Burnout and Physical Health among SocialWorkers: A Three-Year Longitudinal Study

Hansung Kim, Juyeji, and Dennis Kao

The high risk of burnout in the social work profession is well established, but little is knownabout burnout's impact on the physical health of social workers. This article examines therelationship between burnout and physical health, using data from a longitudinal study ofsocial workers. California-registered social workers (N = 406) were surveyed annually overa three-year period. Using structural equation modeling, the authors conducted a pathanalysis to test whether burnout predicted changes in physical health over time. The resultsshowed that social workers with higher initial levels of burnout later reported more physicalhealth complaints. Moreover, higher levels of burnout led to a faster rate of deterioration inphysical health over a one-year period.The potential implications for policy and social serviceorganizations are discussed.

KEYWORDS: bumout; longitudinal data; physical health; social workers

Social workers deliver a broad range of vitalservices and, thus, play a critical role in en-suring the health and well-being of society's

most vulnerable members. Ho'wever, the highdemand for their services—coupled with increas-ingly diminishing resources—can present significantchallenges for social workers. According to a recentnational study (Whitaker,Weismiller, & Clark, 2006),social workers face increasing levels of paperworkand inadequate supervision. Beset by chronic staffshortages and turnover, social workers tend to beoverworked and are often asked to take on large cli-ent caseloads. Given the heavy demands placed onthem, it is not surprising that social workers oftenexperience psychological distress and, eventually,high levels of burnout.

Burnout is a prolonged psychological responseto chronic workplace Stressors and is theorized toinclude three dimensions: emotional exhaustion,depersonalization or cynicism, and diminished per-sonal accomplishment (Maslach, Schaufeli, & Leiter,2001). Over the past decade, clear empirical evidencehas emerged regarding the prevalence of burnoutin the social work profession. For example, on thebasis of a sample of 751 social workers, Siebert (2005)found that about three-fourths reported having hadtrouble with burnout during their careers. Thishigh prevalence of burnout among social workershas been found across the broad range of practicefields. For example, Oktay (1992) found high lev-

els of emotional exhaustion and depersonalizationamong hospital AIDS social workers. Poulin andWalter (1993a) examined burnout among 1,196gerontological social -workers and found that about60 percent of them were experiencing moderateto high levels of emotional exhaustion. Similarly,Lloyd and King (2004) reported that Australiansocial workers working in mental health settingsexhibited high levels of emotional exhaustion. Ina study of 151 fiontline child protective services(CPS) workers, Anderson (2000) found that about62 percent of participants were experiencing highlevels of emotional exhaustion.

Social worker burnout is a serious problem be-cause it can adversely affect the quality and stabilityof social services. The hterature suggests that socialworkers experiencing burnout have an increased riskof psychological distress, such as depression (Evanset al., 2006; Siebert, 2004; N. Stanley, Manthorpe, &White,2007), which can ultimately lead to increasedturnover (Kim & Stoner, 2008; Mor Barak, Nissly,& Levin, 2001). Although a significant number ofresearch studies have shown the negative effects ofburnout on social workers'practice and performanceoutcomes, less is understood about the effects ofburnout on the physical health of social workers.

The role of psychological distress in physical illnessis more established for the general workforce. Forexample, in a 20-year follow-up study based on aU.S. national sample, Ferraro and Nuriddin (2006)

258 ccc Code: oc;" ::ir..;_"_"" _"' I Workers

found that high levels of distress raised mortalityrisk. Darr and Johns (2008) recently conducteda meta-analysis to examine the negative effectsof work strain on the psychological and physicalsymptoms that often lead to absenteeism at work;their results suggest that psychological symptoms doaffect physical illness. Extensive research has shownthat job-related Stressors and strain can have adversehealth effects through several potential mechanisms,such as decreased immunity functioning and poorhealth behaviors (Melamed, Shirom,Toker, Berliner,& Shapira, 2006). For example, Nakamura, Hágase,Yoshida, and Ogino (1999) found that deperson-alization (one of the dimensions of burnout) wasassociated •with diminished cellular immunity. Otherstudies suggest that job Stressors can also lead tonegative health behaviors, such as smoking, alcoholor substance abuse, and less exercise (Johansson,Johnson, & Hall, 1991)—all activities known tohave adverse effects on a person's health (F. Jones& Bright, 2001).

