Estimating Nursing Facility Resiliency for the Next Disaster

ABSTRACT

Objective—To estimate nursing facility (NF) resiliency by complexity and emergency

management (EM) adequacy.

Data Sources/Study Setting—Primary and secondary data from Florida nursing facilities

were collected between November 4, 2014 and January 19, 2015 through surveys and

datamining the 2012 CMS CASPAR and MDS 3.0.

Study Design—quantifiable measurements of NF staff KSAs were collected for a

multivariate analysis of the causal effects of NF complexity and EM plan adequacy upon NF

resiliency. A contingency model demonstrated three quantifiable constructs with correlational

effect that significantly supported the emergency management process in estimating NF

resiliency. Key variables were patient efficacy, workload, size, patient/nurse ratio, and staff

experience.

Data Collection/Extraction Methods—Primary data was collected from a randomized

sample (n=200) of Florida NF administrators (N=680). Surveys of twenty-five (25) questions

were completed by 102 of 200 Florida NFs.

Principal Findings—the prospective study found staff confidence in the adequacy of the

EM plan is significant (.82; p > .01) toward estimating NF resiliency.

Conclusions—it is possible to estimate NF resiliency contingent upon the NF’s

complexity and the adequacy of the EM process in preparing NF staff. However, more detailed

longitudinal studies are needed to quantify EM plan improvements in strengthening future NF

resiliency.

1

Introduction

Healthcare professionals are credited with providing the best possible quality of patient

care during the response and recovery phases of disasters affecting their communities. In 2005,

property losses due to nationally declared disasters were over $115.5 billion dollars and totaled

1,256 disaster related deaths. Six years later the fiancial losses decreased to $23.8 billion with

1,019 souls lost (Hazards & Vulnerability Research Institute, 2012). Disaster research scholars

and healthcare legislators believe contingency planning saves lives and minimizes loss of

property. Therefore, emergency management (EM) plans are required for every healthcare

facility.

This research asked the question: “If emergency management (EM) plans are designed to

save lives and protect property, do EM plans also estimate NF resiliency?”

First, an agreed upon measureable definition was needed for resiliency. The Committee

on Increasing National Resilience to Hazards and Disasters recommends “ a good measure of

resilience…to identify priorities for improvement, determine whether resilience has improved or

worsened, or compare the benefits of resilience with the associated costs” (The National

Academy of Sciences, 2012, p. 10).

Resiliency as defined by the American Psychological Association is “the process of

adapting well in the face of adversity, trauma, tragedy, threats or even significant sources of

threat” (American Psychological Association, 2014, p. 14). For the purposes of this discussion,

the resiliency of a health care organization will be defined as the ability to change from normal

patient care operations to Crisis Standard Care(CSC) during disaster response and recovery

operations (i.e. power outage, tornado, or hurricane). In other words, nursing facility resiliency

can be a measurement of quality of patient care.

2

When changing from normal operations to disaster operations, the quality of patient care

becomes the “ethical basis” for planning the allocation of disaster resources (Institute of

Medicine, 2009, p. 7). Healthcare disaster planners consider CSC as the goal for equitable

quality of patient care when resources are depleted (CDC, 2016, p. 186).

Therefore, the measure of NF resiliency in this research was defined as an estimate of

performance outcomes needed to maintain the predetermined level of CSC. The term NF

resiliency is defined in this study: as the capacity to absorb, cope and recover quality of patient

care from a disruption of normal operations.

An example of normal patient care operations is illustrated in the skilled nursing facility.

Nursing facilities (NFs) are considered among the most complex and therefore most vulnerable

healthcare unit within a system. These facilities face the greatest resiliency challenges because

NF patients are chronically ill with multiple comorbidities (i.e. dialysis, ostomies, and palliative).

Also, the nursing facility is exempted from mandatory evacuations because the frail patients are

already compromised with comorbidities and are extremely susceptible to decompensation,

especially when subjected to sudden changes within their routine care (Bascetta, 2006) (Eiring,

Blake, & Howard, 2012). Therefore, this study used the NF as the unit of analysis because it is

complex and requires more resilience than other organizations.

According to Fielder’s Contingency Theory, an organization’s outcomes are dependent

upon the “least preferred worker” implementation of internal processes and outputs (Ashby,

2007). The NF staff rely upon the internal process of written EM plans as instructions and

authorizations to regain pre-disaster CSC as a resiliency output. NF staff KSAs and the EM

processes correlate to NF resiliency. However, little research has been done to identify empirical

measures of organizational resilience.

