A

su

CaTha Unb Dec Cold Un

International Journal of Nursing Studies 52 (2015) 1735–1745

A R

Artic

Rece

Rece

Acce

Keyw

Earl

Inte

Job

Job

Long

Orga

Pers

Sho

Wor

*

1

180

http

002

structural equation model of turnover for a longitudinalrvey among early career registered nurses

rol S. Brewer a,*, Ying-Yu Chao a,1, Craig R. Colder b, Christine T. Kovner c,omas P. Chacko d

iversity at Buffalo School of Nursing, 210 Wende Hall, 3435 Main Street, Buffalo, NY 14214, United States

partment of Psychology, Park Hall, 227, University at Buffalo, NY 14260-4110, United States

lege of Nursing, New York University, 726 Broadway, 10th Floor, New York, NY 10003, United States

iversity at Buffalo School of Social Work, 205 Parker Hall, 3435 Main Street, United States

T I C L E I N F O

le history:

ived 17 February 2015

ived in revised form 22 June 2015

pted 27 June 2015

ords:

y career nurses

nt to stay

satisfaction

search

itudinal studies

nizational commitment

onnel turnover

ck

k environment

A B S T R A C T

Background: Key predictors of early career nurses’ turnover are job satisfaction,

organizational commitment, job search, intent to stay, and shock (back injuries) based

on the literature review and our previous research. Existing research has often omitted one

of these key predictors.

Objectives: The purpose of this study in a sample of early career nurses was to compare

predictors of turnover to nurses’ actual turnover at two time points in their careers.

Design: A multi-state longitudinal panel survey of early career nurses was used to compare

a turnover model across two time periods. The sample has been surveyed five times.

Participants: The sample was selected using a two-stage sample of registered nurses

nested in 51 metropolitan areas and nine non-metropolitan, rural areas in 34 states and

the District of Columbia.

Methods: The associations between key predictors of turnover were tested using

structural equation modeling and data from the earliest and latest panels in our study.

We used predictors from the respondents who replied to the Wave-1 survey in 2006 and

their turnover status from Wave 2 in 2007 (N = 2386). We compared these results to

the remaining respondents’ predictors from Wave 4 in 2011 and their turnover status

in Wave 5 in 2013 (N = 1073). We tested and found no effect for missingness from Wave

1–5 and little evidence of attrition bias.

Results: Strong support was found for the relationships hypothesized among job

satisfaction, organizational commitment, intent to stay, and turnover, with some support

for shock and search in the Wave 1–2 sample. However, for Wave 4–5 sample (n = 1073),

none of the paths through search were significant, nor was the path from shock to turnover.

Conclusions: Nurses in the second analysis who had matured longer in their career did

not have a significant response to search or shock (back injuries), which may indicate

how easily experienced registered nurses find new jobs and/or accommodation to

jobs requiring significant physicality. Nurse turnover is a major concern for healthcare

organizations because of its costs and related outcomes. The relevant strength and

relationships of these key turnover predictors will be informative to employers for

Corresponding author. Tel.: +1 716 713 7625; fax: +1 716 829 2067.

E-mail address: csbrewer@buffalo.edu (C.S. Brewer).

Current address: Rutgers, the State University of New Jersey School of Nursing, Ackerson Hall, Room 360, 180 University Avenue, Newark, NJ 07102-

3, United States.

Contents lists available at ScienceDirect

International Journal of Nursing Studies

journal homepage: www.elsevier.com/ijns

://dx.doi.org/10.1016/j.ijnurstu.2015.06.017

0-7489/� 2015 Elsevier Ltd. All rights reserved.

C.S. Brewer et al. / International Journal of Nursing Studies 52 (2015) 1735–17451736

What is already known about the topic?

� Turnover is a continual concern for organizationsbecause of its costs and impact that it has on patientoutcomes; however, nothing is known about whatimpacts turnover in the same sample over time.� Four key concepts that are consistently identified in the

literature predicting turnover are intent to stay or leave,search, job satisfaction, and organizational commitment,but much of the research omits at least one or more ofthese.� Additional research has identified shocks as initiating

one of four possible pathways to turnover.

What this paper adds

� This study identifies the significant pathways amongfour key variables (intent, search, job satisfaction, andorganizational commitment, and incorporates a measureof shock in the comparison of the model at two timeperiods in the early careers of RNs.� Using longitudinal data this study shows a causal

relationship between predictor variables and turnoverat both time periods. In the earlier time period search andshock are both significant, but they are not significant inthe later time period.� Our research suggest that when controlling for con-

founding factors, attrition, and missing data, we canconclude that change among the sample over time mayexplain the lack of importance of search in the modelof intent to stay and turnover.

Turnover is a continual concern for organizationsbecause of its costs and impact that it has on patientoutcomes (Jones and Gates, 2007; Li and Jones, 2013), andmany factors have been proposed to explain it. Twosubstantive integrated reviews of turnover literature(Gilmartin, 2012; Griffeth et al., 2005) resulted in tworemarkably consistent models with a set of four variablesforming the core of both reviews. These variables are jobsatisfaction, organizational commitment, job search(search), and intent to stay (or leave; intent). In addition,the meta-analysis of Griffeth et al. (2000) supported thatjob satisfaction, organizational commitment, search, with-drawal cognitions, and quit intentions were the bestpredictors of turnover. Withdrawal cognitions are a groupof constructs that mediate intent and turnover. Lee andMitchell (1994) include these constructs in a voluntaryturnover model but also posit that shocks play a significantrole (Lee et al., 1999; Morrell et al., 2004).

Generally researchers examine a cross section of anorganization’s workforce at one point in time; at best,turnover may be examined one time period later. We foundno literature that uses a panel survey at different points in

time to determine whether the turnover predictor relation-ships vary over time. One problem in evaluating literatureabout turnover is that much of the research is based onunderspecified models in which only a subset of thesevariables theoretically related to turnover are included. Forexample, Coomber and Barriball (2007) reviewed literatureexamining the relationship of job satisfaction to intent tostay (or leave) and turnover. Some of the research studiesincluded organizational commitment (e.g., Gurney et al.,1997; Simon et al., 2010 and some did not (e.g., Galletta et al.,2011; Meeusen et al., 2011). Studies that include all five keyvariables posited as significant in reviews are less commonthan those that include a subset (Blau, 2007; Brewer et al.,2012). Thus, a direct relationship of satisfaction to intent isproposed when in fact that relationship may be indirect(mediated) through organizational commitment. The sameissue exists for turnover.

Another issue is that many researchers (Beecroft et al.,2008; Simon et al., 2010) examine intent rather thanturnover. While there is a moderate relationship betweenintent and turnover (Brewer et al., 2012; Griffeth et al.2000) using a longitudinal data set is a stronger methodto show a causal relationship than cross-sectional studies(Estryn-Behar et al., 2010), but more difficult methodo-logically to accomplish. The purposes of this study areto determine the pathways among all five key variablesusing longitudinal data, and to compare the model attwo time periods for differences in a group of newnurses compared to the nurses at a later point in theircareers.

1. Background

The search procedure captured articles addressingturnover in primarily populations of registered nurses,practical nurses, or other health professionals (i.e.,physicians). To identify potentially relevant studies thatwere published in English from 1981 to January 2014, weconducted searches using CINAHL, MEDLINE, EconLit, Webof Knowledge, and IngentaConnect databases along withmanual searches of the reference lists of the articlesretrieved. The search was limited to quantitative or meta-analytic empirical studies, systematic reviews, and at leasttwo of the three major constructs that have been used inturnover research to predict either intent or turnover, jobsatisfaction, organizational commitment, and search.Studies were included in this review if they providedevidence of level 1 through level 3 based on Evaluation

standards of management research (Reay et al., 2009): Level1 includes randomized controlled trials or meta-analyses;Level 2 includes a high quality review or a systematicliterature review; and Level 3 includes large sample,multisite quantitative studies.

prioritizing strategies to retain their registered nurse workforce. We need more research

on programs that implement changes in the work environment that impact these two

outcomes, as well as research that focuses on the relevant strength or impact to help

administrators prioritize translation of results.

