Nursing Home Culture Change Practices and Survey Deficiencies: A National Longitudinal Panel Study

Abstract

Background and Objectives

Nursing home (NH) adoption of culture change practices has substantially increased in recent decades. We examined how increasing adoption of culture change practices affected the prevalence of health, severe health, and quality of life (QoL) deficiencies.

Research Design and Methods

Novel data on culture change practice adoption from a nationally representative NH panel (N = 1,585) surveyed in 2009/2010 and 2016/2017 were used to calculate change in practice adoption scores in 3 culture change domains (resident-centered care, staff empowerment, physical environment). These data were linked to data on health, severe health, and QoL deficiencies and facility-level covariates. Multinomial logistic regression models, with survey weights and inverse probability of treatment weighting, examined how increased culture change practice adoption related to change in deficiencies.

Results

We generally observed less increase in deficiencies when culture change practices increased. However, after weighting and controlling for baseline deficiencies and culture change scores, we found few statistically significant effects. Still, results show increased physical environment practices resulted in a higher likelihood of decreases or no change (vs increases) in QoL deficiencies; increased resident-centered care practices resulted in decreases or no change (vs increases) in health deficiencies; and increased staff empowerment practices resulted in a higher likelihood of no change (vs increases) in severe health deficiencies.

Discussion and Implications

This study provides some evidence that culture change practices can help reduce the risk of increasing some types of deficiencies, but the impact of increases in each culture change domain related differently to different types of deficiencies.

Nursing homes (NHs) are critical and dynamic components of the United States health care system and major providers of health services for older Americans. In 2016, 15,600 NHs with 1.7 million licensed beds served more than 1.3 million persons on a given day, including nearly 84% of whom were over aged 65 (Harris-Kojetin et al., 2019). Although the culture of NHs was once recognized as dehumanizing and institutional (Goffman, 1961), in recent decades NHs across the country have adopted a variety of multidimensional models and approaches to deinstitutionalize and humanize their cultures and—in alignment with broader humanist social trends supporting patient empowerment (Lepore et al., 2018; Longtin et al., 2010)—have implemented a wide array of culture change practices (Koren, 2010). Three core practice domains are included in many models and studies of NH culture change: resident-centered care, such as enhancing resident choice and control over daily activities; staff empowerment, such as engaging frontline caregiving staff in management decision making; and physical environment, such as removing nursing stations to make the setting appear more like a residential home rather than a medical institution (Koren, 2010; Miller, Lepore, Lima, Shield, & Tyler, 2014).

The majority of NHs are implementing some culture change practices. In a nationally representative survey of NHs, 88% of NH administrators reported some facility-level involvement in culture change in 2016/2017 (Miller et al., 2018). Propelling the culture change movement, several states effectively incentivize NHs to adopt culture change practices through Medicaid pay-for-performance programs that include culture change measures (Miller et al., 2014, 2018). Additionally, implementation of some culture change practices is motivated by business strategy; for example, some NHs implement physical environment practices to attract specific types of clients (e.g., Medicare or private pay clients for short-stay rehabilitation) (Lepore et al., 2015). Although culture change practices are proliferating and are supported by some state programs, the impacts of culture change on NH quality are not well understood.

Mixed evidence has been found on the impacts of culture change on NH quality, but high levels of culture change implementation have been observed to be associated with better quality outcomes in national and statewide studies. Using data from the first national NH survey of culture change practices, conducted in 2009/2010, Miller and colleagues (2014) examined change in outcomes in years following a NH’s reported culture change adoption, and found that NHs reporting high levels of culture change practice implementation achieved several quality improvements, including decreases in hospitalizations, restraints, tube feeding, and pressure ulcers, and increases in the proportion of residents on bladder training programs. In contrast, NHs that reported lower culture change practice adoption showed some quality improvements, including fewer health and quality of life (QoL) survey deficiencies, but also some quality losses, including increases in hospitalizations and urinary tract infections (Miller et al., 2014). Kansas’ Medicaid pay-for-performance program, which provides staff training and education on person-centered care, has had similarly mixed results: NHs at higher stages of program adoption have shown improvement on most long-stay resident health measures examined, better performance on a composite measure of resident health relative to nonparticipating NHs, and a 49% lower prevalence of major depressive symptoms; but during early stages of adoption, poorer performance was observed on measures of resident falls, independence in performing daily living activities, and incontinence (Hermer et al., 2018). Consistent with the mixed evidence from these national and statewide studies, an early review of the literature on comprehensive culture change models indicated potential psychosocial benefits, but also revealed inconsistent effects on resident physical and psychosocial health outcomes, and a low grade of evidence overall (Hill, 2011). More recent literature reviews have continued to reveal the need for more rigorous research on implementation and outcomes of culture change practices to understand how culture change affects quality (Li & Porock, 2014; Shier, Khodyakov, Cohen, Zimmerman, & Saliba, 2014; Zimmerman et al., 2016). Furthermore, a recent panel study of Veterans Administration NHs found no associations between culture change implementation and quality improvement (Sullivan, Shwartz, Stolzmann, Afable, & Burgess, 2018).

