The Impact of State Nursing Home Staffing Standards on Nurse Staffing Levels

594733MCRXXX10.1177/1077558715594733Medical Care Research and ReviewPaek et al. research-article2015 Empirical Research The Impact of State Nursing Home Staffing Standards on Nurse Staffing Levels Medical Care Research and Review 2016, Vol. 73(1) 41 61 The Author(s) 2015 Reprints and permissions: sagepub.com/journalspermissions.nav DOI: 10.1177/1077558715594733 mcr.sagepub.com Seung Chun Paek 1, Ning J. Zhang 2, Thomas T. H. Wan 3, Lynn Y. Unruh 3, and Natthani Meemon 1 Abstract This study investigated the impact of state nursing home staffing standards on nurse staffing levels for the year 2011. Specifically, the study attempted to measure state staffing standards at facility level (i.e., nurse staffing levels that each individual nursing home must retain by its state staffing standards) and analyzed the policy impact. The study findings indicated that state staffing standards for the categories of registered nurse, licensed nurse, or total nurse are positively related to registered nurse, licensed nurse, or total nurse staffing levels, respectively. Nursing homes more actively responded to licensed staffing requirements than total staffing requirements. However, nursing homes did not increase their staffing levels as much as those required by state staffing standards. It is possibly because the quality-oriented inspection allows flexibility in nursing homes control of nurse staffing levels. Keywords state nursing home staffing standards, nurse staffing, U.S. nursing homes This article, submitted to Medical Care Research and Review on December 25, 2014, was revised and accepted for publication on June 1, 2015. 1 Mahidol University, Salaya, Thailand 2 Seton Hall University, South Orange, NJ, USA 3 University of Central Florida, Orlando, FL, USA Corresponding Author: Seung Chun Paek, Department of Society and Health, Faculty of Social Sciences and Humanities, Mahidol University, 999 Phuttamonthon 4 Road, Salaya 73170, Thailand. Email: seungchun.pak@mahidol.ac.th

42 Medical Care Research and Review 73(1) Nursing home staffing standards have become a major long-term care policy issue because of the importance of nurse staffing to quality of nursing home care (Harrington, 2005a, 2005b; Wan, Breen, Zhang, & Unruh, 2010). In 1987, the federal government officially established national nursing home staffing standards through Omnibus Budget Reconciliation Act (OBRA 87) in response to the concern about poor quality of care in nursing homes (Wiener, 2003). The 1987 staffing standards require all nursing homes certified for Medicare and Medicaid to have (a) a registered nurse director of nursing (RN DON), (b) a registered nurse (RN) on duty 8 consecutive hours per day for 7 days a week, (c) a licensed nurse (LN) either RN or licensed vocational nurse (LVN)/licensed practical nurse (LPN) on duty for 24 hours per day for 7 days a week (including the required 8 RN hours), and (d) a minimum of 75 hours of training for nursing aides (NA). The standards allow RN DON and RN to be the same individual for nursing home with fewer than 60 residents. In addition, the law requires that facilities have sufficient nursing staff to provide nursing services to maintain the highest levels of physical, mental, and psychosocial well-being of residents (OBRA 87). Despite OBRA 87, poor quality-of-care problems in nursing homes have been cited in many government reports, and some of them indicated that the quality problems are partially due to nurse staffing issues such as inadequate nurse staffing levels and poor quality staff (U.S. General Accounting Office, 1998, 1999a, 1999b, 1999c; U.S. Office of the Inspector General, 1999a, 1999b). Accordingly, various field experts have called for stronger federal staffing requirements, widely investigating minimum nurse staffing levels for nursing homes (Harrington, 2002, 2005b; Harrington, Kovner, & Harrington, 2000; U.S. Centers for Medicare & Medicaid Services [CMS], 2000, 2001; N. J. Zhang, Unruh, Liu, & Wan, 2006). Nevertheless, the federal government has not changed its staffing standards since OBRA 87. Subsequently, many states have established and continuously updated their own nursing home staffing standards. The state staffing standards are more stringent than the federal ones in general and vary considerably across states. According to Mueller et al. (2006), 40 states, by 2004, had their own stronger staffing standards, while 11 states followed the federal staffing standards. Converting the 33 states staffing standards to hours per resident day (HPRD) unit, the study found that Oregon had the lowest total nursing HPRD (1.76 HPRD) and Florida had the highest one which is 3.60 total nursing HPRD. Florida increased the total staffing requirement to 3.90 HPRD in January 2007 but decreased again to 3.60 HPRD in July 2011. As the relationship between poor quality of care and insufficient nurse staffing has been widely demonstrated, and each state, in response to such concern, has established its own staffing standards which have different levels of stringency, it may be an important question to ask whether the stringency of state staffing standards has made any positive impact on nurse staffing levels in nursing homes. Therefore, the study aims to investigate the relationship between state nursing home staffing standards and nurse staffing levels in nursing homes.