� 2015 Elsevier Ltd. All rights reserved.

revandbetamhadAm200et aSmandsatitur200stushoConnotdracomroletur

as ajobet aet aGurIng199et aKimincto

and(19orgis norgjob

et

com(intmitrevIn tor btha3 omeeviof

a d

1.1.

thaMujob

C.S. Brewer et al. / International Journal of Nursing Studies 52 (2015) 1735–1745 1737

Twenty-two empirical studies were identified in ouriew (see Supplementary Data Table A1). We compared

synthesized studies that investigated the relationshipsween turnover and its four proximal antecedents, andong the antecedents themselves, especially those that

turnover and/or intent as the dependent variable.ong these studies, eight studies (Allen et al., 2003; Blau,7; Brewer et al., 2012; Camerino et al., 2008; Felpsl., 2009; Griffeth et al., 2000; Mueller and Price, 1990;

ith et al., 2011) had turnover as a dependent variable also included the major proximal variables (jobsfaction, organizational commitment, and intent) to

nover. Of these, only two also included search (Blau,7; Brewer et al., 2012). Search was not significant in the

dy of Brewer and colleagues. The study of Blau (2007)wed that intent mediated search’s effect on turnover.trariwise, Hom and Griffeth (1991) included search but

organizational commitment, and posited that with-wal cognitions and the utility of withdrawal (the mostparable concepts to intent) occur before search. The

of search and its measurement is thus uncertain innover models.The following studies have intent to stay or leave a job

dependent variable and also both the major variables satisfaction and organizational commitment (Beecroftl., 2008; Brewer et al., 2012; Cox et al., 2010; De Gieterl., 2011; Garbee and Killacky, 2008; Griffeth et al., 2005;ney et al., 1997; Halbesleben and Wheeler, 2008;

ersoll et al., 2002; Kim and Hwang, 2011; Kim et al.,6; Kovner et al., 2009; Mueller and Price, 1990; Simonl., 2010; Wang et al., 2012). Of these 15 studies, only

et al. (1996) and Hom and Griffeth (1991) alsoluded search, which was significant and antecedentintent. However, Swider et al. (2011) excludes intent

uses search to predict turnover, whereas Somers96) uses search to predict turnover, but omitsanizational commitment and intent: job satisfactionot significant. Conversely, Castle (2006) uses intent andanizational commitment to predict turnover, but omits

satisfaction.Another model of turnover (Brewer et al., 2012; Kovneral., 2009) was developed based on initial workpleted by Price (2001) on the four key concepts

ent, search, job satisfaction, and organizational com-ment) that are consistently identified in the literatureiews on turnover (Gilmartin, 2012; Griffeth et al., 2005).his review, these constructs had a direct relationshipoth a direct and indirect relationships with turnover

t was supported by at least two studies rated as levelr higher on Reay’s evidence-based scale for manage-nt literature (Reay et al., 2009). In summary, there isdence to support both a direct and indirect relationshipssearch (through intent) to turnover, and evidence ofirect relationship of intent to turnover.

Job satisfaction

Job satisfaction is defined as the affective orientationt an employee has toward his or her work (Price andeller, 1981; Price, 2001). There are two approaches to

satisfaction. One approach is to use a global measure of

general satisfaction, such as Minnesota SatisfactionQuestionnaire (Weiss et al., 1967) and global job satisfac-tion (Hackman and Oldham, 1975). The second approach isto use a faceted measurement, such as the Work QualityIndex (Whitley and Putzier, 1994) and the Nurse WorkIndex-Revised (Aiken and Patrician, 2000). For ourpurposes, we use global measures such as used by Breweret al. (2012) and Quinn and Staines (1979). There issubstantial evidence that job satisfaction directly predictsintent to stay and job turnover, but studies often excludeorganizational commitment or intent when examiningsatisfaction’s relationship to turnover (Lee et al., 2008;Murrells et al., 2008; Price and Mueller, 1981; Somers,1996). We propose that the job satisfaction’s relationshipto turnover is indirect through organizational commit-ment, search and or intent.

1.1.1. Job satisfaction and organizational commitment

Studies are inconsistent regarding the causal relation-ship between job satisfaction and organizational commit-ment. Most researchers conclude job satisfaction predictsorganizational commitment (Gaertner, 1999; Gurney et al.,1997; Kim et al., 1996; Mueller and Price, 1990) whichimpacts turnover through intent and/or search. On thecontrary, some researchers report that organizationalcommitment is a direct predictor of job satisfaction (Luet al., 2007; Vandenberg and Lance, 1992).

1.1.2. Impact of job satisfaction on search

Griffeth et al. (2005) described that job satisfaction hada significant direct negative effect on intent to searchamong faculty and staff in a major state university in thesouthwestern United States. However, Kim et al. (1996)and Kovner et al. (2009) found indirect effects of jobsatisfaction on intent through search.

1.1.3. Impact of job satisfaction on intent

Mobley (1982) proposed that the relationship betweenjob satisfaction and turnover is moderated by intent. Thisproposition is supported by several studies (e.g., Allenet al., 2003; Hom and Griffeth, 1991; Murrells et al., 2008;Price and Mueller, 1981). Some researchers establishedthat when both organizational commitment and jobsatisfaction are included in the model, only organizationalcommitment was a significant predictor of intent to stay(Garbee and Killacky, 2008; Ingersoll et al., 2002; Muellerand Price, 1990), and others found that both are significant(De Gieter et al., 2011; Griffeth et al., 2005; Kovner et al.,2009; Wang et al., 2012). Cox et al. (2010) reportedcontradictory results, as only job satisfaction was asignificant predictor on nurses’ intent to stay in the U.S.Navy reserve when job satisfaction and commitment wereboth included. All of these studies omitted search andturnover.

1.1.4. Impact of job satisfaction on turnover

A vast body of literature links nurses’ job satisfactionwith their turnover (Coomber and Barriball, 2007; Hayeset al., 2006). Some researchers report that job satisfactionhas significant direct effects on turnover. For example,Brewer et al. (2012) found that overall job satisfaction was

C.S. Brewer et al. / International Journal of Nursing Studies 52 (2015) 1735–17451738

a direct negative significant predictor of turnover; whileBlau (2007) reported that a faceted job satisfactionmeasure (including satisfaction with co-workers, personalgrowth options, and supervision) was a direct negativesignificant predictor of turnover. On the other hand, someresearchers reported that job satisfaction had significantindirect effects on turnover through intent to stay (Breweret al., 2012; Allen et al., 2003; Price and Mueller, 1981).

To our knowledge, there have been no longitudinalstudies to sort out the direction of effects for job satisfactionand organizational commitment in predicting turnover,although most researchers conclude that job satisfactionpredicts organizational commitment. For the purposes ofthis study, therefore, we make the following hypotheses:

Hypothesis 1a. Job satisfaction directly and positively pre-dicts organizational commitment.

Hypothesis 1b. Job satisfaction directly and negativelypredicts job search.

Hypothesis 1c. Job satisfaction directly and positively pre-dicts intent to stay.

Hypothesis 1d. Job satisfaction directly and negativelypredicts turnover.

1.2. Organizational commitment

Organizational commitment is a central feature ofmodels of intent to stay or turnover in nursing and otherprofessions (Meyer et al., 2002; Mowday et al., 1979;Wagner, 2007). We use the definition of organizationalcommitment as ‘‘loyalty of employees to their employers’’(Kim et al., 1996; Price, 2001).