In the burnout literature, there is substantialevidence on how burnout can affect the physicalhealth of a country's workforce. For example, in anational study of workers in Finland, Honkonenet al. (2006) found that physical illness was morecommon among workers with burnout, includ-ing musculoskeletal disorders among women andcardiovascular diseases among men. In sum, theresearch has shown that burnout can negativelyaffect overall self-rated health status (Peterson etal., 2008) and can lead to a broad range of healthproblems, including somatic complaints (Soares,Grossi, of Sundin, 2007); cardiovascular diseases(Melamed et al., 2006; Toker, Shirom, Shapira, Ber-liner, & Melamed, 2005); sleep disturbances (Grossi,Perski, Evengard, Blomkvist, & Orth-Gomer,2003; Söderström, Ekstedt, Âkerstedt, Nilsson, &Axelsson, 2004); headaches (T. L. Stanley, 2004);flu-hke illnesses, common colds, or incidences ofgastroenteritis (Mohren et al, 2003); and illness-related absences from work (Burke & Mikkelsen,2006). Furthermore, job stress—which can causeburnout—is a significant predictor of gastrointes-tinal problems (EdéU-Gustafson, Kritz, & Bogren,2002; Sveinsdóttir, Gunnarsdóttir, & Riksdóttir,2007). Diminishing physical health can lead to lostworkdays, diminished job effectiveness, permanentdisabilities, and increased compensation for sickleave (Eriksen, Svendsrod, Ursin, & Ursin, 1998;Schwartz, Stewart, Simon, & Lipton, 1998).

The relationship between burnout and physicalillness has also been found among professionalswho may work in settings similar to those of socialworkers. Among Swedish health care •workers, forexample, Peterson et al. (2008) found that burnoutwas associated vî ith lower self-rated health status,sleep disturbance, and neck and back pain.

Despite the high risk of burnout in the socialwork profession, little is known about its impact onthe physical health of social workers.To address thisgap in the research, the present study used longitu-dinal data from a sample of social workers registeredin California to examine the impact of burnout ontheir physical health.

METHOD

Data and SampleData were drawn from a longitudinal study examin-ing the job-related factors associated •with burnoutamong social workers. In this study, participants•were randomly selected from a California registry ofclinical social workers and surveyed three times (ap-proximately annually) for a three-year period from2005 to 2007. Information was gathered throughthe use of mail questionnaires, which includedboth standardized instruments (for example, theMaslach Burnout Inventory—Human Service Survey[Maslach & Jackson, 1986]) and sociodemographicmeasures. The physical health measures were notincluded in the baseline survey (in 2005), but theywere used in the second and third surveys (in 2006and 2007) .When resources •were available, individu-als were sent multiple follow-up questionnaires inan effort to maximize the response rate. The insti-tutional review boards of the authors' universitiesapproved this research.

The current study's sample included participantswho completed the wave 2 questionnaires, in whichthe physical health measure was first collected. Ofthe basehne sample (N = 406), about 70 percentparticipated in the second wave (n = 285), andabout 36 percent participated in the third wave (n= 146). Because the study used social workers whoparticipated in the wave 2 questionnaire (n = 285),it was important to determine whether the studysample was similar to the initial sample and •whetherthe attrition •was random. As a preliminary analysis,•we conducted logistic regression models to examinethe relationship among the study variahles (that is,age, gender, years in the field, income, and initiallevel of burnout) and the likelihood of a partici-

among Social Workers: A Three-Year Longitudinal Study 259

pant dropping out at waves 2 or 3 and found nosignificant relationships (results not reported here).It was therefore assumed that the study's attritionwas random and that our samples were similar acrossall three waves.

MeasuresPhysical Health Complaints. Physical health wasmeasured using the Physical Health Questionnaire(PHQ) (Schat, Kelloway, & Desmarais, 2005). Forthis analysis, we focused on four types of physicalhealth problems: sleep disturbances, headaches, re-spiratory infections, and gastrointestinal infections.Respondents were asked to respond to 14 questionsregarding the frequency of sleep disturbances (fouritems), headaches (three items), respiratory infections(four items), and gastrointestinal problems (threeitems) in the previous six months. Responses •weremade on a seven-point Likert-type scale rangingfrom 1 = not at all to 7 = all of the time. Scores foroverall physical health and for each health problemwere computed by averaging the responses, withhigher scores indicating greater severity. Changescores for each respondent were computed by sub-tracting his or her wave 2 physical health score fromhis or her scores in wave 3. A confirmatory factoranalysis conducted by Schat et al. (2005) empiricallydemonstrated the construct validity of the PHQ.For our sample, we obtained Cronbach's alphas of.80 for sleep disturbance, .92 for headaches, .86 forrespiratory infections, and .78 for gastrointestinalproblems at wave 2.