Paek et al. 43 New Contribution Previous studies examined the impact of state nursing home staffing standards on nurse staffing levels in nursing homes. Harrington (2005a, 2005b) found that median nurse staffing levels in nursing homes (3.16 HPRD) are greater than state average minimum standards (2.32 HPRD) for the year 1999 to 2001. Mueller et al. (2006), reviewing state staffing standards for the year 2004, found that facilities in states with high staffing standards (above 2.50 total staffing HPRD) have higher staffing levels than states with low (below 2.50 total staffing HPRD) or no staffing standards (adhering to the federal staffing requirements), while there is no significant difference in facility staffing levels among states with low and no staffing standards. A study conducted by Harrington, Swan, and Carrillo (2007) found that state RN staffing standards have positive impact on both RN and total nurse staffing levels. Park and Stearns (2009) found that changes in staffing standards from 1998 to 2001 (steady-state effects) are positively associated with nurse staffing levels in low-staff, nonprofit facilities. Bowblis (2011) reviewed state total staffing requirements (called minimum direct care staffing requirements in the study) for the year 1998 to 2004 and showed that the state policy has positive impact on total nurse staffing levels. Also, the state policy was significantly related to higher RN staffing mix for nursing homes with more Medicaid residents. For nursing homes with less Medicaid residents, the state policy was significantly related to higher LN staffing mix. Last, a recent study conducted by Mukamel et al. (2012) showed that state staffing standards are positively related RN and NA hours adjusted for case mix. State staffing standards are much more complex than federal ones and differ considerably across states. Thus, how the state staffing standards are measured may be one of critical issues for evaluating the policy impact. States usually require different nurse staffing hours and the type of nurse according to the facility size (either the number of residents or beds in general). In other words, each individual nursing home, by its state staffing standards, must have different types of nurse and different staffing hours according to how many residents or beds it retains. Furthermore, states regulate their staffing standards using various forms. They usually include the form of minimum staffing hours or the number of staff by shift and the form of staff-to-resident ratio or HPRD. Since the different forms are in conflict with one another according to the facility size, the different forms should be considered together when the staffing standards are measured. For this reason, state-level measurement used in the previous studies could wash out any positive or negative effects of the staffing policy on different size nursing homes. Therefore, combining these points with the previous studies, this study would add to the body of knowledge on how state staffing policy influences actual nurse staffing levels. Specifically, this study attempted to measure state staffing standards at facility level, which are states required nurse staffing levels for individual nursing homes (i.e., nurse staffing levels that each individual nursing home must retain by its state staffing standards). And then, how the states required nurse staffing levels are related to actual nurse staffing levels was investigated.

44 Medical Care Research and Review 73(1) Theory and Literature Review The nursing home market is one of the most highly regulated markets in the United States (Kumar, Norton, & Encinosa, 2006; X. Zhang & Grabowski, 2004). Both federal and state governments have jointly regulated the minimum standards of resident care and safety that all nursing homes must meet to provide Medicare and Medicaid services. Nursing homes compliance is monitored through the annual survey and certification process, and violations of the federal and state standards may lead to sanctions such as civil monetary penalties, depending on the scope and severity (Harrington, Mullan, & Carrillo, 2004). Under state staffing standards, nursing home s staffing levels will be reviewed for the case that violations of the federal quality of care requirements are identified (Harrington, 2005a). Resource Dependence Perspective Characterizing organization as an open system inevitably dependent on contingencies in the external environment, resource dependence theory seeks to explain how environmental uncertainty influences organizations and how organizations manage or adapt overtime (Pfeffer & Salancik, 1978; Shortell & Kaluzny, 2006). The theory premises that no single organization can control all the resources necessary for survival and accordingly depend on its environment which controls the vital resources. While such dependency makes external constraint and control over organization behavior, organizations can actively negotiate with their environment by exercising managerial strategies to reduce unwanted dependencies and enhance survivability. This study, employing the resource dependence perspective, attempts to demonstrate nursing homes strategies on managing their resources in response to state s demand for nurse staffing levels together with organizational structures and characteristics of market. Nursing Homes Control of Nurse Staffing Levels Due to high degree of government involvement in the nursing home market, the government can be seen as the most important regulator and resource provider that nursing home must depend on (X. Zhang & Grabowski, 2004). State staffing standards may affect nursing homes decisions about the type and amount of nursing staff employed. Yet compliance with the staffing standards varies across facilities. Under the resource dependence perspective, organizations apply different strategies to manage their own resource and exchange relationships with the environment (Nienhuser, 2008). Since inspection of nurse staffing adequacy will happen only if serious quality problems are identified, this quality-oriented inspection allows flexibility in nursing homes control of nurse staffing levels. The state staffing standards may only mitigate but do not limit nursing homes autonomous decision making as they tend to struggle for control over their own management and resources, particularly when the values