1.2.1. Impact of organizational commitment on search

Researchers have reported that organizational commit-ment was a significantly negative predictor of searchamong physicians (Kim et al., 1996). However, organiza-tional commitment was not a significant predictor onsearch in hospital nurses (Kovner et al., 2009), or amonguniversity faculty and staff (Griffeth et al., 2005).

1.2.2. Impact of organizational commitment on intent

There is some evidence that organizational commit-ment is a significant predictor of nurses’ intent to stay inhospital jobs (De Gieter et al., 2011; Kovner et al., 2009;Mueller and Price, 1990; Wang et al., 2012) and faculty tostay in nursing education jobs (Garbee and Killacky, 2008;Gurney et al., 1997). Allen et al. (2003) established that therelationship between organizational commitment andturnover was mediated by turnover intent. Particularly,organizational commitment had greater impact on intentthan satisfaction. Castle (2006) used logistic regressionto show a (direct) relationship between organizationalcommitment and intent; however, no indirect relationshipwas tested. Halbesleben and Wheeler (2008), in contrast,showed that affective commitment and job satisfaction didnot have a significant impact on turnover intent. Kim et al.(1996) using Price’s early model (Mueller and Price, 1990)

comprehensively described in Price’s later papers (Price,2001, 2004) found in a random sample of U.S militaryphysicians that organizational commitment was a signifi-cant positive predictor of intent to stay, but job satisfactionwas not. Search behavior was also significant. Kim andHwang (2011) showed that affective commitment was adirect significant positive predictor of intent to stay amongquality improvement nurses; however, search was omittedand job satisfaction was not significant.

1.2.3. Impact of organizational commitment on turnover

There is some empirical evidence that organizationalcommitment directly predicts turnover. Griffeth andcolleagues (2000) explicated that organizational commit-ment predicts turnover more strongly than does overalljob satisfaction in a meta-analysis of antecedents andcorrelates of employee turnover. Castle (2006) also foundthat organizational commitment had a significant negativedirect effect on voluntary turnover as well as search andintent, but a mediating effect was not tested, and jobsatisfaction was omitted. Thus, there is evidence to supporteach of the following hypotheses:

Hypothesis 2a. Organizational commitment directly andnegatively predicts search.

Hypothesis 2b. Organizational commitment directly andpositively predicts intent to stay.

Hypothesis 2c. Organizational commitment directly andnegatively predicts turnover.

1.3. Search

Job search is defined as the degree to which an employeeis looking for another job(s) (Kim et al., 1996; Price, 2001).Job search, job search attitude, job search intention, jobsearch behavior and search behavior are terms used inthe literature to describe activities, attitude toward andbehaviors by which job seekers explore alternative jobopportunities (Griffeth et al., 2000; Kim et al., 1996). Searchitself can be considered either an attitude (Kim et al., 1996;Price, 2001) or set of activities (Griffeth et al., 2000).

1.3.1. Impact of search on intent

The relationship between job search behavior and thevarious dimensions of job, career or professional intentionsis not well established (Brewer et al., 2009). Kim et al.(1996) and Kovner et al. (2009) found that search wassignificant, when job satisfaction, organizational commit-ment, and job search were regressed on intent to stay. Insummary, there is strong evidence that intent directlypredicts turnover, but whether search is mediated by intentor is related directly to turnover is not clear, and in additionmaybe be affected by how or when search is measured.

1.3.2. Impact of search on turnover

There are mixed findings about the impact of search onturnover. The meta-analysis by Griffeth et al. (2000)established that search intention was the only significantturnover predictor among various measures of job search

behjobrelaconet aothSwturall

indBreturtionfoudenon

Howthejobwesign199madepshoadvintewitthis

Hyp

Hyp

1.4.

mediffito s201job(Moet astaythewitHowthadiretheconmaet aMo

direhyp

Hyprela

C.S. Brewer et al. / International Journal of Nursing Studies 52 (2015) 1735–1745 1739

aviors (search intentions, general job search scales, search behaviors, and job search methods). Thetionships between turnover, intent, and the variousstructs of job search have not been widely studied. Kiml. (1996) places search distal to intent as a predictor, buters (Blau, 2007; Hom and Griffeth, 1991; Somers, 1996;ider et al., 2011) have search directly predictingnover (mediating intent). Among five studies that usedfive variables, Blau (2007) found that job search was anependent positive predictor of turnover. However,wer et al. (2012) reported that the effect of intent onnover was mediated by job satisfaction and organiza-al commitment, but not search. Griffeth et al. (2000)

nd both job search and intent significantly indepen-tly predicted turnover. These studies have not focuseda thorough examination of mediators and moderators.

ever, Felps et al. (2009), illustrated that job search was only significant predictor of individual turnover when

satisfaction, organizational commitment, and searchre included. Several authors have found that search was a

ificant direct predictor of turnover (Blau, 2007; Somers,6; Swider et al., 2011). Another reason study findings

y be mixed is that a successful job search may be heavilyendent on the job market at a given time. In a nurse laborrtage, finding a job may require little if any search inance; a tighter job market may require more search. Thent to leave may occur concurrently with job search,

hout clear causality. Because of the mixed support for construct, we do not propose a direct relationship.

othesis 3a. Job search is not related to intent to stay.

othesis 3b. Job search is not related to turnover.

Intent

Measures of intent are varied and the variation inasures and slight differences in concepts make itcult to assess their role (s) in turnover theory. Intent

tay (Kim et al., 1996) and desire to stay (Kirschling et al.,1) indicate employees’ willingness to stay in their

s, professions or organizations. Intention to leaverBarak et al., 2001) and turnover intention (Luml., 1998) are comparable but are the reverse of intent to. Hom and Griffeth (1991) proposed an expansion of

concept of intent into several related concepts calledhdrawal cognitions and expected utility of withdrawal.

ever, the common thread for all of these concepts ist they are generally proposed as the most immediatect precursor to turnover. Most researchers now accept

premise that intent to stay or leave and similarcepts compose the final cognitive steps in the decisionking process of voluntary turnover (Blau, 2007; Brewerl., 2012; Camerino et al., 2008; Griffeth et al., 2000;

bley, 1982; Mueller and Price, 1990).All these studies support that intent has consistent andct effect on turnover. Thus we posit the followingothesis:

othesis 4a. Intent to stay is directly and inverselyted to turnover.

1.5. Shocks

One less tested development in turnover theory hasbeen the concept of shocks. Lee and Mitchell (1994) havedeveloped the unfolding model of voluntary turnover, inwhich the pathways described unfold along one of fourpathways depending on whether shocks are present or not.Shock is defined as a particular, jarring event that initiatesthoughts of quitting a job (Lee et al., 1999). Shock triggersfour pathways of turnover including: (1) leaving withoutconsidering current attachment to the organizations andalternatives; (2) reconsidering organizational attachment;(3) leaving after evaluating current jobs and searchingalternatives; and (4) leaving due to lower levels of jobsatisfaction instead of shock (Lee et al., 1999). Our datasetallows us to test this first pathway, if we assume that theshock proxy (injuries including back sprains and strains) isin fact a shock as suggested by (Brewer et al., 2012). Hencewe postulate the following:

Hypothesis 5a. Work related injury (shock) directly pre-dicts turnover.

2. Methods

2.1. Study design

We used a longitudinal panel design to test a modellinking major turnover variables (job satisfaction, organi-zational commitment, search, intent, shock, and turnover)among a national sample of early career registered nurses(Kovner et al., 2007). We compare the results from theearliest surveys in our panel to those of the latest surveysto accentuate any potential differences in the careertrajectories. The surveys were conducted one year apart forthe first two waves, and every two years after that; at thispoint we have completed five surveys over seven years in2006, 2007, 2009, 2011, and 2013.