3

Thus, for this study contingency theory is the framework for testing the hypothesis: “NF

staff support of EM plan can estimate NF resiliency.”

This study supports the idea that NF resiliency is contingent upon three contextual

elements: 1) the type, extent and duration of the disaster (i.e. tornado F4 damage and three days

without utilities); 2) NF patient complexity (i.e. ADL, acuity and comorbidities); and 3) the NF

staff knowledge, skills and attitudes regarding EM written procedures.

The rationale for using Florida as the context for the study was based upon Florida’s

historical experience with natural disasters and the state’s popularity as a retirement destination.

This provided a cultural mix of skilled nursing facilities and nursing staff exposed to the perils of

hurricanes, wildfires, and related economic shifts.

Florida legislation requires licensed NFs to have EM plans designed with all

contingencies for every type of hazard, including hazardous chemical spills and acts of terrorism

(Florida Statute 59A-4, 2012).

Methodology

The methodology for this study needed a way of capturing the complexity of the NF

patients, staff KSAs, EM plan adequacy and NF capacity for resiliency. So this research study

used three measurement models nested within a modification of the healthcare sector’s standard

for performance measurement.

A covariance structure model or structural equation model (SEM) was modified to

illustrate the causal paths and to capture the correlational relationships between the NF staff’s

confidence in the adequacy of an EM plan as a tool for estimating NF resiliency.

Contingency models traditionally demonstrate causal paths of structure upon internal

processes and subsequent causal paths upon outputs/outcomes (S→P→O). This methodology

4

provided a way to test the research hypothesis: “NF complexity and EM adequacy can be used to

estimate NF resiliency.” Put another way, NF resiliency is contingent upon available resources

and the NF staff’s ability to obtain more through an adequate EM process.

The methodology for this study used three measurement models nested within a

modification of the healthcare sector’s standards for measuring performance. The complexity of

the NF structure (S) correlated with the adequacy of the emergency management (EM) process

(P) in estimating the NF staff’s capacity to maintain CSC as the NF resiliency outcome (O).

This model also allowed for multivariate analysis of large numbers of first-order

exogenous indicators into the second-order latent endogenous indicators for ease of data

management.

Therefore, this research methodology provided the ability to quantify the NF capacity to

absorb, cope and recover quality of patient care from a disruption of normal operations.

Study Validity

Sample size is essential to protecting internal validity of the data. This proposed research

measurement model for the unit of analysis (NF) that is so complex with patient comorbidities

required a robust method for determining a range of sample sizes.

There were thirty-three (33) indicators within the hypothesized composition of three

latent constructs from a population of 680 possibilities. Initially, surveying the entire population

(N=680) seemed impractical and not cost effective. There was also the possibility of a Type I

error of rejecting a true hypothesis (i.e. resiliency is estimated by NF complexity and staff KSA

of EM Plan) (Bickel, 2007).

Therefore, a standard formula of Cronbach α set at > .8, provided more precision in

determining effect significance. An Effect Size Calculator designed by Dr. Daniel Soper for

5

hierarchical multiple regression was chosen to confirm a valid minimum and maximum sample

size (n = 75 < 200; p > .05). This calculator considered multiple latent endogenous constructs

that required individual regression correlation analyses. A sample range of sizes protected the

internal validity of the study with respective levels of accuracy in data collection.

In other words, effect size calculations provided a simulated statistical methodology that

considered three latent constructs: NF complexity, EM plan adequacy, and NF resiliency. These

three latent constructs nested within one final covariance measurement model. Calculations for

sample size were done with proper simulation techniques and power analysis.

The list of assumptions used as threshold parameters are identified and listed here:

Anticipated effect size = .3

Desired statistical power level = .8

Number of latent (unobserved) variables = 19

Number of indicator (observed) variables = 14

Probability or p-value (rule out Type I error) = .05

The recommended optimal sample size was 200 with a minimum sample of 75

participants needed to detect the medium effect (Soper, 2012).

Data Collection

Descriptive data of the nursing facilities in Florida were collected from the Florida

Association of Healthcare Agencies (FAHCA) 2012 database and cross referenced with 2012

CMS data. Descriptive characteristics of the NF and the exogenous variables X(i) were gleaned

and cross checked with other pertinent data into one manageable dataset for regression

correlation analysis. See Table 1 Confirmed Resiliency Variables.

A tracking method was needed to correlate data from multiple reports. So, the CMS

provider number was embedded into the facility’s KSA survey. This CMS provider number was

6

used as a reference point among other reports until all facility related data were collected. Then

the CMS provider number was replaced with an anonymous random case number prior to the

KSA survey distribution.