Paek et al. 45 and interests (i.e., ensuring acceptable quality vs. securing internal resources) are in conflict (Clegg & Rura-Polley, 1998). As such, nursing homes may try to maintain their autonomy in controlling staffing level to secure internal resources through managerial strategies (e.g., staff management emphasizing staff organization, supervision, and motivation) to ensure acceptable quality as well as maintain a certain degree of compliance with the state staffing standards. In general, nursing homes are not expected to violate the minimum staffing standards required by the state (Pfeffer & Salancik, 1978). Therefore, our main hypothesis is that: Nursing homes in states with stronger nursing home staffing standards will have higher nurse staffing levels than those in states with lower nursing home staffing standards, controlling for organizational and environmental factors. Organizational Factors The prediction of nursing homes decision on nurse staffing levels has been based on organizational factors that link to how internal resources can be mobilized and secured. As residents with higher case-mix levels require more extensive care in general (Wunderlich & Kohler, 2001), higher acuity facilities have been found to have higher overall staffing levels (or RN and/or LPN staffing levels) (Harrington, Carrillo, Mullan, & Swan, 1998; Harrington et al., 2007; Harrington & Swan, 2003; Mueller et al., 2006). In addition, case-mix reimbursement methods have been used for Medicare reimbursement and increasingly adopted by many sates (35 states by 2004) for Medicaid reimbursement (Zinn, Feng, Mor, Intrator, & Grabowski, 2008). Thus, higher acuity facilities can take advantage of higher reimbursement and consider having higher staffing levels (Harrington et al., 2007). The proportion of Medicare and Medicaid residents has been considered as important factors influencing nursing homes decision about nurse staffing levels. As stated by many studies, since Medicare reimbursement has higher profit margins than Medicaid one, higher proportion of Medicare residents were related to higher staffing levels, and in contrast, higher proportion of Medicaid residents were related to lower staffing levels (Harrington et al., 1998; Harrington et al., 2007; Harrington & Swan, 2003; Mueller et al., 2006). Hospital-based facilities have more Medicare residents, have higher acuity levels, and have more short-term intensive care residents (Harrington et al., 2007; Harrington, Zimmerman, Karon, Robinson, & Beutel, 2000) thus, have been found to have higher nurse staffing levels (Mueller et al., 2006; Wunderlich & Kohler, 2001). Chainaffiliated facilities have been reported to have significantly low nurse staffing levels as compared with freestanding ones (Kovner & Harrington, 2000; Mueller et al., 2006; Mukamel et al., 2012). A recent study found that for-profit chains have much lower staffing levels than both nonprofit chains and freestanding facilities (Harrington et al., 2012). For-profit nursing homes, possibly because of their organizational nature of profit seeking (Banaszak-Holl, Zinn, & Mor, 1996; McKay, 1991; Zinn, Castle, Intrator, & Brannon, 1999; Zinn, Mor, Feng, & Intrator, 2007), have been reported to have lower nurse staffing levels than nonprofit ones (Harrington et al., 1998; Harrington

46 Medical Care Research and Review 73(1) et al., 2007; Harrington, Olney, Carrillo, & Kang, 2012; Harrington & Swan, 2003; Mueller et al., 2006). Larger nursing homes have been found to have lower nurse staffing levels. This could be partially due to economies of scale in caring for residents (Harrington et al., 1998; Harrington et al., 2007; Kovner & Harrington, 2000; Mukamel et al., 2012). In addition, compliance with both federal and state staffing standards (particularly minimum-staffing-hours requirement) must be met regardless of the facility size and the number of residents that facilities retain (Harrington & Swan, 2003). Occupancy rate has been found to be negatively related to nurse staffing levels due to the same reason (Harrington & Swan, 2003; Mueller et al., 2006). Finally, the study hypotheses regarding organizational factors highlights the nurse staffing levels in directional association with: resident case mix (+), proportion of Medicare residents (+) and Medicaid residents ( ), hospital-based facilities (+), chain-affiliated facilities ( ), for-profit facilities ( ), facility size ( ), and occupancy rate ( ). Environmental Factors Organizations use managerial strategies in response to the environment in which the amount of resources available would allow them to improve their structures (Pfeffer & Salancik, 1978). State Medicaid reimbursement rates have been found to be positively associated with nurse staffing levels in nursing homes (Harrington et al., 2007; Harrington & Swan, 2003; Mueller et al., 2006). As Medicaid reimbursement rates are set partially on the basis of facility costs including nurse staffing, nursing homes in states with higher Medicaid reimbursement rates may have more sufficient financial resources available for increasing their staffing levels. In addition, nursing homes may operate nurse staffing levels even far greater than the state minimum standards if that would improve the chance of securing or attracting potential resources. Since nursing homes in highly competitive market should inevitably share prospective nursing home residents, they may more perceive market competition in the shared pool of limited resources as threats to their survival than those in a less competitive market (Zinn, Weech, & Brannon, 1998). Since higher nurse staffing levels has been proved to associate with better quality of care (Akinci & Krolikowski, 2005; Harrington, Zimmerman, et al., 2000; Kim, Harrington, & Greene, 2009), increasing nurse staffing level may be one of managerial strategies to be more competitive in the market. Nursing homes that are perceived to have better quality can attract more residents and eventually dominate more resources. Last, market demand may also imply available resources in the environment. The use of nursing home care was found to increase considerably for people aged 65 years and older (Kemper & Murtaugh, 1991). Thus, the demand for and use of nursing home services would increase in that market boundaries. Additionally, the higher proportion of the aged 65 years and older adults, in combination with the population s declining physical and mental functioning, may increase overall case-mix levels in nursing homes (Harrington et al., 2007; Harrington & Swan, 2003). Therefore, nursing homes located in areas with a higher percentage of older adults are expected to have higher

