aimed at reducing hospitalization rates among nursing home residents. Specifically, the strong role of nursing home facility attributes in explaining variations in risk of hospital transfer among residents may reflect quality-of-care practices and/or care philosophies regarding the role of rehabilitative and restorative care over more palliative models. The extent to which undesirable nursing home effects exist, however, may be further confounded by the degree to which facilities lack needed equipment, trained and adequate numbers of personnel, and supportive amenities to care for certain residential demands. For example, Intrator and colleagues (1999) found that residents of facilities with special care units and increased physician services (including physicians' extenders) had reduced odds for hospitalizations.
Ideally, of course, hospitalizations occurring from inadequate resources in the nursing home, and thus precluding in-house treatment, would not be targeted for immediate reductions, but rather more comprehensive long-term solutions should be sought, such as increasing on-site services and better targeting of medical needs with facility resources. Hospitalizations resulting from poor care practices, on the other hand, should immediately be addressed by identification of and improvement in care practices in the nursing home. Given the difficulties in identifying specific hospitalizations resulting from poor quality-of-care practices versus inadequate on-site resources, one practical approach to reducing potentially avoidable hospitalizations may be to identify potential contextual factors associated with those hospitalizations stemming from medical conditions most likely to respond to improved daily care practices.
Although many of our empirical findings have fairly broad policy implications, our study population may not be representative of all nursing homes and all residents of nursing homes. Only nursing home residents whose nursing home stays were reimbursed by Medicaid were included in this study, leaving questions about factors affecting hospital use by private pay residents and Medicare beneficiaries in nursing homes largely unexplored. Restricting the study sample to one state may also limit the generalizability of our findings, as prior research suggests that the state of Massachusetts as a whole may exhibit some distinctive market characteristics compared with other states and/or the national average (Weinstein, Freedman, & Randle, 1995). Additionally, the reliance on administrative data for case-mix adjustment poses some study limitations as well. Administrative data lack detailed description of medical conditions and may be more likely to report clinical conditions related to reimbursement rates rather than describing individual health conditions in detail. Also, data limitations affected the specification of nursing home attributes in the study. Several clinical staffing variables—such as physician visits and physician extenders, shown elsewhere (e.g., Intrator et al., 1999) to be important components in avoiding unnecessary hospitalizations—were not available for use in the present study, limiting overall understanding of how such factors contribute to variations in hospitalization rates. Likewise, an inability to control for do-not-hospitalize orders in the present study may limit our understanding of the role of specific facility care philosophies to the extent that facilities with certain approaches to providing end-of-life care may be more or less inclined to respect do-not-hospitalize orders.
Although significant variations in hospitalization rates were observed across counties in this study, the basis for these geographic variations and the extent to which they reflect overutilization versus underutilization of inpatient hospital care remains unknown. It seems reasonable to expect that at least some of the observed variation may reflect a lack of necessary medical care and/or timely access to that care, especially given the large literature base establishing geographic variations for community-based populations. Here, study findings suggest a complex arrangement of organizational and system factors affecting the same population moving between two different health care settings, resulting in varied levels of hospitalization risk, which are, at least in part, unrelated to clinical need. Thus, further attempts to understand the interaction between facility-level and market-level effects, and how these factors affect hospitalization risk among nursing home residents, represent a key area for further research. These continuing efforts are needed to ensure that reduction in hospitalization rates occur only among medically unnecessary cases, and, at the same time, improve the targeting of hospital care to those nursing home residents most apt to benefit medically from inpatient hospital care.
|
Footnotes |
|---|
1 Division of Health Services Research and Policy, School of Public Health, University of Minnesota, Minneapolis.
2 Gerontology Center, University of Massachusetts, Boston.
Received for publication June 21, 2001. Accepted for publication January 29, 2002.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  D. Liu and N. G. Castle Health Information Technology in Nursing Homes Journal of Applied Gerontology, February 1, 2009; 28(1): 38 - 58. [Abstract] [PDF]
|
|
|
|
|
|
N. G. Castle Nursing Home Caregiver Staffing Levels and Quality of Care: A Literature Review Journal of Applied Gerontology, August 1, 2008; 27(4): 375 - 405. [Abstract] [PDF]
|
|
|
|
 |
|
 N. M. Bellows and H. A. Halpin MDS-Based State Medicaid Reimbursement and the ADL-Decline Quality Indicator Gerontologist, June 1, 2008; 48(3): 324 - 329. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. T. Konetzka, W. Spector, and M. R. Limcangco Reducing Hospitalizations From Long-Term Care Settings Med Care Res Rev, February 1, 2008; 65(1): 40 - 66. [Abstract] [PDF]
|
|
|
|
 |
|
 F. H. Decker Dying in a Nursing Home: The Role of Local Bed Supply in Nursing Home Discharges J Aging Health, February 1, 2008; 20(1): 66 - 88. [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]
|
|
|
|
 |
|
 M. W. Carter and F. W. Porell Vulnerable populations at risk of potentially avoidable hospitalizations: The case of nursing home residents with Alzheimer's disease American Journal of Alzheimer's Disease and Other Dementias, November 1, 2005; 20(6): 349 - 358. [Abstract] [PDF]
|
|
|
|
|
|
P. D. Sloane, S. Zimmerman, A. L. Gruber-Baldini, J. R. Hebel, J. Magaziner, and T. R. Konrad Health and Functional Outcomes and Health Care Utilization of Persons With Dementia in Residential Care and Assisted Living Facilities: Comparison With Nursing Homes Gerontologist, October 1, 2005; 45(suppl_1): 124 - 134. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