The University of Central Florida Institutional Review Board approved the KSA survey

instrument and the informed consent document for this research. The survey questions were

adapted from a 2009 nursing facility research grant project conducted by the University of South

Florida for the John Hartford Foundation (Florida Health Care Education and Development

Foundation, 2008, pp. 14-20). Each nursing facility survey was addressed to the administrator

because they are the contact person for local emergency response agencies.

Surveys were distributed only after receiving written consent or approval from each of

277 NF administrators. Then the consenting cases (277 of the 680 contacts) were de-identified of

any personal information that connected the survey participants to their respective facilities.

Then 200 NFs were randomly selected from the de-identified database of 277 pre-screened

participants.

The KSA survey questions were distributed among nursing facility employees and to

minimize social likeability bias, two methods were used to protect the participants. First the

twenty-five (25) survey questions were distributed in a randomized sequence with a 5-point

Likert scale of responses: “none,” “less than enough,” “enough,” “more than enough,” and “do

not know.” Secondly, the KSA survey were de-identified of any personal contact information

and distributed electronically as unique one-time use links through Qualtrics with an “opt out”

option.

7

KSA survey distribution began November 4, 2014. Links to the surveys were closed

January 19, 2015. Thus survey respondents had within ten weeks to reply. Three reminders were

emailed in two week intervals until the study was closed.

It was important that the survey timeframe occurred during the last month of hurricane

season and the last quarter of the NF planning cycle.

The primary KSA survey data were combined as aggregated numbers with the secondary

data collected from the Centers for Medicare and Medicaid (CMS) CASPAR and MDS

databases.

Confirmatory Factor Analyses

A descriptive analysis, ensured that the study sample (n= 200) was a true representation

of the entire NF administrator population (N= 680) within the state of Florida. Consenting

participants came from thirty-one of the sixty-seven counties in Florida. The majority of the 102

completed KSA surveys came from Orange (15.7%) and Pinellas (12.7%) counties.

The first step was to run a confirmatory factor analyses (CFA) using SPSS 22.

Confirmatory factor analysis (CFA) identified the first order elements for each of three latent

constructs: the NF complexity of structure (S), the adequacy of EM process (P), and the expected

NF resiliency outcome (O).

The Confirmatory Factor Analysis (CFA) included a Factor Reduction treatment of the

39 exogenous variables considered for NF complexity(S) into eight (8) easily manageable first-

order factors (Xi) and ten (10) indicators for resiliency (Yi).

These eighteen (18) variables were scientifically significant to the structural equation

model (SEM) of and are described with each data source in Table 1 Confirmed Resiliency

Variables.

8

The Factor Reduction process provided flexibility as regression analyses sorted

collinearity of the proposed variables into three latent endogenous constructs: complexity η1;

adequacy η2; and resiliency η3.

The first-order factors were reduced into second-order components for the latent

construct indicating NF complexity η1. These NF complexity elements were respectively labeled

as: patient acuity ( 1), workload ( 2), and administrative strengths ( See Figure 1 for the

Model for Estimating NF Resiliency.

The following section discusses the results of the CFA of the measurement model for the

latent endogenous construct for the process: EM adequacy (P). The EM plan adequacy

indicators chosen for this study are industry standards based upon the AHCA and CMS criteria

for NF licensure. These healthcare indicators were also used in an earlier research study

conducted by the CDC and the University of South Florida through a grant from the Hartford

Foundation (Eiring, Blake, & Howard, 2012).

Three of the nine proposed indicators for EM plan adequacy (P) did not meet the

predetermined threshold for significance (< .30). Therefore, six indicators remained as first

order factors: alternate facility (Y1), primary communications (Y2), written transportation

agreement (Y3), additional disaster staff (Y4), back-up power (Y5) and medication reserves (Y6).

The next paragraph discusses the factor analysis results of indicators scientifically significant for

estimating NF Resiliency (O).

Confirmatory factor analysis (CFA) of resiliency indicators has never been done for a

skilled nursing facility. However, the CFA of primary data collected from this study’s KSA

survey did provide four quantifiable indicators for NF resiliency. The four strongest factors

9

were: travel caregiver to accompany patient (Y7), knowledge of travel time (Y8), knowledge of

EM resources process (Y9), and authorization to activate the EM resources system (Y10).