2.2. Participants

Data were obtained from surveys mailed and emailed toearly career registered nurses (RNs). The first survey wasmailed to those RNs who passed the National CouncilLicensure Examination (NCLEX) between September2004 and August 2005. The sample was selected using atwo-stage sample of RNs nested in 51 metropolitan areas(MSA) and nine non-MSA rural areas in 34 states and theDistrict of Columbia. The sampling method and theeligibility criteria were described in detail elsewhere(Kovner et al., 2007). For each wave of the survey, we usedthe Dillman survey method with a $5 incentive (Dillman,2007). The sample sizes were: Wave 1 (2006; N = 3370),Wave 2 (2007; N = 2386), Wave 3 (2009; N = 2007), Wave 4(2011; N = 1544), and Wave-5 (2013; N = 1073). For testingthe SEM model, we used the respondents who replied toboth Wave-1 and Wave-2 and Wave-4 and Wave-5. Wave1 and Wave 2 were one year apart, but Wave 4 and Wave5 were two years apart. We compared the nurses whoresponded in Wave 1 to Wave 5 and found only onesignificant difference on non-local job opportunities

C.S. Brewer et al. / International Journal of Nursing Studies 52 (2015) 1735–17451740

(p < .01), thus little evidence of attrition bias. Wave5 respondents had found a more difficult non-local jobopportunities environment.

2.3. Variables and measures

2.3.1. Major variables

Job satisfaction (Quinn and Staines, 1979), organiza-tional commitment, search, and intent to stay (Price, 2001)were measured by scales described in Table 1. Jobsatisfaction was defined as ‘‘a worker’s general affectivereaction to the job without reference to any specificjob facet’’ (Quinn and Staines, 1979; p. 205); e.g., ‘‘all in all,how satisfied would you say you are with your job?’’ Thesecond variable, organizational commitment was definedas ‘‘loyalty of employees to their employers,’’ (Price, 2001;p. 608); e.g., ‘‘my present employer inspires the very best inme in the way of job performance.’’ Price (2001) definedintent to stay as ‘‘the extent to which employees plan tocontinue membership with their employers,’’ (p. 608); e.g.,‘I plan to stay with my employer as long as possible.’Finally, search behavior was defined as ‘‘the degree towhich employees are looking for other jobs,’’ (Price, 2001;p. 608); e.g., ‘I almost always follow up on job leads.’

We defined shock for this study as the rate of strains,sprains, including back injury per year for each respondent.Price’s original theory and our original study was notdesigned to measure shock, so we used this measure as aproxy. These variables were chosen from Wave 1. Turnoverwas calculated using RN-reported employer changesbetween Wave 1 and Wave 2, and then Wave 4 and Wave5 datasets.

2.3.2. Control variables

We identified variables that were significantly corre-lated with turnover in our previous study (Kovner et al.,2009) and supported by the literature that might beconsidered potential confounders of our hypothesized

associations, and added Wave 1 variables that weresignificantly correlated with Wave 2 turnover as statisticalcontrol variables in our path models. The followingdemographic and work attributes were included asstatistical control variables: gender (Estryn-Behar et al.,2010; Nooney et al., 2010), affectivity (Chen et al., 2008;Price, 2001), overtime (Paquet et al., 2013), direct care(Harrington and Swan, 2003; Kash et al., 2007), unit(Staggs and Dunton, 2012), types of health institution(Camerino et al., 2008; Josephson et al., 2008), andperceived availability of alternative job opportunities(Josephson et al., 2008; Kovner et al., 2009), and had leftat least one job prior to our first data collection (Suzukiet al., 2008, 2010).

3. Data analysis

SPSS (version 21) was used to compute descriptivestatistics. Pearson correlation and Chi-square were usedto test the associations between variables and turnover.Mplus (version 7.11) was used to estimate path models.Our dependent variable (turnover) was dichotomous, andtherefore, we used weighted least squares estimation forthe path models. While there is some skewness in theshock variable, Weighted Least Squares (WLS) estimationyields asymptotically normal estimates regardless of thedistribution in the population (Bollen, 1989; Muthenet al., 1997). Indirect effects were tested using boot-strapped confidence intervals computed using 5000 boot-strapped samples (MacKinnon, 2008). We reportstandardized coefficients to allow comparison of differ-ent scale metrics. Several criteria were used to evaluatemodel fit, including the model chi-square (x2), Compar-ative Fit Index (CFI), and Root Mean Square Error ofApproximation (RMSEA). A CFI � 0.90 and RMSEA < .06suggest a good fit (Hu and Bentler, 1999; Kline, 2005). Wealso used several methods to control for missing dataand sample attrition.

Table 1

Major variables and time of data collection.

Variable Possible range Wave N Mean � SD Reliability

(Cronbach’s a)

Turnover Yes/no

0 = no change

1 = change employer

W2 2386 0 = 1910

1 = 476

NA

W5 1073 0 = 872

1 = 201

NA

Shock 1 = totally unexpected

7 = totally expected

Item range: 1–5

W1 2197 1.03 � 2.70 NA

W4 984 0.98 � 1.63 NA

Job satisfaction (Quinn and Staines, 1979) 1 = very dissatisfied

7 = very satisfied

Item range: 1–7

W1 2326 5.21 � 1.54 0.829

W4 1009 5.31 � 1.46 0.843

Organizational commitment (Price, 2001) 1 = strongly disagree

5 = strongly agree

Item range: 1–5

W1 2323 3.79 � 0.78 0.862

W4 1003 3.74 � 0.77 0.859

Search behavior (Price, 2001) 1 = strongly disagree

5 = strongly agree

Item range: 1–5

W1 2378 2.83 � 0.44 0.765

W4 1039 2.81 � 0.46 0.813

Intent to stay (Price, 2001) 1 = strongly disagree

5 = strongly agree

W1 2321 3.41 � 0.95 0.893

W4 993 3.56 � 0.99 0.886

Item range: 1–5

4. R

4.1.

anasecsumDatageoveposturin acor

(91degfulldiretheMapatdueMo(IPWsamin t

4.2.

satishoeacfor

parandwitWa

Fig.

not

C.S. Brewer et al. / International Journal of Nursing Studies 52 (2015) 1735–1745 1741

esults

Sample characteristics

A sample of 2386 (Wave 1–2) was used for the first datalysis, and a sample of 1073 (Wave 4–5) was used for theond analysis. Findings from descriptive analyses are

marized in Tables A2 and A3 in the Supplementarya. In the sample of 2386, participants had an average

of 32.45 years (S.D. = 8.84). In Wave 1 mandatoryrtime, negative affectivity, job change, direct careition and gender were positively correlated withnover. In both Wave 1 and 4 job opportunities, working

hospital (negatively), and type of unit (positively) wererelated with turnover.The majority of the sample was White (82.3%), female.1%), and held an associate (56.5%) or baccalaureateree (39.3%) as their basic nursing degree. Most worked-time (84.5%) in hospital settings (87.1%) and providedct care (92.3%). About 10.8% of participants changed

ir job after working a year as a RN. Full Informationximum Likelihood (FIML) estimation was used in ourh models to allow inclusion of cases with data missing

to attrition in our analysis (Wave 1–2 sample).reover, we also used inverse probability weighting

; Seaman and White, 2014) based on Wave 1–5ple attrition to control for any potential attrition bias

he Wave 4–5 sample due to attrition.