The present study builds on the findings and addresses the limitations of past research in two key ways. First, building on previous national and statewide research showing differences in the impacts of culture change by the extent of practice implementation (Hermer et al., 2018; Miller et al., 2014), we examined impacts of culture change in NHs stratified by their extent of baseline culture change practice implementation. Second, given the substantial differences between resident-centered care, staff empowerment, and physical environment culture change practices, and NH administrators’ different motivations for implementing specific types of culture change practices (Lepore et al., 2015), we examined how each domain of culture change practice separately related to change in the prevalence of deficiencies—specifically, health, severe health, and QoL deficiencies—from state surveyor inspections.

Conceptual Framework

In developing the culture change surveys used in this study, we used an ecological framework, consistent with the Holistic Approach to Transformational Change (HATCh) model (Quality Partners of Rhode Island, 2005). We conceptualized resident-centered care, staff empowerment, and physical environment domains of culture change as being related to survey deficiencies and other quality outcomes (Miller et al., 2018). We adopted a difference-in-differences design to evaluate whether decreases or no change (vs increases) in the prevalence of deficiencies were significantly more likely when NHs had greater increases in culture change practice adoption.

Given the findings from prior research examining impacts of culture change on quality outcomes and the nature of practices in each culture change domain, we hypothesized that increasing resident-centered care and staff empowerment practices would have beneficial impacts on health, severe health, and QoL deficiencies (i.e., an increase in resident-centered care and staff empowerment practices would be associated with either a decrease or with no increase—vs an increase—in the prevalence of health, severe health, and QoL deficiencies), and physical environment practices would have beneficial impacts on QoL deficiencies but not health or severe health deficiencies (i.e., an increase in physical environment culture change practices would be associated with a decrease or with no increase—vs an increase—in the prevalence of QoL deficiencies, but would not be associated with change over time in the prevalence of health deficiencies or severe health deficiencies).

Study Data and Methods

This study merges novel survey data on culture change practice implementation from a nationally representative panel of NHs with NH-level data from the Centers for Medicare and Medicaid Services (CMS)’ Certification and Survey Provider Enhanced Reporting (CASPER) database, and NH resident-level clinical assessment data from the Minimum Data Set (MDS). The national panel study of NHs consisted of two survey rounds. The 2009/2010 baseline information was collected as part of a large program project covering several topics related to NH quality. Separate surveys were sent to Directors of Nursing (DONs) and NH administrators from 4,035 eligible U.S. NHs (i.e., Medicare- and Medicaid-certified NHs with at least 30 beds that had not taken part in developing the survey). The 2016/2017 follow-up survey (i.e., T2 survey) focused solely on culture change and was funded by a separate federal grant. Eligible NHs included all those that had returned baseline surveys from NH administrators and were still in operation (n = 2,142). A response rate of 74% (n = 1,584) was obtained and no response bias was detected for either wave. Weights are used to account for the overall stratified, proportionate random sample design.

The CASPER database from CMS includes data on NH characteristics, such as size and ownership, and information from NH inspections which are conducted by the state approximately every 12–15 months (Harrington & Carillo, 1999). State inspectors assign deficiencies by severity and scope to indicate poor quality (Antonova & Zimmerman, 2012). Resident-level MDS data are federally mandated assessments carried out at admission, discharge, and regular intervals during a resident’s stay for each NH resident in a Medicare-/Medicaid-certified NH in the United States.

Culture Change Scores

While the T2 survey more comprehensively measured culture change practices than the baseline survey, it repeated the survey items that collected information at baseline across three domains—physical environment (six items), staff empowerment (seven items), and resident-centered care (four items). Physical environment survey items assessed the extent to which facilities have homelike features, such as the proportion of residents with private rooms, whether kitchen areas are accessible to residents and families 24/7, whether play areas for children are available, and the presence of residents’ personal items in common areas. Staff empowerment survey items assessed how often staff work together to cover shifts when someone cannot come to work, how common it is for staff to be cross-trained to perform tasks outside of their assigned job duties, how often nursing assistants take part in quality improvement teams, whether nursing assistants are able to choose which residents they care for, and other issues. Items pertaining to the resident-centered care index focused on the extent to which residents have choices related to their activities and care, including when they get up in the morning, when they have meals, and which nursing assistants care for them.

At baseline and T2, all staff empowerment and physical environment items were answered by NH administrators. Resident-centered care items were answered by DONs at baseline and by NH administrators at T2, accounting for a smaller sample size for this domain because only 76% of eligible study facilities had DON surveys at baseline (Lima, Schwartz, Clark, & Miller (2020) report sample details, survey items, and point allocations). The change in survey design—from completion by DON to administrator—was made because budget constraints allowed for only one survey per facility in 2016/2017 (Miller et al., 2018) and cognitive interviews conducted with NH DONs and administrators before fielding the T2 survey indicated that administrators were better informed about culture change practices (Shield, Tyler, Berridge, Clark, & Miller, 2018). Additionally, nonresponses to culture change survey items resulted in missing data that reduced the numbers of NHs included in the final models. Among the 1,584 study facilities with surveys at both time points, 1,459 (92.1%) NHs had sufficient information to calculate physical environment scores at both time points, and 1,447 (91.4%) and 1,105 (69.8%) NHs had sufficiently complete data to calculate staff empowerment and resident-centered care scores, respectively. The prevalence of missing data ranged from 1.8% to 9.9% of items. Items were considered missing at random as no theoretical reason suggests missingness relates to culture change practices. Item responses were imputed within NHs using the NH’s domain-specific average based on the nonmissing items, but only when a facility had two or fewer missing items in the physical environment (1.2% in 2009/2010, 1.4% in 2016/2017) and staff empowerment (3.0% and 5.8%, respectively) domains or one missing item in the resident-centered care domain (0.3% and 2.2%, respectively). Missing scores were not imputed for one physical environment item whose response was not a Likert scale rating like the other items in the domain. Additional details about the survey methodology are reported elsewhere (Clark et al., 2018; Lima et al., 2020; Miller et al., 2018).