Goodness of Fit

Factor loadings were checked for significant effects (direct and cumulative) of indicators

on latent constructs. The larger the loading, then the more effect of that indicator on correlated

constructs. In other words, the larger the effect of NF complexity (S) indicators such as patient

acuity, workload and admin strengths then the more effect is possible on the correlated construct

of EM plan adeqaucy (P) as a process and then a cumulative effect of NF complexity plus EM

plan adequacy (S+P) on the correlated construct of NF resiliency (O) outcomes!

The t-value or Critical Ratio (C.R.) range was larger than 1.96 or less than -1.96. This

ensured that there was a good model fit and that the stastically significant effect remain

between .05 and .09 (Byrne, 2010).

Consideration of the importance of an indicator to the study was determined when

estimating the sample size. The indicator threshold established at .30 meant that any indicators

above .30 and below .90 remained within the final contingency model for measuring and

ultimately for estimating NF resiliency.

Other measures for model reliability were likelihood ratios: CMIN (x2/df), goodness of

fit index (GFI), adjusted goodness of fit index (AGFI), parsimony goodness of fit index (PGFI),

root mean squared (RMS), root mean squared equity adjustment (RMSEA) and Tucker-Lewis

index (TLI). Pearson distribution coefficient was also considered for normal curve distribution as

another indication of goodness of fit (Spatz, 2008; Wan, 2002).

Some analysts consider the Hoelter critical N of .05 as confirmation of a significant

sample size. Hoelter rejected models with more than 200 participants because results can be

10

misconstrued as chi-square moves from nonsignificant to significant in samples greater than 200.

Instead, Hoelter believed sample sizes between 75 and 200 are acceptable confirmation of the

chi-square goodness of fit test (Pallant, 2007). However, some scholars continue to ignore

Hoelter and prefer Cronbach’s α (alpha) to protect the internal validity of data.

Cronbach α ensured the data collected were reliable with the probability of being

repeated in future studies. The acceptable minimum threshold for this study’s Cronbach’s alpha

was set at .8 as recommended in the literature (Bickel, 2007; Pallant, 2007). However, the

findings were closer to 1 and therefore the study more robust than expected before data

collection.

As the last reliability step, the modification index (MI) list of values were used as

recommended adjustments to the model for a parsimonious fit. The MIL identified measurement

error correlations that were too high and were considered redundant among the latent constructs

(Bickel, 2007). Thus the original 39 varaibles became the final 18 varibles in the final sturctural

equation model for estimating NF resiliency within a contingency framework.

Findings

Significant indicators were compared between the FL NF population (N=680) and the NF

sample (n=200). See Table 2 Descriptive Statistics for Population Sample.

These six elements for sample integrity comparisons were: number of dialysis patients

(DIALYSIS); the number of patients with ostomies (OSTOMY); the mean of patients requiring

assistance with daily living (ADLINDEX); the mean of patient acuity (ACUINDEX); the

number of times changes were made in ownership within the previous twelve months

(CHG_OWN); and the size of the NF as determined by the number of certified beds

(CERTBEDS).

11

DIALYSIS findings— Facilities had an average of 2 +/- 2 patients on dialysis.

The highest number in one facility was 15 dialysis patients.

OSTOMY findings—Ostomy patients per facility were higher in numbers with an

average of 5 +/- 7 patients. In 2012, according to the CMS data, one of the Florida

NFs in the sample has the capability to house 42 ostomy patients (CMS, 2013)!

ADLINDEX findings—Assistance with daily living (ADL) Index is 5 units to 12

units per skilled nursing facility. It is based upon the overall patient ADL status as

reported to CMS online through CASPER. In this study the mean ADL Index for

patient severity among the NF population (N=680) was 10.42 +/- 1.21. The NF

study sample (n=200) mean for the ADL Index was 10.47 +/- 1.63 units. ADL

Index findings were skewed highest for NFs providing Hospice care (CMS,

2013).

ACUINDEX findings— The Acuity Index measures patient cognitive abilities

and ranges from 0 to 19 units. Similarly to the ADL Index, the patient acuity

index is based upon the overall patient status of the facility. The study sample

(10.77 +/- 1.8) was a good representation of the population (10.70 +/- 1.3).

OWNERSHIP findings—The 102 survey participants included nineteen (19)

facilities not affiliated with a healthcare system and eighty-three (83) facilities

were affiliated with large corporations. Change in ownership did not occur among

any of the NFs within the twelve months prior to this research.

CERTBEDS—Facility size is customarily represented by the number of beds

certified by Medicaid to accept patients. The average size of the sampled NFs was

116 beds.