Paek et al. 47 nurse staffing levels. Finally, the study hypothesizes nurse staffing levels to have directional association with these following environmental factors: state Medicaid reimbursement rates (+), market competition (+), and market demand (+). Method Data Sources and Variables This study used a cross sectional design with four different data sets which are (a) State Nursing Home Staffing Standards of 2011, (b) Average State Medicaid Reimbursement Rates of 2007, (c) Online Survey Certification and Reporting System (OSCAR) of 2011, and (d) Area Resource File of 2011. State nursing home staffing standards of 2011 were obtained through each website of the states department of health and human services. Additionally, this study referred to Harrington s published study titled Nursing Home Staffing Standards in State Statutes and Regulations (Harrington, 2008, 2010) when state staffing rules or regulations were not available through the Internet. The published study specifies nursing home staffing requirements of the 50 states and the District of Columbia in detail. Furthermore, the staffing requirements were converted to a number by estimating HPRD for a 100-bed nursing home to obtain standardized values of nurse staffing levels required by states. Average state Medicaid reimbursement rates were obtained from the research titled State Data Book on Long Term Care, 2007: Program and Market Characteristics (Harrington et al., 2008). These data represent the average Medicaid reimbursement rate for nursing homes in dollars from the 50 states and the District of Columbia. Since the 2007 data are not matched with this study s target year (2011), we obtained some available data sets (1999, 2000, 2001, 2002, and 2007) and examined their patterns to check possibility of using the 2007 data for the study. Similar distribution pattern was noticed through all data sets with correlation coefficients ranging from 0.82 to 0.98. OSCAR, as a national database of all nursing homes federally certified for Medicaid and Medicare in the United States, includes three types of comprehensive facilitylevel information including (a) facility characteristics, including all categories of nurse staffing; (b) resident census and characteristics; and (c) deficiency citations about regulatory compliance of nursing homes. For this study, OSCAR 2011 was used to obtain information about nurse staffing levels and specified organizational factors. Last, Area Resource File 2011 which is a national county-level health resources information system was used to obtain environmental factors which are county-level market competition and market demand which is percentage of population aged 65 years and older. Dependent Variables Nurse staffing levels in nursing homes were measured in HPRD unit by using the following formula: (FTE 70 14) Number of Total Residents. This formula

48 Medical Care Research and Review 73(1) multiplies the nurse staff payroll FTEs (Full-time equivalent) reported for a 2-week period by 70 hours for the period, and then divides by 14 days and the total number of residents in the reporting period. This study divided nurse staffing levels into four different categories according to the categories that states commonly have used in their staffing requirements: (a) RN (RN DON and RN); (b) LN (RN DON, RN, and LVN/LPN); (c) total (RN DON, RN, LVN/LPN, and NA); and (d) NA. For the RN category, this study combined RN DON with RN to measure nurse staffing HPRD for the reason that many states, like the federal standards, allow RN to serve as RN DON for smaller homes, while they require a separated body of RN DON for larger homes. Independent Variables This study measured state nursing home staffing standards at facility level, which are states required nurse staffing levels for nursing homes (i.e., nurse staffing levels that each individual nursing home must retain by state staffing standards) by the following steps: (a) state nursing home staffing standards were classified into four nurse categories (RN, LN, total nurse, and NA) like the dependent variables; (b) required minimum hours for each category were divided by the number of residents in nursing homes in order to convert the required hours to HPRD unit; (c) if states have duplicated staffing requirements within or between nurse categories, more stringent ones were selected; assuming that federal staffing standards preempt state staffing standards; (d) for states which do not clearly indicate their minimum requirements, the federal requirements were applied; and (e) if states required staffing levels were lower than the federal requirements, the federal ones were applied. For the facility-level measurement, different forms within or between nurse categories in state staffing standards were considered together since the different forms are in conflict with one another according to the facility size. For LN staffing category, when a state uses different forms for the same category of nurse, this study converted the state s required staffing levels to HPRD unit and then applied more stringent one to each individual nursing home. For instance, Florida requires both 24 hours and 1.0 HPRD for LN staffing category. In HPRD unit, for nursing homes with fewer than 24 residents, the 24 hours requirement is more stringent than the 1.0 HPRD requirement. For total nurse staffing category, this study considered states LN requirements together with their total staffing HPRD since for smaller nursing homes, LN requirements are usually more stringent than total staffing requirements. For example, Illinois requires both 24 LN hours and 2.5 total staffing HPRD. For nursing homes with nine residents, the required total staffing levels will be 2.7 HPRD instead of 2.5 HPRD. Among the 50 states and the District of Columbia, 33 states required total nurse staffing levels were finally obtained. While 31 states have their own total staffing requirements, other 2 states (OR and SC) do not have specific total staffing requirements but have NA staffing requirements. Thus, for OR and SC, their LN and NA requirements were summed up together to obtain those states required total staffing levels.

Paek et al. 49 States required NA staffing levels were also obtained from the same 33 states above. Among the 33 states, for seven states (DE, FL, MT, OH, OR, SC, and VT) which have their own NA requirements, this study directly used their required NA staffing levels. For other 26 states which do not have their NA requirements but have their LN and total staffing requirements, their required NA staffing levels were measured by subtracting their required LN staffing levels from their required total nurse staffing levels. State- and facility-level measurements of all 50 states and District of Columbia are presented in the appendix. Control Variables For organizational factors, facility size represents the total number of beds in each facility. Ownership is measured as categorical variable representing three categories: for-profit, nonprofit, and government-owned nursing homes. Chain affiliation is used as a dichotomous variable coded as 1 for chain-affiliated and 0 or nonchain-affiliated status. Occupancy rate is quantified by total number of residents divided by total number of beds. The percentage of Medicaid and Medicare residents is measured by the ratio of the number of residents with Medicaid and Medicare residents to the total number of residents in each facility. Resident acuity index is used for resident case mix. Resident acuity index, which is the aggregated facility level, represents the severity of residents living in nursing homes, reflecting both activities of daily living and health status measures. For environmental factors, Market competition was measured by the Herfindahl n Hirschman index and calculated as: H H index = (number of beds in a nursing home/total number of beds in a county) 2, where i is the number of nursing homes in a county. Higher value of the H-H score indicates less competition. Market demand was measured by the percentage of people aged 65 years or older in the county where a nursing home is located. Last, state Medicaid reimbursement rates are the dollar amount of average daily payment rates in state level. Operational definition of the study variables are presented in Table 1. Data Cleaning and Analyses To eliminate outliers and erroneous values, data cleaning rules used in relevant literatures using OSCAR were applied as follows: this study excluded (a) facilities with more residents than beds or no residents, (b) facilities reporting no total nursing HPRD (RN + LPN/LNV + NA) or more than 24 total nursing HPRD, and (c) facilities in the top 1% and bottom 1% within each staffing category that this study used to eliminate outliers having extremely high or low values. Since facilities in the top 1% and bottom 1% within each staffing category were excluded independently, four different data sets (RN, LN, total, and NA data sets) were used separately for the study analyses. i= 1