Correlations between major variables

As seen in Table A4 in the supplementary data, jobsfaction, organizational commitment, intent, search,ck, and turnover were significantly correlated withh other for the Wave 1 and Wave 2 samples, as well asWave 4 and Wave 5 samples, except for search. In

ticular, job satisfaction, organizational commitment, intent to stay were significantly negatively correlatedh turnover for both Wave 1–2 samples (p < .01) andve 4–5 samples (p < .01). Search and shock were

significantly correlated with turnover for the Wave 1–2samples (p < .01), but not for the Wave 4–5 samples.

4.3. Test of the hypothesized model

For Waves 1–2 sample (N = 2386) the initial analysis(Model 1, not shown) of the hypothesized model withoutcontrol variables revealed a good fit to the data(x2 = 24.348, df = 3, p = 0.00, CFI = 0.988, RMSEA = 0.055,R2: organizational commitment = 0.437, intent = 0.626,search = 0.037, turnover = 0.170). When control variableswere added (Model 2, not shown), the model continued toprovide a good fit to the data (x2 = 224.894, df = 36,p = 0.00, CFI = 0.968, RMSEA = 0.047, R2: organizationalcommitment = 0.672, intent = 0.704, search = 0.040, turn-over = 0.210). We examined modification indices forpotential specification errors, and modification indicessuggested that paths from work units and local jobopportunities be changed from turnover to organizationalcommitment (modification indices > 132.266). The finalmodel (Model 3, see Fig. 1) including these two additionalpaths provided a good fit to the data (x2 = 132.786, df = 36,p = 0.00, CFI = 0.983, RMSEA = 0.034; R2: organizationalcommitment = 0.566, intent = 0.693, search = 0.040, turn-over = 0.213).

We tested the same model (Model 4) with the Wave 4–5 sample (N = 1073). Model 4 (see Fig. 2) also provided agood fit to the data x2 = 82.336, df = 36, p = 0.00, CFI = 0.983,RMSEA = 0.035, R2: organizational commitment = 0.409,intent = 0.657, search = 0.012, turnover = 0.165.

In none of the models was there direct significantrelationships between job satisfaction or organizationalcommitment and turnover, but there were significantdirect positive relationships between job satisfaction ororganizational commitment and intent to stay, and thenbetween intent to stay and turnover (see Table 2 for tests ofindirect effects). In the final model for Wave 1–2 sample(Model 3), all of the indirect pathways through search toturnover were statistically significant (see Table 2 for testsof indirect effects). However, when Model 4 was rerun with

1. Model 3: Wave 1–2; Sample (N = 2386). Note: *p < .05, **p < .01; unstandardized coefficient (standard coefficients); paths from control variables are

presented working units** and local job opportunities** significantly predict OC; settings** and changed job** significantly predict turnover.

C.S. Brewer et al. / International Journal of Nursing Studies 52 (2015) 1735–17451742

the respondents from Wave 4–5 sample (N = 1073), none ofthe paths through search were statistically significant, norwas the path from shock to turnover.

Models 5 to7 (see these were not included to proof???-Supplementary Data Figures A1, A2, and A3) weredeveloped to identify whether the lack of significancefor these two pathways was a sample attrition effect orhistory effect: we proceeded in three ways. All three testsproduced fairly comparable results in terms of themagnitude and significance of the pathways. First, were-ran the Wave 1–2 model using the subsample that waspresent at Wave 4 and 5 (N = 1073). If the model (Model 5)was the same as the Wave 1 and 2 model estimated withthe entire sample (N = 2386), then differences between

the Waves 1 and 2 and Waves 4 and 5 models could beattributable to a history or experience effect. The Model5 yielded a good fit to the data (x2 = 107.891, df = 36,p = 0.00, CFI = 0.969, RMSEA = 0.043, R2: organizationalcommitment = 0.558, intent = 0.629, search = 0.037, turn-over = 0.196). Model 5 results suggested that the path fromsearch to intent was still significant, but shock was notsignificantly related to turnover. The findings suggestedthat our results are robust to a small attrition bias andmissing data.

Thus taken together these tests suggest that differencesin our Wave 1 to Wave 2 and Wave 4 to Wave 5 models areattributable to history effects such as changes in behaviorof nurses over time rather than sample bias or differences

Fig. 2. Model 4: Wave 4–5; Sample (N = 1073). Note: *p < .05, **p < ; unstandardized coefficient (standard coefficients); paths from control variables are not

presented; working units** and local job opportunities** significantly predict OC; settings** significantly predict turnover.

Table 2

Indirect effects for structural paths.

All structural paths from: Model 3 (Fig. 1) Model 4 (Fig. 2)

JS ! OC ! search �.025** (�.084**) �.009 (�.027)

JS to intent to stay .219** (.337**) .181** (.260**)

� JS ! OC ! intent .213** (.327**) .178** (.256**)

� JS ! OC ! search ! intent .003** (.005**) .001 (.001)

� JS ! search ! intent .003* (.005*) .002 (.003)

JS to turnover �.159** (�.233**) �.201** (�.286**)

� JS ! OC ! turnover .041 (.061) .025 (.036)

� JS ! OC ! intent ! turnover �.089** (�.131**) �.080** (�.114**)

� JS ! OC ! search ! turnover �.003 (�.005) .001 (.001)

� JS ! OC ! search ! intent ! turnover �.001* (�.002*) .000 (.000)

� JS ! intent ! turnover .105** (�.154**) .145** (�.207**)

� JS ! search ! intent ! turnover �.001 (�.002) �.001 (�.001)

OC ! search !intent .009** (.007**) .002 (.002)

OC to turnover �.277**(�.213**) �.252** (�.192**)

� OC ! intent ! turnover �.263**(�.202**) �.253** (�.193**)

� OC ! search ! turnover �.010 (�.008) .002 (.002)

� OC ! search ! intent ! turnover �.004* (�.003*) �.001 (�.001)

Search ! intent ! turnover .052** (.023**) .037 (.017)

Note. *p < .05; **p < .01; Unstandardized coefficient (standardized coefficients); JS = job satisfaction; OC = organiza-

tional commitment.

in texpnurhyp

5. D

comstabwaturmo

comindmeactconin tbutpatsam

encdeg200thaprefactindinteit mto

tha

feequiif aandThecanstagthonurma

Tab

Resu

#

1a

1b

1c

1d

2a

2b

2c

3a

3b

4

5

Note

C.S. Brewer et al. / International Journal of Nursing Studies 52 (2015) 1735–1745 1743

he markets. It also points to the value to employers thaterienced nurses may have in the market over newses. In summary, Table 3 summarizes the results for theotheses as follows from Models 3 and 4).

iscussion

Relationships among job satisfaction, organizationalmitment, intent to stay, shock, and turnover werele as the early career RN sample matured. We also

nted to sort out the multiple pathways that impactnover. We believe this is the only study to test a turnoverdel in one sample at two different time periods.The central roles of job satisfaction, organizational

mitment, and intent are supported. There was noication that job satisfaction or organizational commit-nt had direct effects on turnover; instead they clearlyed indirectly through intent. The role of search was notsistent. All of the paths through search were significanthe larger and earlier Wave 1–2 sample (Models 1–3);, in the smaller Wave 1–2 sample (Model 5) only theh from search to intent was significant; in the Wave 4–5ple (Model 4) this path was not significant.

There are several possible explanations for these differ-es. First, our measure of search was global, indicating theree to which employees are looking for other jobs (Price,1). The meta-analysis by Griffeth et al. (2000) identifiedt search intention was the only significant turnoverdictor; but Hom and Griffeth (1991) have shown that in

search could be a complex construct that needs toicate more specific behaviors. If our measure is closer tont to search rather than degree of actual search, thenay overlap considerably with the formation of the intent

leave, although Griffeth and colleagues (2000) foundt search intent was a significant predictor.A second explanation is that while newer nurses mayl the need to search more carefully for a job before theyt, as nurses mature they may not spend significant time,ny, searching for a job. They know the job market

are likely have friends or colleagues in many settings.y are likely to be relatively confident that they can ornot get a job when they want it. Thus, search as ae may be less important for later career nurses than for

se in other occupations. The historical shortages in thesing labor market, particularly for experienced nurses,

Lastly, another explanation could be that the two yeartime difference between Wave 4 and 5 compared to theone year time difference in Wave 1–2 may have attenuatedthe relationships, so that search became unimportant.