NH change in each of the three culture change domain scores at baseline was subtracted from T2 scores to determine whether, and to what extent, NHs had increased implementation of practices in the domain during the study period. Because we were interested in assessing how increasing adoption of culture change practices affected the changing prevalence of survey deficiencies, our explanatory variables were conceptualized in terms of “increase” in culture change practices. NHs were categorized as having an “increase” or “no increase” of practices in each culture change domain, based on the difference in domain scores between their baseline and T2 surveys. NHs with a two-point or greater increase in their resident-centered care domain score from baseline to T2 were categorized as having an “increase” in the domain, and NHs with a three-point or greater increase in their staff empowerment or physical environment domain scores were categorized as having an “increase” in those domains, and all other NHs were categorized as having “no increase” per domain (Lima et al., 2020).

Quality Outcome Variables

The three outcome variables in this study were NH change from baseline to T2 in the prevalence of health, severe health, and QoL deficiencies. Consistent with other research examining how culture change affects deficiencies (Grabowski et al., 2014), we used CASPER to identify the count of each of the three types of survey deficiency (health, severe health, and QoL) in each study NH at baseline and T2. When surveyors find deficiencies that pose potential or immediate danger to residents, they classify them with a letter rating (A through L), with deficiencies coded by letters closer to the beginning of the alphabet (i.e., A through F) indicating “no actual harm” and deficiencies coded by later letters in the alphabet (i.e., G through L) indicating “actual harm” or “immediate jeopardy” (Office of the Inspector General, 2019). We used A through F deficiencies to identify the prevalence of health and QoL deficiencies and used G through L deficiencies to identify the prevalence of severe health deficiencies. No G through L QoL deficiencies were observed. Deficiencies were weighted for severity and scope (Grabowski et al., 2014) and standardized by state to account for interstate variations in survey results (Kelly, Liebig, & Edwards, 2008). NH changes from baseline to T2 in health, severe health, and QoL survey deficiencies were calculated. Three categories of change in deficiencies were used in statistical analyses: more deficiencies (two or more deficiencies at T2 than baseline), no change (a difference of one deficiency at most between baseline and T2), and fewer deficiencies (two or more deficiencies at baseline than T2). Multivariate analyses used more deficiencies as the reference group and generated results that indicate the relative risk or likelihood that a NH which increased a given culture change domain score had no change or fewer deficiencies as opposed to more deficiencies.

Propensity Score Analysis

Propensity score analysis was used to minimize bias from the imbalance of NH and resident characteristics between NHs that adopted more culture change practices over time and NHs that did not increase adoption. To balance the characteristics, we calculated propensity scores that represented each NH’s probability of increasing adoption of each type of culture change practice, based on observed characteristics. Inverse probability treatment weights, equal to the inverse of each NH’s probability of increasing practice uptake in each culture change domain, were incorporated into multinomial analyses to minimize the effects of observed confounding (Austin, 2016).

Propensity scores were tested before use to estimate the effects of increased culture change practices on changes in the prevalence of deficiencies. Final propensity scores were calculated using 15 NH-level covariates (Table 1)—including NH, staff, and resident characteristics—that prior research found to be associated with culture change implementation or with survey deficiencies. Data on NH and staffing characteristics mainly came from CASPER (Grabowski et al., 2014; Harrington, Olney, Carillo, & Kang, 2012; Harrington, Zimmerman, Karon, Robinson, & Beutel, 2000), and from our survey we included data on administrator turnover. Data on resident characteristics came from CASPER or were aggregated from MDS at the time of the baseline survey, including: payor source (Chisholm et al., 2018; Miller et al., 2014), race and ethnicity (Fennell, Feng, Clark & Mor, 2010; Grabowski, 2004), and cognitive impairment (Lepore & Leland, 2015).

Table 1.

Variables in Propensity Scores

Variable	Operationalization	Data source
NH characteristics
NH profit status	For profit or nonprofit	CASPER
NH size	Total number of beds	CASPER
NH chain membership	Part of a chain or not	CASPER
NH hospital-based status	Hospital-based or freestanding	CASPER
NH special care unit status	Have a special care unit or not	CASPER
NH occupancy rate	Percent of beds occupied	CASPER
Staffing characteristics
Registered nurse (RN) staffing	RN hours per resident day (HRPD)	CASPER
Licensed professional nurse (LPN) staffing	LPN HPRD	CASPER
Certified nursing assistant (CNA) staffing	CNA HPRD	CASPER
Administrator turnover	Whether or not there was a change in administrator between the T1 and T2 surveys	Survey
Resident characteristics
Medicaid payor source	Percentage Medicaid residents	CASPER
Medicare payor source	Percentage Medicare residents	CASPER
Race	Percentage black residents	MDS
Ethnicity	Percentage Hispanic residents	MDS
Cognitive impairment	Percentage of residents with a high cognitive performance score (CPS) indicating a high level of cognitive impairment	MDS

Note: NH = nursing home; CASPER = Certification and Survey Provider Enhanced Reporting; MDS = Minimum Data Set.