50 Medical Care Research and Review 73(1) Table 1. Operational Definitions of the Study Variables. Variables Operational definition Sources Dependent variables RN staffing levels RN staffing HPRD in nursing home OSCAR LN staffing levels LN staffing HPRD in nursing home OSCAR Total staffing levels Total staffing HPRD in nursing home OSCAR NA staffing levels NA staffing HPRD in nursing home OSCAR Independent variables States required RN staffing levels States required LN staffing levels States required RN staffing HPRD for nursing home (all states) States required LN staffing HPRD for nursing home (all states) Nursing home staffing standards from website of each state s States required total staffing States required total staffing HPRD for department of levels nursing home (33 states a ) health and human States required NA staffing States required NA staffing HPRD for services and levels nursing home (33 states b ) Harrington (2010) Control variables Organizational factors Facility size Total number of beds in nursing home OSCAR Acuity index Resident acuity index in nursing home OSCAR Ownership 1 = for-profit; 2 = nonprofit; 3 = OSCAR government Chain affiliation 1 = chain-affiliated; 0 = nonchain-affiliated OSCAR Hospital affiliation 1 = hospital-based; 0 = nonhospital-based OSCAR Occupancy rate Total number of residents divided by total OSCAR number of beds in nursing home Percent Medicare The number of Medicare residents divided OSCAR by total number of residents in nursing home Percent Medicaid The number of Medicaid residents divided OSCAR by total number of residents in nursing home Environmental factors Market competition Herfindahl Hirschman index in county OSCAR and ARF Market demand Percent of 65 or over population in county ARF State Medicaid reimbursement rate Each state s average daily Medicaid reimbursement rate in dollar Harrington et al. (2007) Note. RN = registered nurse; HPRD = hours per resident day; OSCAR = online survey certification and reporting system; LN = licensed nurse; NA = nursing aides; ARF = Area Resource File. RN = RN DON and RN; LN = RN DON, RN, and LPN/LVN; total = RN DON, RN, LPN/LVN, and NA. a Thirty-one states required total staffing HPRD for nursing home was obtained from their total staffing requirements, while two states (OR and SC) required total staffing HPRD was obtained by summing up their required LN and NA staffing levels. b Seven states (DE, FL, MT, OH, OR, SC, and VT) required NA staffing HPRD obtained from their NA staffing requirements, while 26 states required NA staffing HPRD was obtained by subtracting their required LN HPRD from their required total staffing HPRD. OSCAR 2011 originally includes a total of 15,703 facilities. Of those facilities, 202 facilities with missing or unidentified values in ownership status, 11 facilities with

Paek et al. 51 more residents than beds or no residents, and 78 facilities with no total nursing HPRD or more than 24 total nursing HPRD were excluded. Additional data cleaning was done for each data set. For RN and LN data sets, facilities in the top 1% and bottom 1% (150 facilities each) within each staffing category were excluded independently. For total and NA data sets, 18 states which do not have their own total and NA staffing standards were excluded. Then, facilities in the top 1% and bottom 1% (100 facilities each) within each staffing category were excluded independently. But, for NA category, smaller nursing homes that comply with their state LN requirements can hypothetically have no NA staffing levels since that will not violate their state total staffing requirements. Thus, for NA data set, facilities only in the top 1% (101 facilities due to rounding) were excluded. After data cleaning, a total of 15,112 facilities for the RN and LN staffing analyses, 10,498 facilities for the total staffing analysis, and 10,597 facilities for the NA staffing analysis remained, accounting for 96.2%, 96.2%, and 97.1% of all nursing facilities in OSCAR 2011, respectively. The study employed hierarchical linear modeling (HLM) to investigate the impact of state nurse staffing standards on nurse staffing levels in nursing homes. HLM, also known as multilevel analysis, allows variance in outcome variables to be analyzed at multiple hierarchical levels, while in linear regression, all effects are modeled to occur at a single level (Singer, 1998). Nursing homes are nested within states the variables of interest are in two different levels including facility level (nurse staffing level in nursing homes) and state level (state nursing home staffing standards). Therefore, HLM is appropriate for the purpose of dealing with the nested data (Mueller et al., 2006). The HLM was conducted by using Proc Mixed of SAS program, treating all 51 intercepts (or 51 state effects) as randomly varying. Specifically, the study performed HLM analysis for each of three nurse categories (RN, LN, and total). For NA category, since only seven states (DE, FL, MT, OH, OR, SC, and VT) regulate their NA staffing requirements, the policy impact may not be sufficiently representative. Furthermore, for other 26 states, this study measured their required NA staffing levels by subtracting their required LN staffing levels from their required total nurse staffing levels. In this sense, variation in the NA policy stringency across states may not be accurately captured because it depends greatly on states required LN and total staffing levels. Thus, for NA category, this study performed only descriptive analysis. Results Descriptive statistics for the study variables are presented in Table 2. Since facilities in the top 1% and bottom 1% within each staffing category were excluded independently, four different data sets were used separately for the analyses. In Table 2, descriptive statistics for the dependent variables (nurse staffing levels) and independent variables (states required nurse staffing levels) were obtained from their respective data sets. But, descriptive statistics for control variables were obtained from the data set for RN staffing analysis because we found that there is not much variation in descriptive statistics of the variables among the four data sets.