The other hypothesis that was not supported was thelack of significance in the model for Wave 4–5 for shock.The theory (Morrell et al., 2004; Morrell, 2005) suggeststhat any number of unexpected events could be shock, anda back strain for nurses working in hospitals, as the majorityof the nurses in this study did at Wave 1–2 (87.1%) couldsignificantly impair the RNs ability to care for patients. ByWave 4–5, however, only 72.6% of the sample were workingin hospitals. Fewer nurses thus were exposed to thisoccupational hazard, and if they were working for theirpreferred employer by that time, they may have moved to asetting with that employer that reduced the risk of injury orthe intent to stay was stronger than the risk of injury. Wealso explicitly defined turnover as leaving the employer, soit is possible that nurses who experienced back strains couldstay with the employer but move to a less physicallydemanding position. One of the major historical effects wasthe economic recession beginning in 2008 and resultinghigh unemployment rates may also have influenced thesenurses. While we controlled for perception of job opportu-nities, it could be that nurses were less likely to leave dueto shocks of any kind because of economic uncertainties. Asdiscussed above, it could also be due to the time differencebetween the surveys attenuating the relationships.

A research implication is that studies that excludeeither organizational commitment or job satisfaction havean omitted variable bias, as both job satisfaction andorganizational commitment are clearly important factorsin influencing intent to stay. Employer-modifiable pre-dictors of job satisfaction and organizational commitmentneed to be identified and tested as solutions to turnoverbecause of the clear and strong relationship to intent. A greatdeal of research has been conducted on job satisfaction, andsomewhat less on organizational commitment, but fewsystematically tested interventions exist. Reducing turnoveris a major goal of most employers, especially in a growtheconomy. Helping employers understand what factors havethe ‘‘biggest bang for the buck,’’ or in research language,effect size, would be a useful direction for the field.Implementation of interventions to improve job satisfactionand organizational commitment must be developed and

le 3

lts of Hypotheses Testing.

Hypotheses Wave 1–2 Wave 4–5

JS significantly directly and positively predicts OC Yes Yes

JS has a significant negative direct effect on job search Yes No

JS has a significant positive direct effect on intent to stay Yes Yes

JS does not have a significant effect on turnover Yes Yes

OC has a significant negative direct effect on search Yes No

OC has a significant positive direct effect on intent to stay Yes Yes

OC does not have a significant effect on turnover Yes Yes

Job search has a significant negative direct effect on intent to stay Yes No

Job search has a significant positive direct effect on turnover Yes No

Intent to stay has a significant negative direct effect on turnover Yes Yes

A work related injury/shock significant direct predicts turnover Yes No

. JS = job satisfaction; OC = organizational commitment.

ed in order to control and manage personnel turnover.

ke this especially likely. test

C.S. Brewer et al. / International Journal of Nursing Studies 52 (2015) 1735–17451744

5.1. Limitations

While the large sample size is a strength of this study,over time there has been a drop in the sample size fromWave 1 to Wave 5. While we carefully controlled forpotential bias that may have resulted due to the change insample sizes, we can confidently state that the differencesin the pathways is not a result of differences in the sample.Our data indicated that there have been only a few changesin the sample characteristics. We found only one signifi-cant difference on non-local job opportunities betweennurses who did respond in Wave 1 to Wave 5 compared tothose who did not (*p < .05; **p < .01).

Unfortunately, when we started this study we did notdistinguish involuntary leaving from other reasons forleaving. However, for Waves 2 and 4 we did ask if theywere laid off; for these nurse respondents, only 0.3% and3.1%, respectively, admitted to being laid off. It is also likelythat in any study asking this question, social response biaswould result in a lack of honesty in the answers anyway. Tomitigate this problem we used the phrase laid off ratherthan fired. We think the potential influence is very small,given the very small percentages in Wave 2 and 4.

6. Conclusions

There is strong support over time for the relationshipshypothesized among job satisfaction, organizationalcommitment, intent and turnover, with some supportfor shock and search for nurses who have just entered theworkforce. We controlled for both missing data andattrition bias, strengthening the conclusion that nurseswho have been in the workforce for longer time periodsare less likely to turnover if they have a back injury, andwere less likely to search for a job before finding theirnext one. In both cases the important role of jobsatisfaction as well as organizational commitment, andthe impact of job satisfaction on organizational commit-ment, in reducing turnover and enhancing retention areclear and have been addressed in many studies. We needmore research on programs that implement changes inthe work environment that impact these two outcomes,as well as research that focuses on the relevant strength orimpact (e.g. ‘‘bang for the buck’’) to help administratorsprioritize translation of results. Employers that focuson factors directly contributing to the job satisfactionand organizational commitment of their employees arelikely to improve turnover and retention.

Conflict of interest: None.

Funding: The Robert Wood Johnson Foundation.

Ethical approval: Approved by the Institutional Review Boards

of both the University at Buffalo and New York University.

Appendix A. Supplementary data

Supplementary data associated with this article can befound, in the online version, at http://dx.doi.org/10.1016/j.

References

Aiken, L.H., Patrician, P.A., 2000. Measuring organizational traits of hos-pitals: the revised nursing work index. Nurs. Res. 49 (3), 146–153.

Allen, D.G., Shore, L.M., Griffeth, R.W., 2003. The role of perceivedorganizational support and supportive human resource practicesin the turnover process. J. Manage. 29 (1), 99–118.

Beecroft, P.C., Dorey, F., Wenten, M., 2008. Turnover intention in newgraduate nurses: A multivariate analysis. J. Adv. Nurs. 62 (1), 41–52.

Blau, G., 2007. Does a corresponding set of variables for explainingvoluntary organizational turnover transfer to explaining voluntaryoccupational turnover? J. Vocat. Behav. 70 (1), 135–148.

Bollen, K.A., 1989. Structural Equations with Latent Variables. John Wiley& Sons, New York.

Brewer, C.S., Kovner, C.T., Poornima, S., Fairchild, S., Kim, H., Djukic, M.,2009. A comparison of second degree baccalaureate and traditionalbaccalaureate new graduate RNs: implications for the workforce.J. Profess. Nurs. 25 (1) , 5-5-14.

Brewer, C.S., Kovner, C.T., Greene, W., Tukov-Shuser, M., Djukic, M., 2012.Predictors of actual turnover in a national sample of newly licensedregistered nurses employed in hospitals. J. Adv. Nurs. 68 (3), 521–538,http://dx.doi.org/10.1111/j.1365-2648.2011.05753.x.

Camerino, D., Conway, P., van der Heijden, B., Estryn-Behar, M., Costa, G.,Hasselhorn, H., 2008. Age-dependent relationships between workability, thinking of quitting the job, and actual leaving among italiannurses: a longitudinal study. Int. J. Nurs. Stud. 45 (11), 1645–1659.

Castle, N.G., 2006. Organizational commitment and turnover of nursinghome administrators. Health Care Manage. Rev. 31 (2), 156–165.

Chen, H., Chu, C., Wang, Y., Lin, L., 2008. Turnover factors revisited: a longi-tudinal study of Taiwan-based staff nurses. Int. J. Nurs. Stud. 45 (2), 277.