Importantly, because propensity scores can only consider data for observed characteristics, to control somewhat for unmeasured confounders we created our propensity scores separately for facilities with culture change baseline scores below the median or at or above the median. We did this because we expect that the degree of baseline adoption is correlated with unmeasured confounders and that facilities with similar patterns of culture change adoption at baseline will be more likely to also have similar unmeasured confounders affecting change in adoption of culture change practices.

Statistical Analysis

A multinomial logistic regression model of the likelihood of a NH having fewer or no change in state-standardized health-, severe health-, and QoL-related survey deficiencies was estimated for each culture change practice– outcome combination (across the three culture change practice domains and three deficiency outcomes) stratified by baseline culture change domain scores (below median score vs at and above median score). Regression models incorporated the inverse probability treatment weighting described above and T2 survey weights (allowing us to generalize findings nationally). Analyses were conducted using Stata 15.

Brown University’s Institutional Review Board reviewed and approved the study and included the appropriate Waiver of Informed Consent and HIPAA Waiver of Authorization for the use of person-level claims, enrollment, and assessment data. The use of CMS data was covered by a Data Use Agreement. Supplemental information about the data and methods used for these analyses can be found in Brown’s Digital Repository (https://repository.library.brown.edu/studio/item/bdr:847097/).

Results

Table 2 shows how the prevalence of deficiencies changed between baseline and T2 among NHs with and without increases in culture change domain scores. Results are stratified by whether a NH’s baseline score was at or above, or below, the median of all study NHs’ baseline scores. Descriptive findings for most strata revealed that a lower proportion of NHs with increases in culture change domain scores (vs those without increases) had increases in the prevalence of deficiencies. For example, among NHs with baseline resident-centered care domain scores equal to or above the median (n = 693), approximately 28% had increases in the number of health deficiencies compared to 44% of NHs with no increase in the resident-centered care scores. Also, for these same NHs, 10% of those with increased resident-centered care scores had increases in the number of QoL deficiencies while this proportion was 19% for NHs without score increases (Table 2). However, some increases did not follow this pattern. For example, among NHs with staff empowerment domain scores less than the median at baseline (n = 579), about 46% that increased staff empowerment domain scores had an increase in health deficiencies, whereas fewer than 43% of NHs that did not increase staff empowerment domain scores had an increase in health deficiencies (Table 2).

Table 2.

Pre–Post Changes in Health, Severe Health, and Quality of Life Survey Deficiencies by Whether NHs Increased Their Culture Change Domain Scores, Stratified by Baseline Culture Change Domain Scores

Domain	Baseline domain score (n)	Score increase 0 = no 1 = yes	Change in health deficiencies			Change in severe health deficiencies			Change in quality of life deficiencies
			Increase	No change	Decrease	Increase	No change	Decrease	Increase	No change	Decrease
Resident-centered care	≥ Median (693)	0	43.56	19.85	36.59	13.89	70.27	15.83	18.76	66.38	14.85
		1	27.67	31.90	40.43	13.28	69.32	17.40	10.17	73.32	16.51
	< Median (392)	0	48.65	17.62	33.72	17.74	65.39	16.87	24.29	60.63	15.08
		1	45.22	14.38	40.40	15.52	66.67	17.81	16.36	67.86	15.78
Staff empowerment	≥ Median (843)	0	42.22	19.33	38.45	13.03	69.39	17.58	16.51	68.61	14.87
		1	37.33	21.18	41.49	13.66	73.73	12.60	19.12	67.64	13.23
	< Median (579)	0	42.58	17.12	40.30	18.84	63.32	17.84	21.80	61.80	16.40
		1	45.99	15.74	38.27	13.04	68.15	18.81	17.41	64.55	18.05
Physical environment	≥ Median (760)	0	42.70	18.32	39.98	13.76	69.44	16.81	16.54	68.35	15.10
		1	41.35	22.80	35.85	13.26	74.15	12.59	11.74	73.23	15.03
	< Median (676)	0	44.47	16.97	38.56	14.40	67.53	18.08	21.13	63.65	15.23
		1	40.09	19.73	40.19	18.43	63.23	18.34	19.28	63.04	17.68

Note: NH = nursing home.

Increases in Culture Change Scores and Changes in Health, Severe Health, and QoL Deficiencies Over Time—Multivariate Analyses

Multivariate models controlling for baseline survey deficiencies and culture change domain scores, and including survey weights and with inverse probability of treatment weighting, show how change (from baseline to T2) in culture change domain scores is related to whether health, severe health, and QoL deficiencies decreased or did not change (vs increased) between baseline and T2 using relative risk ratios (RRR). Among NHs with baseline resident-centered care scores equal to or above the median, the relative risk of having no change versus an increase in health deficiencies was well over two times greater in NHs that increased resident-centered care scores (RRR 2.82; 95% confidence interval [CI] 1.46, 5.41). Additionally, increased resident-centered care scores from baseline to T2 were marginally associated with NHs having fewer versus more health deficiencies over time (RRR 1.83; 95% CI 0.91, 3.70) (Table 3). Resident-centered care scores were not related to change in the prevalence of severe health or QoL deficiencies.