Burnout. Burnout was measured using theMaslach Burnout Inventory—Human Service Survey(MBI-HSS), a 22-item scale that conceptualizesburnout as having three dimensions: emotionalexhaustion, depersonalization or cynicism, anddiminished personal accomplishment (Maslach &Jackson, 1986; Maslach et al, 2001). Emotionalexhaustion (nine items) is related to the worker'sfeelings of being overextended and depleted ofemotional and physical resources. Depersonaliza-tion (five items) addresses the worker's negative orexcessively detached responses to various aspects ofthe job. Finally, diminished personal accomplishment(eight items) reflects the worker's feelings of incom-petence and lack of achievement at work (Maslach& Jackson, 1986). For each item, respondents areasked to report the extent of their experiences alonga seven-point continuum, ranging from 0 = never to6 = every day. Although all three burnout dimen-

sions are commonly used together in the literature.Acker (1999) found that social workers workingwith severely mentally ill chents were more likely toexperience emotional exhaustion and depersonaliza-tion but not diminished personal accomplishment.A recent study by Kim and Ji (2009) also showedthat burnout among social workers may be largelyexplained by emotional exhaustion and deperson-ahzation. In this study, findings from a longitudinalfactorial invariance test of the MBI-HSS suggestedthat personal accomphshment may be associatednot only with burnout, but also with professionaldevelopment. Therefore, for the present study, weonly used the 13 items for emotional exhaustionand depersonahzation to calculate a single burnoutscore for each participant, leaving out the personalaccomplishment items. For our sample, we obtainedCronbach's alphas of .91 for emotional exhaustionand .75 for depersonalization.

Control Variables. The analysis also controlled forother demographic and work-related characteristicsthat may affect burnout or physical health, includinga respondent's age, gender, field tenure, and annualsalary. Gender was coded as follows: Female = 1,and male = 0. Field tenure refers to the self-reportednumber of years a respondent had •worked as a social•worker. Annual salary •was based on a respondent'sself-reported income from his or her current job(at the time of the first survey). Age, field tenure,and annual salary were all included in the analysisas continuous variables.

AnalysisThe analysis consisted of t̂ wo stages. First, •we ex-amined the bivariate relationships between burnoutand physical health complaints. Following previousstudies (for example, Lau,Yuen, & Chan, 2005), re-spondents were evenly divided into three groups onthe basis of their baseline MBI-HSS scores: low [n= 92), moderate (« = 93), and high (n = 91). One-way analyses of variance (ANOVAs) were used totest whether the three burnout groups differed intheir subsequent physical health complaint scores(measured at wave 2). Second, using structuralequation modeling (SEM), we used path analysis totest whether basehne levels of burnout predictedchanges in physical health over time with controlvariables accounted for. As shown in Figure 1, wemodeled the change in the respondents' physicalhealth (from wave 2 to wave 3) as being predictedby their burnout levels at •wave 1, •while account-

260

Figure 1: A Path Model of Changes in Physical Health Complaints among Social Workers

Burnoutat wave 1

Annualsalary

Physical health:omplaints at wave 2

Gende

Years inthe field

Change in physicalhealth complaints(wave 3-wave 2)

ing for their physical health at wave 2. Baselinedemographic and •work-related characteristics werealso controlled for in the model.

The one-way ANOVAs were conducted usingSPSS 17.0, and the parameters for the path modelswere estimated •with Mplus 5.0 software (Muthén& Muthén, 2007). To estimate the path modelparameters, fuU-information maximum likelihoodestimation (FIML) was used to account for any miss-ing data (Arbuckle, 1996). FIML allows the estima-tion to proceed using all available data by breakingdown the likehhood function into componentson the basis of the patterns of missing data. SEMapproaches using FIML requires data with normaldistribution—absolute values of univariate skewnessindex greater than 3.0 and absolute values of uni-variate kurtosis index greater than 10.0 consideredas problematic (Kline, 1998). For the current study'sdata, univariate skewness values ranged from —1.48to 0.46, and kurtosis values ranged from -0.94 to7.01, so practically acceptable distributions wereassumed for all study variables.