52 Medical Care Research and Review 73(1) Table 2. Descriptive Statistics for the Study Variables (n = 15,112). Variables M or percent SD Minimum Maximum Dependent variables RN staffing levels (n = 15,112) 0.529 0.465 0.078 4.143 LN staffing levels (n = 15,112) 1.373 0.655 0.530 6.137 Total staffing levels (n = 10,498) 3.764 1.046 1.151 10.371 NA staffing levels (n = 10,597) 2.342 0.671 0 5.110 Independent variables States required RN staffing levels 0.219 0.196 0.010 6 (n = 15,112) States required LN staffing levels 0.555 0.487 0.022 24 (n = 15,112) States required Total staffing levels 2.646 0.620 1.022 24 (n = 10,498) States required NA staffing levels 2.032 0.539 0 4.615 (n = 10,597) Control variables For-profit 68.52% Nonprofit 25.70% Government 5.78% Chain-affiliated 54.45% Nonchain-affiliated 45.55% Hospital-based 6.15% Nonhospital-based 93.85% Facility size 108.988 63.977 4 1,389 Occupancy rate 0.822 0.158 0.014 1 Percent Medicare 0.153 0.155 0 1 Percent Medicaid 0.603 0.230 0 1 Acuity index 10.237 1.568 3.643 23.288 Market demand 0.143 0.037 0.056 0.434 Market competition 0.205 0.241 0.003 1 State Medicaid reimbursement rate 145.254 32.326 99.580 384.160 Note. RN = registered nurse; LN = licensed nurse; NA = nursing aides. For seven states (DE, FL, MT, OH, OR, SC, and VT), descriptive statistics of actual NA staffing levels in nursing homes (N = 2,106) are (M = 2.592; SD = 0.667; Minimum = 0; and Maximum = 5.110), while those of states required NA staffing levels are (M = 2.291; SD = 0.455; Minimum = 0; and Maximum = 4.615). For dependent variables, the mean of RN staffing levels in nursing homes was 0.53 HPRD. The means of LN, total, and NA staffing levels were 1.37, 3.76, and 2.34 HPRD, respectively. For independent variables, the mean of states required RN and LN staffing levels were 0.22 and 0.56 HPRD, respectively, while the means of states required total nurse and NA staffing levels were 2.65 HPRD and 2.03 HPRD, respectively. The medians of states required staffing levels (not presented in Table 2) were

Paek et al. 53 0.18 (RN), 0.50 (LN), 2.62 (total), and 2.00 (NA) HPRD. Additionally, the means of actual and required NA staffing levels for the seven states (DE, FL, MT, OH, OR, SC, and VT) which have their own NA staffing requirements were 2.59 and 2.29 HPRD, respectively (presented in the footnote of Table 2). Overall, Table 2 shows that actual nurse staffing levels, on average, were much greater than state staffing standards for all categories of nurse. For control variables, all facilities averaged 109 beds and had an average occupancy rate of 82.2%. Among those facilities, 68.52% were for-profit, 25.70% were nonprofit, and 5.78% were government-owned facilities. Of all facilities, 54.45% were chain-affiliated, while 45.55% were independent. Last, an average of state Medicaid reimbursement rates were $145.25. The results of HLM analysis are presented in Table 3. To examine variations in nurse staffing levels in nursing homes within and between states, intraclass correlations, which are between-state variance as a proportion of total variance, were calculated for unconditional models without entering any independent and control variables as fixed effects. The intraclass correlations indicated that about 89% of variance in RN staffing levels is explained within states, while about 11% of variance in RN staffing levels is explained between states. For the LN and total staffing models, about 95% and 91% of variances in LN and total staffing levels were explained within states, while 5% and 9% of variances in LN and total staffing levels was explained between states, respectively. Thus, nurse staffing levels varied more within states than between states for all three categories of nurse. For independent variables, higher states required RN (0.861), LN (0.445), and total nurse staffing levels (0.202) were found to be significantly associated with higher nurse staffing levels in nursing homes. Rejecting the null hypothesis, it indicates that nursing homes in states with stronger RN, LN, and total nurse staffing requirements had higher RN, LN, and total nurse staffing levels, respectively. The coefficients specifically mean that states that differ by 1.0 HPRD in state staffing standards differ by 0.861, 0.445, and 0.202 HPRD in their actual RN, LN, and total staffing levels, respectively. For organizational factors, both facility size and occupancy rate, as expected, were negatively related to nurse staffing levels for all categories of nurse. Like the study expectation, higher acuity index was found to be significantly related to higher nurse staffing levels in all three staffing models. For proportions of Medicare and Medicaid residents, as expected, higher proportion of Medicare residents were related to higher RN, LN, and total staffing levels, respectively, and higher proportion of Medicaid residents were related to lower RN, LN, and total staffing levels, respectively. Nonprofit nursing homes were found to have higher nurse staffing levels than forprofit ones for all nurse categories as expected. As compared with government-owned nursing homes, nonprofit ones had relatively low nurse staffing levels except RN category. Like the study expectation, chained nursing homes, as compared with independent ones, were negatively related to nurse staffing levels for LN and total categories, but they were not significantly related to RN staffing levels. Last, hospital-based nursing homes were related to higher RN, LN, and total nurse staffing levels as expected.