Table 3.

Logistic Regression of Increased Resident-Centered Care Practices on Change in Health Deficiencies, Severe Health Deficiencies, and Quality of Life Deficiencies, With Inverse Probability Weighting * Survey Weights, Stratified by Baseline Resident-Centered Care Practice Score (≥ Median or < Median)

				RRR	p > |z|	95% CI
		Change in deficiencies	More deficiencies	Reference
Health deficiencies	≥ Median	No change	Baseline health deficiencies	1.19	.014	1.04	1.37
			Baseline resident-centered care score	0.53	.110	0.24	1.16
			Increase resident-centered care score	2.82	.002	1.46	5.41
		Fewer	Baseline health deficiencies	1.61	.000	1.39	1.88
			Baseline resident-centered care score	1.27	.554	0.57	2.89
			Increase resident-centered care score	1.83	.090	0.91	3.70
	< Median	No change	Baseline health deficiencies	1.16	.016	1.03	1.32
			Baseline resident-centered care score	1.07	.744	0.72	1.59
			Increase resident-centered care score	1.00	.994	0.50	1.99
		Fewer	Baseline health deficiencies	1.46	.000	1.30	1.64
			Baseline resident-centered care score	0.95	.738	0.72	1.26
			Increase resident-centered care score	1.39	.278	0.77	2.54
Severe health deficiencies	≥ Median	No change	Baseline severe health deficiencies	2.01	.116	0.84	4.81
			Baseline resident-centered care score	0.90	.826	0.36	2.26
			Increase resident-centered care score	1.00	.911	0.46	2.15
		Fewer	Baseline severe health deficiencies	31.46	.000	10.70	92.46
			Baseline resident-centered care score	1.87	.381	0.46	7.57
			Increase resident-centered care score	0.89	.830	0.30	2.63
	< Median	No change	Baseline severe health deficiencies	1.90	.168	0.76	4.67
			Baseline resident-centered care score	1.02	.906	0.73	1.43
			Increase resident-centered care score	1.12	.723	0.40	2.84
		Fewer	Baseline severe health deficiencies	11.75	.000	3.95	35.00
			Baseline resident-centered care score	1.11	.676	0.68	1.80
			Increase resident-centered care score	1.07	.899	0.40	2.84
Quality of life deficiencies	≥ Median	No change	Baseline quality of life deficiencies	2.36	.140	0.75	7.38
			Baseline resident-centered care score	1.05	.906	0.44	2.55
			Increase resident-centered care score	1.80	.138	0.83	3.94
		Fewer	Baseline quality of life deficiencies	31.46	.000	10.70	92.46
			Baseline resident-centered care score	1.00	.996	0.23	4.35
			Increase resident-centered care score	2.03	.260	0.59	7.00
	< Median	No change	Baseline quality of life deficiencies	2.94	.002	1.51	5.75
			Baseline resident-centered care score	0.94	.656	0.70	1.25
			Increase resident-centered care score	1.65	.105	0.90	3.03
		Fewer	Baseline quality of life deficiencies	30.84	.000	12.09	78.67
			Baseline resident-centered care score	0.88	.622	0.52	1.48
			Increase resident-centered care score	1.75	.525	0.05	4.42

Note: Bold values indicate statistically significant (p < 0.05) relationships between increased culture change scores and change in the prevalence of survey deficiencies.

Among NHs with baseline staff empowerment scores below the median, the relative risk of having no change versus an increase in severe health deficiencies was significantly greater in NHs that increased empowerment scores (RRR 1.74; 95% CI 1.04, 2.91). Additionally, among NHs with baseline staff empowerment scores equal to or above the median, increased staff empowerment scores were marginally associated with NHs having no change versus an increase in health deficiencies (RRR 1.73; 95% CI 0.94, 3.02). Likewise, among NHs with below median QoL deficiencies, the relative risk of decreasing versus increasing QoL deficiencies was marginally greater in NHs that increased staff empowerment scores (RRR 1.87; 95% CI 0.89, 3.95) (Table 4).

Table 4.

Logistic Regression of Increased Staff Empowerment Practices on Change in Health Deficiencies, Severe Health Deficiencies, and Quality of Life Deficiencies, With Inverse Probability Weighting * Survey Weights, Stratified by Baseline Staff Empowerment Practice Score (≥ Median or < Median)