Coomber, B., Barriball, K.L., 2007. Impact of job satisfaction componentson intent to leave and turnover for hospital-based nurses: a review ofthe research literature. Int. J. Nurs. Stud. 44 (2), 297–314.

Cox, C.W., Relf, M.V., Chen, R., Zangaro, G.A., 2010. The retention ofrecalled United States navy nurse reservists. Nurs. Outlook 58 (4),214–220, http://dx.doi.org/10.1016/j.outlook.2010.03.001.

De Gieter, S., Hofmans, J., Pepermans, R., 2011. Revisiting the impact of jobsatisfaction and organizational commitment on nurse turnover in-tention: an individual differences analysis. Int. J. Nurs. Stud. 48 (12),1562–1569, http://dx.doi.org/10.1016/j.ijnurstu.2011.06.007.

Dillman, D.A., 2007. Mail and Internet Surveys: The Tailored DesignMethod, 2nd ed. John Wiley & Sons, Inc., New Jersey.

Estryn-Behar, M., van der Heijden, B.I.J.M., Fry, C., Hasselhorn, H., 2010.Longitudinal analysis of personal and work-related factors associatedwith turnover among nurses. Nurs. Res. 59 (3), 166–177, http://dx.doi.org/10.1097/NNR.0b013e3181dbb29f.

Felps, W., Mitchell, T.R., Hekman, D., Lee, T., Holtom, B.C., Harman, W.S.,2009. Turnover contagion: how coworkers’ job embeddednessand job search behaviors influence quitting. Acad. Manage. J. 52(3), 545–561, http://dx.doi.org/10.5465/AMJ.2009.41331075.

Gaertner, S., 1999. Structural determinants of job satisfaction and orga-nizational commitment in turnover models. Hum. Resour. Manage.Rev. 9 (4), 479–493.

Galletta, M., Portoghese, I., Penna, M.P., Battistelli, A., Saiani, L., 2011.Turnover intention among italian nurses: the moderating roles ofsupervisor support and organizational support. Nurs. Health Sci. 13(2), 184–191, http://dx.doi.org/10.1111/j. 1442-2018.2011.00596.x.

Garbee, D.D., Killacky, J., 2008. Factors influencing intent to stay inacademia for nursing faculty in the southern united states of America.Int. J. Nurs. Educ. Scholars. 5 (1), 1–15.

Gilmartin, M.J., 2012. Thirty years of nursing turnover research: lookingback to move forward. Med. Care Res. Rev., http://dx.doi.org/10.1177/1077558712449056.

Griffeth, R.W., Hom, P.W., Gaertner, S., 2000. A meta-analysis of ante-cedents and correlates of employee turnover: update, moderatortests, and research implications for the next millennium. J. Manage.26 (3), 463–488, http://dx.doi.org/10.1016/S0149-2063(00)00043-X.

Griffeth, R.W., Steel, R.P., Allen, D.G., Bryan, N., 2005. The development of amultidimensional measure of job market cognitions: the employ-ment opportunity index (EOI). J. Appl. Psychol. 90 (2), 335–349,http://dx.doi.org/10.1037/0021-9010.90.2.335.

Gurney, C.A., Mueller, C.W., Price, J.L., 1997. Job satisfaction and organi-zational attachment of nurses holding doctoral degrees. Nurs. Res. 46(3), 163–171.

Hackman, J.R., Oldham, G.R., 1975. Development of the job diagnosticsurvey. J. Appl. Psychol. 60 (2), 159–170.

Halbesleben, J.R.B., Wheeler, A.R., 2008. The relative roles of engagementand embeddedness in predicting job performance and intention to

leave. Work Stress 22 (3), 242. ijnurstu.2015.06.017.

Harr

Hay

Hom

Hu,

Inge

Jone

Jose

Kash

Kim

Kim

Kirs

Klin

Kov

Kov

Lee,

Lee,

Lee,

Li, Y

Lu, H

Lum

Mac

Mee

Mey

Mob

Mor

Mor

Mor

C.S. Brewer et al. / International Journal of Nursing Studies 52 (2015) 1735–1745 1745

ington, C., Swan, J.H., 2003. Nursing home staffing, turnover, and casemix. Med. Care Res. Rev. 60 (3), 366–392, http://dx.doi.org/10.1177/1077558703254692.es, L.J., O’Brien-Pallas, L., Duffield, C., Shamian, J., Buchan, J., Hughes, F.,Stone, P.W., 2006. Nurse turnover: a literature review. Int. J. Nurs. Stud.43 (2), 237–263, http://dx.doi.org/10.1016/j.ijnurstu.2005.02.007., P.W., Griffeth, R.W., 1991. Structural equations modeling test of a

turnover theory: cross-sectional and longitudinal analyses. J. Appl.Psychol. 76, 350–366.L.T., Bentler, P.M., 1999. Cutoff criteria for fit indexes in covariancestructure analysis: conventional criteria versus new alternatives.Struct. Eq. Model. 6 (1), 1–55.rsoll, G.L., Olsan, T., Drew-Cates, J., DeVinney, B.C., Davies, J., 2002.Nurses’ job satisfaction, organizational commitment, and career in-tent. J. Nurs. Admin. 32 (5), 250–263.s, C.B., Gates, M., 2007. The costs and benefits of nurse turnover: abusiness case for nurse retention. OJIN: Online J. Issues Nurs. 12 (3) .phson, M., Lindberg, P., Voss, M., Alfredsson, L., Vingard, E., 2008. Thesame factors influence job turnover and long spells of sick leave – a 3-year follow-up of Swedish nurses. Eur. J. Public Health 18 (4), 380–385,http://dx.doi.org/10.1093/eurpub/ckn009., B.A., Castle, N.G., Phillips, C.D., 2007. Nursing home spending,

staffing, and turnover. Health Care Manage. Rev. 32 (3), 253–262., E., Hwang, J., 2011. Characteristics associated with intent to stayamong quality improvement nurses. Int. Nurs. Rev. 58 (1), 89–95,http://dx.doi.org/10.1111/j. 1466-7657.2010.00841.x., S., Price, J.L., Mueller, C.W., Watson, T.W., 1996. The determinants ofcareer intent among physicians at a U.S. air force hospital. Hum. Relat.49 (7), 947–976, http://dx.doi.org/10.1108/EUM0000000006233.chling, J., Marie Colgan, C., Andrews, B., 2011. Predictors of registerednurses’ willingness to remain in nursing. Nurs. Econ. 29 (3), 111–117.e, R.B., 2005. Principles and Practice of Structural Equation Modeling,2nd ed. Guilford, New York.ner, C.T., Brewer, C.S., Fairchild, S., Poornima, S., Kim, H., Djukic, M.,2007. Newly licensed RNs’ characteristics, work attitudes, and inten-tions to work. Am. J. Nurs. 107 (9), 58–70.ner, C.T., Brewer, C.S., Greene, W.F.S., 2009. Understanding new regis-tered nurses’ intent to stay at their jobs. Nurs. Econ. 27 (2), 81–98.

T.H., Gerhart, B., Weller, I., Trevor, C.O., 2008. Understanding voluntaryturnover: path-specific job satisfaction effects and the importanceof unsolicited job offers. Acad. Manage. J. 51 (4), 651–671.

T.W., Mitchell, T.R., 1994. An alternative approach: the unfolding modelof voluntary employee turnover. Acad. Manage. Rev. 19 (1), 51–89.

T.W., Mitchell, T.R., Holtom, B.C., McDaneil, L.S., Hill, J.W., 1999. Theunfolding model of voluntary turnover: a replication and extension.Acad. Manage. J. 42 (4), 450–462.., Jones, C.B., 2013. A literature review of nursing turnover costs. J.Nurs. Manage. 21 (3), 405–418, http://dx.doi.org/10.1111/j.1365-2834.2012.01411.x.