12

GENERALIZABILITY—Findings confirmed that the sample (n=200) represented the

research population (N=680). Findings from 102 participant surveys (51% response) were

generalizable and testable for future studies of NF complexity and resiliency. See Table 3 for the

parameter estimates (standardized and unstandardized regression weights) for this study’s

confirmed variables X(i) and Y(i).

Empirical evidence demonstrated the significance of EM plans in estimating resiliency

outcomes. For example, staff confidence in using cell phones and texting as communications

was one of the strongest indicators for plan adequacy with a correlation coefficient of .89 at a

confidence level of .95, p < .01.

The majority of the NH administrators surveyed (.92 +/- .27) required staff to use text

messages through cell phones during disasters. Other EM plan elements displayed significant

relevance in association with resiliency capabilities as follows: Transportation agreement (.499,

p < .01), Additional staff (.573, p < .01), back-up power (.322, p < .01) and medications (.485, p

< .01).

However, the findings may not be generalized to organizations that have fewer

regulations, because the surveyed NFs were isomorphic, and adhered to regimented licensure

criteria. EM plans mimic other EM plans among other NF units within the similar corporate

healthcare systems.

The study findings specific to the exogenous variables for NF complexity of structure (S)

begin in the next paragraph. What follows is a discussion regarding findings related to the other

two latent endogenous constructs: EM plan adequacy (P) and NF resiliency (O).

First consideration was the context of the nursing facility. The context of the nursing

facility included a mixture of patients described as frail, elderly, chronically ill, with disabilities

13

and comorbidities. Three control variables for protecting the homogeneity of the sample were:

license status, percentage of elderly in the county where the NF was located and the urban/rural

service area.

Besides patient acuity and chronic health conditions, some social science scholars believe

cultural barriers can also influence patient and staff dynamics within the context of healthcare

(Azlina & Jamaluddin, 2010) (Drabek & McEntire, 2002). Group diversity affects community

members’ expectations and can either lessen or enhance the organization’s problem-solving

experiences. Issues of trust and embarrassment create additional disaster resiliency problems as

the human phenomenon of not appearing needy and therefore not reaching out for assistance

(Rivera, 2012).

There may have been a cultural barrier between the administrators of St. Rita’s Nursing

Home and local emergency managers when 34 patients died during Hurricane Katrina in 2005

(Dosa, et al., 2010). Another example of a cultural barrier can be observed through differences

in language and personal rationale. The ability to overcome obstacles may be derived from the

aspect of an internal culture through patient expectations and NF staff qualifications. During

times of disaster, the nursing facility patients expect to use the language that they are most

familiar and are comfortable. This is more probable during times of emotional distress. Thus,

communication barriers are a consideration in EM plans. The number of patients using another

primary language are reported each quarter to CMS as Non-English. Therefore, in this study, the

variable NON_ENG was used as a proxy indicator for cultural diversity within FL NFs.

The count used for this study came from the last quarter of the 2012 Minimum Data

Statistics (MDS) database. Facilities in the survey sample with Non-English speaking patients

14

averaged 4 +/- 7 patients. This was significant to organizational complexity because translators,

interpreters, or multi-lingual assistance may be needed to provide CSC quality during a disaster.

Another descriptive statistic was the organization’s dependence upon Medicaid for

revenue. The indicator PCTMCAID, demonstrated the percentage of Medicaid reimbursements

for each facility. Of the 200 facilities in the sample, only 4.9% +/- .22 did not have Medicaid

patients while the other 95.1 % +/- .22 received Medicaid compensation. This may be because

the sample data came from the CMS database. As a socioeconomic indicator, this confirmed the

NF sample is composed disproportionately of Medicaid patients. The patient care population in

the State of Florida is 55.7% (+/- .8) Medicaid.

The percentages of beds occupied in each facility were identified by the variable,

OCCUPANCY. In the NF sample, facilities averaged 87% +/- .15% of available beds filled in

2012. This revenue indicator affected workload levels of staffing hours and required resources.

Physical therapy per patient per day (PTHRD) was extremely small when compared with

other healthcare services. Perhaps due to the endurance levels of frail patients and the frequency

of therapy sessions offered. In other words, the therapist may not be meeting with individual

patients each day, but in groups or as weekly sessions. The CASPER reports for the survey

sample indicated the average PTHRD was .13 +/- .09 hours per day (or 5 to 12 minutes a day).

This variable was ruled out in the correlation regression analysis as less significant and was

eventually removed from the structural equation model. However, PTHRD was included in the

variable for total staff/patient contact hours (TOTHRD).