				RRR	p > |z|	95% CI
		Change in deficiencies	More deficiencies	Reference
Health deficiencies	≥ Median	No change	Baseline health deficiencies	1.23	.000	1.11	1.37
			Baseline staff empowerment score	1.18	.012	1.04	1.35
			Increase staff empowerment score	1.73	.077	0.94	3.02
		Fewer	Baseline health deficiencies	1.68	.000	1.50	1.88
			Baseline staff empowerment score	1.06	.351	0.94	1.19
			Increase staff empowerment score	1.33	.320	0.77	2.35
	< Median	No change	Baseline health deficiencies	1.13	.003	1.04	1.30
			Baseline staff empowerment score	1.13	.185	0.95	1.34
			Increase staff empowerment score	0.95	.846	0.58	1.56
		Fewer	Baseline health deficiencies	1.58	.000	1.45	1.71
			Baseline staff empowerment score	1.02	.787	0.86	1.21
			Increase staff empowerment score	0.85	.521	0.53	1.38
Severe health deficiencies	≥ Median	No change	Baseline severe health deficiencies	1.37	.261	0.79	2.37
			Baseline staff empowerment score	1.05	.446	0.92	1.21
			Increase staff empowerment score	1.09	.798	0.57	2.07
		Fewer	Baseline severe health deficiencies	6.69	.000	3.28	13.62
			Baseline staff empowerment score	0.95	.600	0.78	1.15
			Increase staff empowerment score	0.68	.423	0.27	1.74
	< Median	No change	Baseline severe health deficiencies	2.28	.046	1.02	5.12
			Baseline staff empowerment score	1.13	.175	0.85	1.34
			Increase staff empowerment score	1.74	.037	1.04	2.91
		Fewer	Baseline severe health deficiencies	17.38	.000	6.05	49.88
			Baseline staff empowerment score	1.00	.989	0.76	1.31
			Increase staff empowerment score	1.62	.233	0.73	3.57
Quality of life deficiencies	≥ Median	No change	Baseline quality of life deficiencies	5.21	.000	2.88	9.45
			Baseline staff empowerment score	1.00	.996	0.87	1.15
			Increase staff empowerment score	0.88	.703	0.47	1.67
		Fewer	Baseline quality of life deficiencies	97.39	.000	30.57	310.26
			Baseline staff empowerment score	0.99	.897	0.81	1.20
			Increase staff empowerment score	0.75	.566	0.28	2.00
	< Median	No change	Baseline quality of life deficiencies	2.28	.001	1.39	3.75
			Baseline staff empowerment score	1.10	.203	0.95	1.27
			Increase staff empowerment score	1.48	.106	0.92	2.37
		Fewer	Baseline quality of life deficiencies	24.93	.000	12.61	49.29
			Baseline staff empowerment score	1.22	.121	0.95	1.57
			Increase staff empowerment score	1.87	.099	0.89	3.95

Note: Bold values indicate statistically significant (p < 0.05) relationships between increased culture change scores and change in the prevalence of survey deficiencies.

Finally, among NHs with baseline physical environment scores equal to or greater than the median, the relative risk of having a decrease versus an increase in QoL deficiencies was over six times greater in NHs that increased physical environment scores (RRR 6.14; 95% CI 1.64, 23.02), and among NHs with baseline physical environment scores below the median, the relative risk of having no change versus an increase in QoL deficiencies was over three and a half times greater in NHs that increased physical environment scores (RRR 3.64; 95% CI 1.98, 6.70). Physical environment culture change scores were not related to change in health or severe health deficiencies (Table 5).

Table 5.

Logistic Regression of Increased Physical Environment Practices on Change in Health Deficiencies, Severe Health Deficiencies, and Quality of Life Deficiencies, With Inverse Probability Weighting * Survey Weights, Stratified by Baseline Physical Environment Practice Score (≥ Median or < Median)

				RRR	p > |z|	95% CI
		Change in deficiencies	More deficiencies	Reference
Health deficiencies	≥ Median	No change	Baseline health deficiencies	1.13	.043	1.00	1.29
			Baseline physical environment score	1.07	.593	0.83	1.38
			Increase physical environment score	1.40	.401	0.64	3.08
		Fewer	Baseline health deficiencies	1.48	.000	1.31	1.68
			Baseline physical environment score	0.95	.669	0.77	1.19
			Increase physical environment score	1.20	.586	0.63	2.28
	< Median	No change	Baseline health deficiencies	1.21	.00	1.11	1.32
			Baseline physical environment score	1.00	1.00	0.84	1.19
			Increase physical environment score	1.21	.45	0.74	1.98
		Fewer	Baseline health deficiencies	1.49	.00	1.36	1.62
			Baseline physical environment score	0.98	.836	0.83	1.16
			Increase physical environment score	1.05	.823	0.67	1.64
Severe health deficiencies	≥ Median	No change	Baseline severe health deficiencies	2.89	.008	1.31	6.36
			Baseline physical environment score	0.84	.153	0.66	1.07
			Increase physical environment score	1.02	.968	0.45	2.28
		Fewer	Baseline severe health deficiencies	15.57	.000	5.51	44.04
			Baseline physical environment score	0.83	.27	0.60	1.15
			Increase physical environment score	0.91	.82	0.30	2.75
	< Median	No change	Baseline severe health deficiencies	2.23	.021	1.13	4.40
			Baseline physical environment score	1.91	.904	0.84	1.22
			Increase physical environment score	0.77	.295	0.47	1.26
		Fewer	Baseline severe health deficiencies	18.95	.000	8.00	44.90
			Baseline physical environment score	0.74	.047	0.54	1.00
			Increase physical environment score	0.88	.721	0.42	1.82
Quality of life deficiencies	≥ Median	No change	Baseline quality of life deficiencies	2.67	.005	1.35	5.28
			Baseline physical environment score	1.01	.931	0.82	1.24
			Increase physical environment score	1.72	.213	0.73	4.02
		Fewer	Baseline quality of life deficiencies	74.35	.000	29.58	186.83
			Baseline physical environment score	1.06	.742	0.748	1.50
			Increase physical environment score	6.14	.007	1.64	23.02
	< Median	No change	Baseline quality of life deficiencies	2.11	.003	1.28	3.49
			Baseline physical environment score	1.01	.858	0.87	1.18
			Increase physical environment score	3.64	.000	1.98	6.70
		Fewer	Baseline quality of life deficiencies	26.96	.000	12.45	58.36
			Baseline physical environment score	1.13	.406	0.85	1.51
			Increase physical environment score	1.49	.32	0.67	3.30

Note: Bold values indicate statistically significant (p < 0.05) relationships between increased culture change scores and change in the prevalence of survey deficiencies.