54 Medical Care Research and Review 73(1) Table 3. Results of Hierarchical Linear Modeling Analysis. RN staffing model LN staffing model Total staffing model (n = 15,112) (n = 15,112) (n = 10,498) Variables Estimate SE Estimate SE Estimate SE States required RN staffing levels 0.8613** 0.0168 States required LN staffing levels 0.4451** 0.0094 States required total 0.2022** 0.0257 staffing levels Facility size 0.0003** <0.0001 0.0006** 0.0001 0.0021** 0.0001 Occupancy rate 0.2345** 0.0198 0.6396** 0.0283 1.3857** 0.0617 Acuity index 0.0166** 0.0018 0.0559** 0.0026 0.0945** 0.0054 Percent Medicare 0.8575** 0.0230 1.1965** 0.0333 1.3605** 0.0721 Percent Medicaid 0.1353** 0.0157 0.1836** 0.0227 0.5174** 0.0489 For-profit (vs. 0.0677** 0.0066 0.0891** 0.0094 0.3449** 0.0205 nonprofit) Government (vs. 0.0094 0.0122 0.0416* 0.0175 0.1793** 0.0392 nonprofit) Chain (yes vs. no) 0.0034 0.0054 0.0183* 0.0078 0.1799** 0.0168 Hospital-based (yes 0.4771** 0.0121 0.6580** 0.0172 0.7858** 0.0364 vs. no) Market competition 0.0960** 0.0135 0.2134** 0.0192 0.2078** 0.0445 Market demand 0.3617** 0.0828 0.7230** 0.1184 0.7681** 0.2594 State Medicaid reimbursement rate 0.0010* 0.0004 0.0020** 0.0005 0.0048** 0.0013 Unconditional model variance Level 1 (Facility) 0.1988** 0.0023 0.4121** 0.0048 0.9977** 0.0138 Level 2 (State) 0.0258** 0.0058 0.0238** 0.0059 0.0957** 0.0256 Fitted model residual variance Level 1 (Facility) 0.0949** 0.0011 0.1955** 0.0023 0.6447** 0.0089 Level 2 (State) 0.0118** 0.0025 0.0157** 0.0035 0.0504** 0.0136 Reduction in residual variance Level 1 (Facility) 52% 53% 35% Level 2 (State) 54% 34% 47% Note. RN = registered nurse; LN = licensed nurse. *Significant at the 0.05 level. **Significant at the 0.01 level. For environmental factors, higher state Medicaid reimbursement rates, as expected, were found to be related to higher nurse staffing levels in all three staffing models. For market competition, nursing homes in highly competitive market had higher RN, LN, and total nurse staffing levels as expected. Unlike the study expectation, market demand was found to be negatively related to nurse staffing levels for all categories of nurse.

Paek et al. 55 Discussion The study investigated the impact of state nursing home staffing standards on nurse staffing levels for the year 2011. Specifically, this study attempted to measure state staffing standards at facility level and analyzed the policy impact. The study findings indicated that nursing homes in states with stronger staffing standards for the categories of RN, LN, or total nurse are more likely to have higher RN (0.861), LN (0.445), or total (0.202) nurse staffing levels, respectively. The results may imply that nursing homes could more actively respond to licensed staffing requirement than total staffing requirement. When states increase RN, LN, or total staffing requirement by 1.0 HPRD, nursing homes would increase their respective nurse staffing levels by 0.861 (RN), 0.445 (LN), or 0.202 (total) HPRD. Thus, if future policy aims to increase actual nurse staffing levels, increasing licensed staffing requirement may be more effective, though additional costs for increasing licensed and total staffing requirements need to be further compared and adjusted. Either increasing RN requirement by 0.25 HPRD or increasing total requirement by 1.0 HPRD would bring similar effect that actual total nurse staffing levels would increase by around 0.215 HPRD. And, the increase in actual staffing levels would be mostly done by increase in RN staffing levels. Furthermore, since 1.0 HPRD increase in state staffing standards did not increase 1.0 HPRD in nurse staffing levels, there may be a gap between states expected staffing levels and nursing homes actual staffing levels. Nursing homes would not increase their staffing levels as much as the staffing levels that state staffing standards require. It is possibly because inspection process is quality-oriented but not staffing adequacyoriented. Such inspection may allow nursing homes to have management flexibility to control nurse staffing levels. For nursing homes, complying with state staffing policy may be direct financial burden. Thus, they may focus on staff management until the point that quality standards are not met, instead of increasing nurse staffing levels at or above the state policy. The appendix showed that almost all the state staffing standards are far below those recommended by the CMS (Phase II) report for 2001 and the expert report (CMS, 2001; Harrington et al., 2000). When state staffing standards increase to the recommended levels, the actual nurse staffing levels would increase as much as the impact that the study findings indicated. For organizational factors, chained nursing homes were negatively related to RN staffing levels, but the relationship was not statistically significant. It is possibly because this study used simplistic categorization of chain status (chain-affiliated vs. nonchain-affiliated) for the analysis, and it could not clearly capture the difference of RN staffing levels. A study conducted by Harrington et al. (2012) used detailed categorization by combining chain status with ownership status. They found that nurse staffing levels (including RN staffing levels) in nonprofit chains are much greater than those in both for-profit chains and freestanding facilities. Besides, another study by Mukamel et al. (2012) found that chained nursing homes are negatively related to RN, LPN, and NA staffing levels, when case-mix adjusted staffing levels were used. These