The dataset also provided the Total Staff hours per patient per day (TOTHRD). The

average among the survey sample was 4.33 +/- .98 hours (or 260 +/-59 minutes). Both

CNAHRD and TOTHRD were assumed to have a strong correlation because CNA’s provide the

15

majority of TOTHRD patient contact hours. Now, that we have discussed the nursing hours, we

will describe the staff ratings found within the quality assurance reports included in 2012

CASPER data.

The NF facility staff (STAFFRATING) are evaluated on a scale of 1 (low) to 5 (high).

All patient care services and administrative staff are included. STAFFRATING of the research

sample averaged 3.45 with a deviation +/- 1.02. The highest percentile rating was 4 and 58.8% of

the sample were rated 4 while 5.9% were rated a 5. So approximately 64% of the sample NF

staff were rated above average (3).

Registered nurse ratings (RNRATING) averaged 2.96 +/- 1.12 with 56% of the sample

ratings at 3 or above. Only 32% of the RN staff were rated below 3. Staff ratings are only one

measure of the organizational complexity within a nursing facility; another complexity measure

is the nurse to patient ratio.

The nurse/patient ratio average was 4.29 +/- .98 or roughly 4.3 hours per day per patient

per licensed professional. This statistic was derived from self-reported times collected from each

certified nurse, licensed nurse and registered nurse per facility.

Delimitations

There was concern regarding social desirability bias affecting the resiliency outcomes.

Since the 1960’s some scholars believe there are two paths of social desirability bias. One is self-

deception and the other path is other-deception. Self-deception occurs whenever the participant

chooses an answer based upon their belief that socially unacceptable characteristics should be

ignored. Therefore, study participants may choose to provide a preferred or morally acceptable

answer. Other-deception bias is evident whenever participants provide inaccurate data to impress

16

an observer. Either type of bias is intended to protect that participant’s self-image. Therefore, it

was imperative that the validity of the results from the KSA survey be protected.

It was important to keep in mind that not all data in this study was cross checked among

other data sources. The self-reported primary data were anonymously linked with other

secondary data, therefore the need for social approval, various response styles, evaluation

apprehension, pre-survey effects, and social desirability bias were minimized as aggregated

results (Azlina & Jamaluddin, 2010; Nederhof, 1985).

Conclusions

In conclusion, the research question: “If emergency management (EM) plans are

designed to save lives and protect property, do EM plans also estimate NF resiliency?” can be

answered “it depends.” This study supported a significant causal path (β = .82, p < .001) toward

NF Resiliency and thus contingent upon the adequacy of a written EM plan. It can further be

deduced that estimating NF resiliency depends upon the facility’s capacity to implement the

Critical Standards of Care (CSC) post-disaster. More detailed comparisons need to be made

between NF and other complex healthcare organizations that require EM Plans.

The NF working environment is so highly regulated that it may be concluded that this

study’s findings were subject to isomorphic organizational behavior and therefore the data were

prone to suffer from “regulatory capture.”

Highly complex organizations such as a skilled nursing facility may present with highly

correlated characteristics and often adopt routines learned from other similar organizations

(Drabek & McEntire, 2002; Salamon, 2002). In this study, regulatory capture could explain why

the bivariate correlations met or exceeded the maximum threshold (.9) for significance among

the first-order factors within the complex NF structure (S). Some of the greater than .9

17

regression correlations in this dataset were not interpreted as a multicollinearity issue because

they were attributed to regulatory capture.

Examples of these bivariate correlations were ostomy patients and patients requiring

dialysis. These licensed services may have construed the increased demand for licensed

practitioners (the regulated workforce) into a negative effect of diminishing returns on

investment (ROI) while simultaneously increasing a positive effect of elevating ratings of

licensed staff. This phenomenon complicated efforts to identify which characteristics between

the facilities were responsible for resiliency, as the EM plan followed a regulated design criteria.

This multivariate non-experimental research incorporated only the most significantly

related indicators extracted from three nested measurement models. It can therefore be

concluded that structural equation modeling (SEM) simplifies traditional analysis perspectives

(Byrne, 2010, p. 17).

It can be concluded that it is possible to create a methodology for estimating NF

resiliency as CSC performance during and after a disaster. However, this is only one

recommended method for estimating resiliency.

Other techniques may originate among the emergent disaster research collaborations

developing across disciplines. Healthcare providers, emergency management practitioners and

public health systems conduct annual disaster training exercises and disaster after action reports

in anticipation of the next disaster. Future implementation studies are needed for more precise

resiliency estimations on three levels: individual, organization and community.