As noted above, multivariate analyses controlled for NHs’ baseline culture change domain scores and baseline deficiencies. Baseline culture change scores were a significant predictor of change in deficiencies in 89% of total stratified results, whereas baseline deficiencies were a significant predictor of change in deficiencies in just 5.5% of results.

Sensitivity Analysis Using Survey Weights (Without Inverse Probability Weighting)

In addition to the results reported above, which were derived from analytic models that used both inverse probability of treatment weighting and survey weights, we also ran the models using only survey weights. Differences of significance between the models were limited mainly to the QoL models. For example, in the model examining staff empowerment impacts on QoL deficiencies, when using survey weights (without inverse probability treatment weighting), increased staff empowerment scores were significantly associated with the likelihood of having no change versus an increase in QoL deficiencies (RRR 1.07; 95% CI 1.01, 1.15; not reported in table).

Discussion and Implications

This study shows increased adoption of practices in three core domains of culture change—resident-centered care, staff empowerment, and physical environment—is not consistently associated with NH quality, as indicated by no change or a decrease (vs an increase) in a NH’s prevalence of health, severe health, and QoL survey deficiencies. However, when increasing culture change practice adoption was associated with a higher likelihood of no change or a decrease in deficiencies (compared to increased deficiencies) we found the categories of deficiencies affected differed by the culture change domain being evaluated. For example, we found significant relationships between increases in resident-centered care practices and health deficiencies; between staff empowerment practices and severe health deficiencies; and between physical environment practices and QoL deficiencies. By measuring implementation of these culture change domains longitudinally within NHs and controlling for the selection bias associated with baseline domain scores, this study advances our understanding of whether and how culture change practices (and what domains of practices) may be able to advance differing aspects of quality, measured here by survey deficiencies.

Study findings indicate that increasing resident-centered care practices, among NHs with median or higher-than-median baseline levels of practice implementation, protects against increasing health deficiencies. Likewise, increasing staff empowerment practices, among NHs with lower-than-median baseline levels of practice implementation, protects against increasing severe health deficiencies. Additionally, increasing physical environment practices, among NHs with lower-than-median baseline levels of practice implementation, protects against increasing QoL deficiencies, and similarly, increasing physical environment culture change practices, among NHs with median or higher-than-median baseline levels of physical environment practice implementation, increases the likelihood of decreasing rather than increasing QoL deficiencies.

Results provide some but not overwhelming support for our hypotheses. While we anticipated increased resident-centered care, staff empowerment, and physical environment practices would be associated, respectively, with the observed beneficial impacts on the frequency of health, severe health, and QoL deficiencies, we hypothesized additional favorable relationships between culture change practices and deficiencies that were not observed. For example, we also expected to observe statistically significant relationships between resident-centered care practices and severe health and QoL deficiencies, and between staff empowerment practices and health and QoL deficiencies.

The association between increasing resident-centered care scores and having no change in health deficiencies was only observed among NHs with baseline resident-centered care scores greater than or equal to the median. This finding reflects prior national and statewide studies that show positive outcomes among NHs with high levels of culture change practice implementation, but mixed outcomes at lower implementation levels (Hermer et al., 2018; Miller et al., 2014).

Findings suggest that increasing staff empowerment practices may help NHs protect against increasing severe health deficiencies, and this aligns with other studies that emphasize the value of staff empowerment for quality outcomes. For example, a large-scale survey of DONs linked with CMS outcomes data showed that NHs with DONs who engage in staff empowerment practices achieve better care outcomes, including fewer survey deficiencies (McKinney, Corazzini, Anderson, Sloane, & Castle, 2016). To the extent that Kansas’s PEAK education and training program is considered to be grounded in staff empowerment, Hermer and colleagues’ (2018) findings of improvements in several outcomes among more advance program participants also lend support to the importance of staff empowerment for NHs achieving quality outcomes. The value of staff empowerment for supporting quality outcomes is an important issue for NHs, researchers, and policymakers to consider in relation to implementing quality improvement programs and worthy of further longitudinal study.

Findings also suggest that increasing physical environment culture change practices protects against increasing QoL survey deficiencies. These findings are consistent with prior research linking the establishment of homelike NH environments to psychosocial benefits (Hermer, Bryant, Pucciarello, Mlynarczyk, & Zhong, 2017) and showing that residents of Green House model NHs, compared to residents of control facilities, reported better QoL in several domains (Kane, Lum, Cutler, Degenholtz, & Yu, 2007; Yoon, Brown, Bowers, Sharkey, & Horn, 2015).