Model goodness-of-fit was evaluated on the basisof several indices; the chi-square statistic divided bythe degrees of freedom (x^/dfj, the comparative fitindex (CFI), and the root mean square error of ap-

proximation (RMSEA). A %-/df value of less than3 •would indicate a reasonable fit (Kline, 1998). CFIvalues can range firom 0 to 1, with a value above0.90 suggesting an acceptable fit between the modeland the data (Kline, 1998). Finally, an RMSEA of.05 or less would indicate a good fit (Browne &Cudeck, 1993).

RESULTS

The characteristics of our sample are summarizedin Table 1. Among the 285 social workers in wave2, 65 percent were licensed clinical social workers(LCSWs), whereas 35 percent were associate socialworkers (ASWs [that is, registered in the state butnot yet licensed]). Almost half of the sample workedin the mental health field. At the time of the initialsurvey, the mean age of the sample was 46 years.Participants had worked in the social work fieldfor an average of 17.4 years and earned an averageannual income of $57,100. For the total sample, theaverage physical health complaints score decreasedmarginally from 39.6 in wave 2 to 39.5 in wave 3.

The means, standard deviations, and correlationcoefficients for the study variables are presented inTable 2. Physical health was positively associatedwith burnout experience (r = .50) but negatively

.:th among Social Workers: A Three- Year Longitudinal Study 261

Table 1: Description of the Sample

characteristic

Wave(W)

Sample

Gender (%)

Male

Female

License status (%)

Associate social worker

Licensed clinical social worker

Service field (%)"

School social work

Child welfare/family

Healthcare

Mental health

Mean age (in years)

Mean years in the field

Mean annual salary (in thousands of dollars)

Mean level of hurnout

Mean physical health complaints scores'"

Sleep disturbances

Headaches

Gastrointestinal prohlems

Respiratory infections

Overall physical health

Mean change in physical health (W3 – W2)

406 285 146

19.9

80.1

36.7

63.3

6.9

17.0

21.7

46.1

45.6 (12.0)

17.1 (10.6)

57.0 (17.0)

31.3 (14.3)

21.778.3

35.4

64.6

6.0

16.5

21.4

48.4

46.0(11.9)

17.4 (10.4)

57.1 (17.1)

29.9 (13.9)

13.3 (4.8)

9.1 (4.6)

10.1 (5.2)

7.1 (3.3)

39.6(13.3)

21.079.0

33.6

66.4

7.5

17.1

22.6

47.3

46.1 (11.3)

18.0 (10.0)

59.9 (16.1)

31.1 (14.3)

13.7 (4.5)

8.4 (4.4)

10.0 (5.1)

7.3 (3.3)

39.5 (12.9)

-0.8 (8.7)Notes: The sample for the present consisted of the 285 social workers who participated in the W2 survey. Where means are reported, standard deviations are included in paren-theses. Burnout scores were calculated by summing participants' responses to 13 questions on a seven-point scale (coded as 0 to 6), with total scores ranging from 0 to 78. Scoresfor overall physical health complaints were calculated by summing participants' responses to 14 questions on a seven-point scale (coded as 1 to 7), with total scores ranging from7 to 98. The scores for each physical health complaint were calculated in a similar manner, but their ranges varied on the basis of the number of items for each subscale: sleepdisturbances (four questions; scores ranged from 4 to 28); headaches (three questions; scores ranged from 3 to 21); gastrointestinal problems (three questions; scores rangedfrom 3 to 21); respiratory infections (four questions; scores ranged from 4 to 28).'Totals equal more than 100 percent because the participants were allowed to select multiple responses. Participants indicated their areas of praaice by using the categorizationsof service area developed for the NASW Center for Workforce Study (Whitaker, Weismiller, & Clark, 2006).''Physical health was not assessed at W l .

associated with age (r = —.21) and years in the field(r = -.25). In addition, burnout was significantly-correlated with age (r= -.19) and years in the field(r=-.16).

The one-way ANOVA results (see Table 3)showed significant relationships between burnoutand overall physical health and between burnout andeach individual health problem. In general, physicalhealth problems were the least severe among socialworkers with low burnout levels and the mostsevere among those with high burnout levels. Foroverall physical health, the mean score was the lowestfor social workers with low burnout levels (31.7),gradually increased for the moderate burnout group(39.7), and was even higher for the high burnoutgroup (47.8). Similar patterns were also found for

each health problem. Results from Tukey post-hocanalyses also revealed significant low versus moderateand moderate versus high group differences: Thehigh-burnout group was significantly worse thanthe moderate group, which, in turn, was significantlyworse than the low group, for overall physical healthand for each individual physical health problem.