So, there remains a continuing need for future longitudinal studies of ongoing

improvements in the emergency management planning process. This is only one alternative in

estimating a nursing facility’s future resiliency capabilities within a disaster environment.

18

References

American Health Care Association. (2014). Emergency Management Preparedness Guide. Retrieved from https://www.ahcancal.org/facility_operations/disaster_planning/Documents/Emergency%20Preparedness%20Guide_February%202014.pdf

American Psychological Association. (2014). The road to resilience. Washington D.C.: American Psychological Association. Retrieved March 3, 2016, from http://www.apa.org/helpcenter/road-resilience.aspx

Ashby, W. R. (2007). Strategy, Structure, and Performance: The Sociology of Organizational Strategy. In W. R. Scott, G. F. Davis, & L. Jewell (Ed.), Organizations and Organizing Rational, Natural, and Open System Perspectives (pp. 310-339). Upper Saddle River, New Jersey: Pearson Education, Inc. Retrieved 2012

Azlina, A. M., & Jamaluddin, S. (2010). Assessing Reliability of Resiiency Belief Scale (RBS) in the Malaysian Context. International Journal for Cross-Disciplinary Subjects in Education, 3-8. Retrieved 2014

Bascetta, C. (2006). Disaster Preparedness Preliminary Observations on the Evacuation of Vulnerable Populations due to Hurricanes and Other Disasters. Washington DC: U.S. Government Accountability Office.

Bickel, R. (2007). Multilevel Analysis of Applied Research. New York, N.Y., USA: Guilford Press.

Byrne, B. (2010). Structural Equation Modeling with AMOS (Second ed.). New York: Routledge Taylor & Francis Group.

CDC. (2016). Long-Term, Home Health, and Hospice Care Planning Guide for Public Health Emergencies. Pandemic Influenza Preparedness and Response in Long-Term Care Facilities and Their Communities Stakeholder Meeting, April 2010 (p. 296). Centers for Disease Control and Prevention (CDC) Healthcare Preparedness Activity (HPA).

Centers for Medicaid and Medicare. (2013, January 17). Nursing Homes Survey & Certification. Retrieved February 4, 2013, from Certification and Compliance: http:/www.cms.gov/Medicare/Provider-Enrollment-and-Certification/CerficationandCompliance/NHs.html

Dosa, D., Feng, Z., Hyer, K., Brown, L. M., Thomas, K., & Mor, V. (2010). Effects of Hurricane Katrina on Nursing Facility Resident Mortality, Hospitalization, and Functional Decline. Disaster Medicine and Public Health Preparedness, 4, S28-S31.

19

Drabek, T. E., & McEntire, D. A. (2002). Emergent Phenomena and Multiorganizational Coordination in Disasters: Lessons from the Research Literature. International Journal of Mass Emergencies and Disasters, 20(2), 197-224.

Eiring, H., Blake, S., & Howard, D. (2012). Nursing homes' preparedness plans and capabilities. American Journal of Disaster Medicine, 7(2), 127-135.

Florida Health Care Education and Development Foundation. (2008). National Criteria for Evacuation Decision-Making in Nursing Homes. Tampa: University of South Florida.

Hazards & Vulnerability Research Institute. (2012). 1960-2011 U.S. Hazards Losses. Columbia: University of South Carolina, Department of Geography. Retrieved from http://webra.cas.sc.edu/hvri/docs/summary_1960_2011.pdf

Hox, J., & Bechger, T. (2011). An Introduction to Structural Equation Modeling. Family Science Review (11), 354-373.

Institute of Medicine. (2009). Crisis Standardized Care: A systems framework for castastrophic disaster response. Institute for Crisis, Disaster & Risk Management. Washington DC: George Washington University Press.

Nederhof, A. J. (1985). Methods of coping with social desirability bias: a review. European Journal of Social Psychology, 15, 263-280.

Pallant, J. (2007). SPSS Survival Manual. Berkshire, MA: Open University Press McGraw Hill.Rivera, F. (2012). Cultural mechanisms in the exchange of social support among Puerto Ricans

after a natural disaster. Qualitative Health Research, 22(6), 801-809. Retrieved from http://qhr.sagepub.com/content/22/6/801

Soper, D. (2006-2016). Statistical Calculators. Retrieved from Daniel Soper.com: http://danielsoper.com/statcalc3/calc.aspx?id=13

Spatz, C. (2008). Basic Statistics: Tales of Distributions (ninth ed.). Belmont, CA: Thomson Higher Education.