Several marginal statistically significant associations (p ≤ .10) together with the other significant findings noted support the importance of further study of culture change practices in relation to survey deficiencies. In particular, given the statistically significant association observed between increases in resident-centered care practice scores and no change in health deficiencies among NHs with baseline scores at or above the median, two marginally significant associations strengthen the importance of further study of these practices. These include the finding that increases in resident-care practices were associated with decreases in severe health deficiencies among NHs with baseline resident-centered care scores at or above the median (RRR 1.83, 95% CI 0.91, 3.70), and with no change in severe health deficiencies among NHs with baseline scores below the median (RRR 1.65, 95% CI 0.90, 3.03) (Table 3). In addition, considering also observed marginally significant associations, future study of staff empowerment practices and both health and QoL deficiencies appears warranted. Increases in staff empowerment practices were associated with no change in health deficiencies among NHs with baseline scores at or above the median (RRR 1.73, 95% CI 0.94, 3.02); and, among NHs with baseline scores below the median, increases in staff empowerment practice scores were associated with both no change in QoL deficiencies (RRR 1.48, 95% CI 0.92, 2.37), and decreases in QoL deficiencies (RRR 1.87, 95% CI 0.89, 3.95).

Implications for NH Policy and Regulation

This study has important implications for NH policy and regulation pertaining to culture change, particularly pay-for-performance programs. Many states have adopted NH pay-for-performance programs that use financial rewards to incentivize NH practices, and states with pay-for-performance programs that use culture change measures effectively promote NH adoption of culture change practices (Miller et al., 2018). Furthermore, in states that measure the degree of culture change implementation in NHs as part of their Medicaid pay-for-performance programs, evidence is mixed but shows some beneficial clinical outcomes of increased implementation of culture change practices (Hermer et al, 2018; Miller et al., 2018). While our findings are not consistent across all outcomes of interest, the findings that are significant show that specific domains of culture change practices affect health, severe health, and QoL survey deficiencies differently. Therefore, when designing Medicaid pay-for-performance programs, states may find it valuable to incentivize domains of culture change practices, rather than global indicators of culture change. Such measurement of specific culture change practice domains could enhance the validity of culture change measurement and strengthen the capacity of the state and providers to focus measurement on key practices (e.g., culture change practices associated with high priority outcomes). For NHs and states aiming to stem increasing health or severe health survey deficiencies, our findings indicate that adoption of resident-centered care or staff empowerment culture change practices has better odds of success, as hypothesized, compared to physical environment culture change practices. However, physical environment practices appear to be an important component for achieving higher QoL for residents.

Although this study was strengthened by its use of linked longitudinal culture change and survey deficiency data from a nationally representative panel of NHs, important limitations are noted. First, our longitudinal data on culture change practices were limited to the three core domains of culture change and did not include more peripheral domains in the HATCh model (e.g., family and community and leadership) (Quality Partners of Rhode Island, 2005). Second, because the measurement of each culture change domain at baseline was restricted to a low number of survey items (Miller et al., 2014), we were only able to measure change for those survey items, rather than for a more comprehensive list of practices. The resident-centered care domain included just four items, and baseline scores were so high that a quarter of NHs were excluded from the resident-centered care change analyses because by our definition they were unable to further increase scores in this domain. However, all three of the indexes used had good measurement properties, including construct validity (Miller et al., 2014). Repeating this survey using indexes with more items would be more likely to show statistically significant rather than the observed marginal associations, and thus the relationships between these culture change practices and survey deficiencies remain worthy of further study. Third, social desirability bias may be reflected in NH administrator responses to culture change survey items. Although this possibility cannot be discounted, previous research indicates that NH administrators show less bias than DONs when responding to culture change survey questions (Shield, Tyler, Berridge, Clark, & Miller, 2018). Fourth, inadequate power may have accounted for some of nonstatistically significant findings observed, especially in relation to the resident-centered care domain. This domain had a lower starting number and a large proportion of NHs (>30%) were removed since improvement was not possible (in the only four-item domain). Finally, responses to resident-centered care survey items were provided by DONs at baseline, and by NH administrators at follow-up. As a result, the resident-centered care analyses included a reduced sample size and may have introduced some measurement error. For example, if DONs are more subject to social desirability bias, then baseline scores may have been inflated compared to follow-up scores, thus reducing the observed domain score change. Future research attending to more culture change domains and practices is needed to more fully evaluate the associations studied here, and thus to inform NH policy and practice.

Conclusion

NH culture change is spreading, with the majority—and an increasing proportion—of NHs adopting culture change practices (Miller et al., 2018), but research has tended to view NH culture change as a composite black box concept rather than examining implementation and outcomes of specific culture change practices. This study opens the black box of NH culture change and shows the differential impacts of three core domains of culture change practices—resident-centered care, staff empowerment, and physical environment practices—on health, severe health, and QoL survey deficiencies. Study findings indicate that increased NH adoption of resident-centered care, staff empowerment, and physical environment practices, as measured in our national survey, related to change in the prevalence of health, severe health, and QoL deficiencies in different ways. This and similar knowledge is needed to help guide NH policy and practice by promoting targeted quality outcomes through incentivizing adoption of specific domains of culture change practice.

Funding

This work was supported by the National Institute on Aging (grant number NIA R01 AG048940-01A1). J. C. Lima received funding from the NIA during the conduct of this study. J. C. Lima also received funding from NIA outside of the submitted work in 2016. S. C. Miller received funding from the NIA during the conduct of this study.

Conflict of Interest

None reported.