., While, A.E., Louise Barriball, K., 2007. A model of job satisfaction ofnurses: a reflection of nurses’ working lives in mainland china. J. Adv.Nurs. 58 (5), 468–479., L., Kervin, J., Clark, K., Reid, F., Sirola, W., 1998. Explaining nursingturnover intent: job satisfaction, pay satisfaction, or organizationalcommitment? J. Org. Behav. 19 (3), 305–320.Kinnon, D.P., 2008. Introduction to Statistical Mediational Analysis.Lawrence Erlbaum Associates, NY.usen, V.C.H., Van Dam, K., BrownMahoney, C., Van Zundert, A.A.J.,Knape, H.T.A., 2011. Understanding nurse anesthetists’ intentionto leave their job: how burnout and job satisfaction mediate theimpact of personality and workplace characteristics. Health CareManage. Rev. 36 (2), 155–163, http://dx.doi.org/10.1097/HMR.0b013e3181fb0f41.er, J.P., Stanley, D.J., Herscovitch, L., Topolnytsky, L., 2002. Affective,continuance, and normative commitment to the organization: ameta-analysis of antecedents, correlates, and consequences. J. Vocat.Behav. 61 (1), 20–52.ley, W.H., 1982. Employee Turnover: Causes, Consequences, and

Control. Addison-Wesley, Reading, MA.Barak, M.E., Nissly, J.A., Levin, A., 2001. Antecedents to retention andturnover among child welfare, social work, and other human serviceemployees: what can we learn from past research? A review andmetanalysis. Soc. Serv. Rev. 75 (4), 625–661, http://dx.doi.org/10.1086/323166.rell, K., Loan-Clarke, J., Wilkinson, A., 2004. The role of shocks inemployee turnover. Br. J. Manage. 15 (4), 335–349.rell, K., 2005. Towards a typology of nursing turnover: the role ofshocks in nurses’ decisions to leave. Nurs. Health Care Manage. Policy49 (3), 315–322.

Mowday, R.T., Steers, R.M., Porter, L.W., 1979. The measurement oforganizational commitment. J. Vocat. Behav. 14 (2), 224–247.

Mueller, C.W., Price, J.L., 1990. Economic, psychological, and sociologicaldeterminants of voluntary turnover. J. Behav. Econ. 19 (3), 321–335.

Murrells, T., Robinson, S., Griffiths, P., 2008. Is satisfaction a directpredictor of nursing turnover? Modelling the relationship betweensatisfaction, expressed intention and behaviour in a longitudinalcohort study. Hum. Resour. Health 6 (1), 22–33, http://dx.doi.org/10.1186/1478-4491-6-22.

Muthen, B., du Toit, S.H.C., Spisic, D., 1997. Robust inference usingweighted least squares and quadratic estimating equations in latentvariable modeling with categorical and continuous outcomes. Re-trieved from http://www.statmodel.com/bmuthen/articles/Article_075.pdf (accessed 14.05.15).

Nooney, J.G., Unruh, L., Yore, M.M., 2010. Should I stay or should I go?Career change and labor force separation among registered nurses inthe U.S.. Soc. Sci. Med. 70 (12), 1874–1881, http://dx.doi.org/10.1016/j.socscimed.2010.02.037.

Paquet, M., Courcy, F., Lavoie-Tremblay, M., Gagnon, S., Maillet, S., 2013.Psychosocial work environment and prediction of quality of care indi-cators in one Canadian health center. Worldviews Evidence-Based Nurs.10 (2), 82–94, http://dx.doi.org/10.1111/j. 1741-6787.2012.00250.x.

Price, J.L., 2001. Reflections on the determinants of voluntary turnover.Int. J. Manpower 22 (7), 600–624.

Price, J.L., 2004. The development of a causal model of voluntary turnover.In: Griffeth, R.W., Hom, P.W. (Eds.), Innovative Theory of EmpiricalResearch on Employee Turnover. pp. 3–32.

Price, J.L., Mueller, C.W., 1981. A causal model for turnover for nurses.Acad. Manage. J. 24 (3), 543–565.

Quinn, R.P., Staines, G.L., 1979. The 1977 Quality of Employment Survey:Descriptive Statistics, with Comparison Data from the 1969–70 andthe 1972–73 Surveys. Survey Research Center, Institute for SocialResearch, University of Michigan, Ann Arbor, MI.

Reay, T., Berta, W., Kohn, M.K., 2009. What’s the evidence on evidence-based management? Acad. Manage. Perspect. 23 (4), 5–18.

Seaman, S., White, I., 2014. Inverse probability weighting with missingpredictors of treatment assignment or missingness. Commun. Stat. –Theor. Methods 43, 3499–3515.

Simon, M., Muller, B.H., Hasselhorn, H.M., 2010. Leaving the organizationor the profession? A multilevel analysis of nurses? intentions. J. Adv.Nurs. 66 (3), 616–626, http://dx.doi.org/10.1111/j.1365-2648.2009.05204.x.

Smith, D.R., Holtom, B.C., Mitchell, T.R., 2011. Enhancing precision in theprediction of voluntary turnover and retirement. J. Vocat. Behav. 79(1), 290–302, http://dx.doi.org/10.1016/j.jvb.2010.11.003.

Somers, M.J., 1996. Modelling employee withdrawal behaviour over time:a study of turnover using survival analysis. J. Occup. Org. Psychol. 69(4), 315–326.

Staggs, V.S., Dunton, N., 2012. Hospital and unit characteristics associatedwith nursing turnover include skill mix but not staffing level: anobservational cross-sectional study. Int. J. Nurs. Stud. 49 (9), 1138–1145, http://dx.doi.org/10.1016/j.ijnurstu.2012.03.009.

Suzuki, E., Itomine, I., Saito, M., Katsuki, T., Sato, C., 2008. Factors affectingthe turnover of novice nurses at university hospitals: A two yearlongitudinal study. Jpn. J. Nurs. Sci. 5 (1), 9–21.

Suzuki, E., Tagaya, A., Ota, K., Nagasawa, Y., Matsuura, R., Sato, C., 2010.Factors affecting turnover of Japanese novice nurses in universityhospitals in early and later periods of employment. J. Nurs. Manage.18 (2), 194–204, http://dx.doi.org/10.1111/j. 1365-2834.2010.01054.x.

Swider, B.W., Boswell, W.R., Zimmerman, R.D., 2011. Examining the jobsearch–turnover relationship: the role of embeddedness, job satis-faction, and available alternatives. J. Appl. Psychol. 96 (2), 432–441,http://dx.doi.org/10.1037/a0021676.

Vandenberg, R.J., Lance, C.E., 1992. Examining the causal order ofjob satisfaction and organizational commitment. J. Manage. 18 (1),153–167, http://dx.doi.org/10.1177/014920639201800110.

Wagner, C.M., 2007. Organizational commitment as a predictor variablein nursing turnover research: literature review. J. Adv. Nurs. 60 (3),235–247.

Wang, L., Tao, H., Ellenbecker, C.H., Liu, X.H., 2012. Predictors of hospitalnurses’ intent to stay: A cross-sectional questionnaire survey inshanghai, china. Int. Nurs. Rev. 59 (4), 547–554, http://dx.doi.org/10.1111/j.1466-7657.2012.01009.x.

Weiss, D.J., Dawis, R.V., England, G.W., Lofquist, L.H., 1967. Manual forthe Minnesota satisfaction questionnaire. University of Minnesota,Minneapolis.

Whitley, M.P., Putzier, D., 1994. Measuring nurses’ satisfaction with thequality of their work and work environment. J. Nurs. Care Qual. 8 (3),43–51.