56 Medical Care Research and Review 73(1) methodological points need to be further considered for better understanding of staffing variation between chained and freestanding facilities. For environmental factors, both state Medicaid reimbursement rates and market competition were positively related to nurse staffing levels as expected. However, market demand was negatively related to nurse staffing levels. This may imply that nursing homes staffing decisions could be influenced by perceived market demand, rather than actual market demand. According to Zinn et al. (1998), perceived market factors such as market competition do contribute to nursing homes strategic decision making, while other, presumably objective, indicators such as Herfindahl index do not. Further studies may need to focus more on subjective assessment of market demand (e.g., the manager s perceived market demand and perceived scarcity of potential resources in managerial processes). The study findings are consistent with previous studies, in which stronger RN and total staffing standards are related to higher RN and total staffing levels (Bowblis, 2011; Harrington et al., 2007; Mueller et al., 2006). And, the facility-level measurement used in this study could provide additional understanding of the policy impact. The state-level measurement does not consider facility size as well as duplicated requirements for same category of nurse or different requirements for another category of nurse. Nursing homes which have different sizes are subjected to different requirements under the same state. For instance, in a state, both 24 LN hours and 2.5 total staffing HPRD are required. Assuming that the nursing homes must comply with the stronger requirement, nursing homes with nine residents should have 24 LN hours which is equal to 2.7 HPRD, while nursing homes with greater than nine residents can follow 2.5 total HPRD requirement. When the state-level measurement was applied, the staffing levels required by state standards could be underestimated, which could overestimate the policy impact. The facility-level measurement in the study could offer more precise estimation. In addition, while the previous studies focused mostly on state total staffing standards, this study could provide additional findings of policy impact for other categories of nurse. Nevertheless, since analytical methods vary widely across the studies, systematic investigation of the difference of policy effects across studies needs to be conducted to provide better understanding of the policy impact. Last, several limitations found in the study provide motivation for future study. First of all, like the study by Mueller et al. (2006), this study also found that actual nurse staffing levels varied more within states than between states. Accordingly, it could be concluded that state staffing standards may be one of many factors which influence nurse staffing levels. However, our post hoc analyses to evaluate the relative importance of the predictors indicated that state staffing standards, particularly state RN and LN staffing standards, have stronger contribution to nurse staffing levels than other organizational and environmental factors. Due to the quality-oriented inspection process, the staffing standards could be associated with many other regulations of resident care and safety, which can cause sanctions. As such, nursing homes may perceive that the state staffing policy is minimal but fundamental safety policy that they must comply with for securing their vital

Paek et al. 57 resources. Thus, how differently state staffing policy and organizational factors explain variation in nurse staffing levels needs to be investigated. In addition, how nursing homes actually perceive and react to their state staffing standards would be one of important areas for assessing the policy impact. Second, this study used OCSAR as a main data source. But, information regarding nurse staffing levels has been known to be inaccurately reported in OSCAR (Mueller et al., 2006; N. J. Zhang et al., 2006). Even though this study eliminated the nursing homes with outliers or erroneous values, the study findings may include some undetected errors. Third, states cultural factors, such as the degree of consumer advocacy involvement in long-term care policy making and average nursing staff wage, may be possible confounding factors associated with variation in state staffing standards and nurse staffing levels. Thus, state-by-state investigation together with those states cultural factors would be useful for understanding the policy impact. Fourth, several sensitivity tests of the facility-level measurement were conducted, and consistent coefficients were noticed through correlation, regression, and HLM analyses. Nevertheless, validity of the measurement should be evaluated in future study. This study did not control for potential endogeneity between nurse staffing levels, case-mix, and state Medicaid reimbursement rates, and it might over or underestimate the policy impact. Also, case-mix adjusted nurse staffing levels which were not used in this study would provide another meaningful understanding of how state staffing policy stringency reflects different levels of resident needs. Last, longitudinal analysis, considering variation in length of staffing policy implementation, is encouraged to investigate long-term effects of state staffing standards on nurse staffing levels and quality of care. Appendix State- and Facility-Level Measurements of State Nursing Home Staffing Standards. RN staffing standards LN staffing standards Total staffing standards NA staffing standards States State-level measurement Facility-level measurement a State-level measurement Facility-level measurement State-level measurement b Facility-level measurement State-level measurement Facility-level measurement e AK 0.32 0.62 0.32 0.96 AL 0.08 0.15 0.14 0.34 AR 0.06 0.20 0.56 0.63 2.80 2.88 2.23 AZ 0.06 0.19 0.38 0.44 0.44 CA 0.30 0.22 0.30 0.47 3.20 3.26 2.78 CO 0.24 0.51 0.48 0.50 2.00 2.07 1.53 CT 0.30 0.34 0.70 0.71 1.90 c 1.96 1.25 DC 0.30 0.38 0.57 0.91 3.50 3.83 2.91 DE 0.32 0.30 1.20 1.57 3.67 4.02 2.47 2.47 FL 0.06 0.15 1.00 1.07 3.90 3.95 2.90 2.90 GA 0.06 0.16 0.30 0.41 2.00 2.06 1.64 HI 0.24 0.62 0.24 0.65 IA 0.08 0.20 0.32 0.57 2.00 2.03 1.44 ID 0.30 0.30 0.30 0.65 2.40 2.58 1.91 IL 0.18 0.22 0.58 0.69 2.50 2.59 1.91 (continued)