The National Academy of Sciences. (2012). Disaster Resilience: A National Imperative. Committee on Increasing National Resilience to Hazards and Disasters. Washington DC: National Academies Press. Retrieved from http://www,nap.edu/catalog.php?record_id=13457

Wan, T. T. (2002). Evidence-Based Health Care Management: Multivariate Modeling Approaches. Norwell, MA: Kluwer Academic Publishers Group.

20

Table 1 Confirmed Resiliency VariablesSecond Order

First Order Variable

Database Code Factor Description

Data Source

CO

N-

TRO

LS COUNTY% elderly/county

CENSUS

RURAL Rural or Urban

CASPER

NF Complexi

ty

1

Patients Acuity

1

X1 ADLSCORE Patient severity

CASPER

X2 ACUINDEX Patient acuity

CASPER

X3 CNAHRD CNA time with patient

CASPER

Workload2

X4 LPNHRD LPN time with patient

CASPER

X5 OCCUPANCYPercent beds occupied

CASPER

Administrative Strengths

3

X6 STAFFRATING Staff Rating CASPER

X7 RNRATING Nurse Rating

CASPER

X8 NURSERATIO RN per patients

CASPER

Plan Adequacy 2

Y1 ALTFACILITY Evacuation point

KSA Survey

Y2 PRIMCOM Cellphones & text

KSA Survey

Y3 TRANS

Written agreement for disaster transport

KSA Survey

Y4 ADDSTAFFAdditional staff for disaster

KSA Survey

Y5 POWER Back-up for outages

KSA Survey

Y6 MEDSReserves for 3 days/patient

KSA Survey

NH Resiliency 3

Y7 TRAVCAREGIVER

Evacuation staff w/patient

KSA Survey

Y8 TRAVTIME

Know time needed to travel out of harm's way

KSA Survey

Y9 ESSPROC EM Procedures

KSA Survey

Y10 ESSEM Authorization

KSA Survey

21

Table 2 Descriptive Statistics for Population Sample

FL NH Population (N=680)

Randomly Selected Sample FL NH (n=200)

Completed KSA Surveys (n=102)

Min

Max Mean S.D.

Min

Max

Mean S.D.

Min

Max Mean S.D

Dialysis 0 13 2.09 2.26 0 38 6.5 2.33 0 15 2.26 2.33

Ostomy 0 52 5.06 6.69 0 41 5.15 7.14 0 41 5.15 7.14

ADL Index 6 16 10.42 1.21 10 12

10.32 1.63 0 16 10.47 1.63

Acuity Index 6 19 10.70 1.30 10 12

10.81 1.79 0 19 10.77 1.79

Owner Change 0 10 2.03 2.13 1 1 1 0 1 1 1 0

# of Beds 15 462

120.83

49.79 20 300 120

44.52 20 300

116.44

44.42

22

Table 3 Parameter Estimates and Regression Weights for NFSRW URW S.E. C.R. P

Plan_Adequacy .143 .479 .346 1.383 .167NH_Resiliency ( .819 .792 .119 6.650 ***Patient_Acuity(F1) .282 1.879 .775 2.429 .015Workload (F2) 1.051 1.000Admin_Strengths (F3) .802 11.916 2.434 4.895 ***ADLSCORE (X1) .918 1.000ACUINDEX (X2) .673 2.305 .978 2.357 .018CNAHRD (X3) .893 7.630 1.309 5.827 ***LPNHRD (X4) .714 2.691 .516 5.214 ***OCCUPANCY (X5) .504 1.000STAFFRATING (X6) .679 .594 .091 6.525 ***RNRATING (X7) .393 .372 .110 3.385 ***NURSERATIO (X8) 1.189 1.000ALTFACILITY (Y1) .885 1.000PRIMCOM (Y2) 1.000 1.022 .065 15.618 ***TRANS (Y3) .499 .753 .136 5.536 ***ADDSTAFF (Y4) .573 .880 .134 6.593 ***POWER (Y5) .322 .329 .098 3.362 ***MEDS Y6) .222 .117 .051 2.269 .023TRAVCAREGIVER (Y7)

.703 1.000

TRAVTIME (Y8) .594 1.016 .149 6.801 ***ESSPROC (Y9) .441 .789 .210 3.756 ***ESS (Y10) .667 .871 .159 5.485 ***

23

Figure 1 Model for Estimating NF Resiliency

24