The final path model, including only the signifi-cant parameters, is shown in Figure 2. We initiallyestimated all the parameters in our model and foundthat neither age nor annual salary significantly in-fluenced physical health at wave 2 or the change inphysical health from wave 2 to wave 3. To derivethe most parsimonious model, we removed the non-significant paths, but only if their removal did notsignificantly influence the overall model fit (on the

262

Variable

Table 2: Correlation Coefficients, Means, andStandard Deviations for Observed Variables

4

l.Wage

2. Gender

5.Ag,e

4. Years in the Field

5. Burnout at Wl

6. PHC (W2)

7. Change in PHC (W3 – W2)

M

SD

1.00

-.28*

.23*

.34*

.05

-.09

.13

57.07

17.12

—

1.00

-.23*

-.28

.01

-.17

-.01

0.78

0.41

—

1.00

.70*

-.19*

-.21*

-.03

46.02

11.88

—

1.00

-.16*

-.25*

.05

17.45

10.42

—

1.00

.50*

.05

30.96

14.36

1.00

-.34*

39.59

13.27

l.ÓO

-0.82

8.71Note: Scores for overall physical health complaints (PHC) were caiculated by summing participants' responses to 14 questions on a seven-point scale (1 to 7), v^ith totai scoresranging from 7 to 98. Scores for change in overall PHC were caiculated by subtracting PHC scores at wave 2 from PHC scores at wave 3. A positive change in PHC score indicatesan increase in PHC over time, w = wave,•p < .05.

basis of nonsignificant chi-square difference tests).This model trimming process is discussed in greaterdetail in KUne (1998).This process led to our finalpath model, which was a good fit to the data {"/^/df= 0.57, RMSEA = .00).This model explained 29percent of the variance in physical health at •wave2 and 18 percent of the variance in the change inphysical health fiom wave 2 to wave 3.

The path analysis results showed that burnout waspositively associated with both the initial physicalhealth complaints (ß = .47) and the change in physi-cal health complaints from wave 2 to wave 3 (ß =.29), even after initial physical health was controlledfor. In other words, higher levels of burnout led toworse physical health. Moreover, social workerswith higher levels of burnout experienced a greater

Table 3: Physical Health Complaints and Burnout amongSocial Workers: One-Way Analysis of Variance Results

Physical Health Complaints'

Sleep disturbances

M

SD

Headaches

M

SD

Gastrointestinal problems

M

SD

Respiratory infections

M

SD

Overall physical health

M

SD

Low (n = 92)

11.2

3.9

7.0

3.8

7.9

3.5

5.7

2.3

31.7

9.4

Level of Burnouf

Middle (n = 93)

13.3

4.3

9.34.1

10.4

4.9

7.0

3.0

39.7

10.6

High (n = 91)

15.6

5.0

11.1

5.0

12.1

5.9

8.8

3.7

47.8

UA

Hcffe) 123.4(2, 273)*

20.9(2, 271)*

"ÜB17.6(2, 272)*

"A23.8(2, 273)*

•m

44.1(2,272)*

mNotes; Scores for overall physical health complaints were calculated by summing participants' responses to 14 questions on a seven-point scale (coded as 1 to 7), with totalscores ranging from 7 to 98. Scores for each physical health complaint were calculated in a similar manner, but their ranges varied on the basis of the number of items for eachsubscale: sleep disturbances (four questions; scores ranged from 4 to 28), headaches (three questions; scores ranged from 3 to 21), gastrointestinal problems (three questions;scores ranged from 3 to 21), and respiratory infections (four questions; scores ranging from 4 to 28). All Tukey post-hoc tests for low versus middle and middle versus high levelsof burnout were significant at p < .05.'Assessed at wave 2."•Assessed at wave 1.' p < .01.

.^^.a/j among Social Workers: A Three-Year Longittidinal StuAy 263

Figure 2: Final Model of Changes in Physical Health Complaints among Social Workers

Burnoutat wave 1

Female(versus male)

.47*

Physical healthcomplaints at wave 2

– .50*

Change in physicalhealth complaints(wave 3—wave 2)

Years in the field

– .14*

Notes: Results are reported as standardized path coefficients (ßs). Model fit: x'(6, N = 285) = 3.44. Comparative fit index = 1.00, root mean square error of appro:(Pelóse = .94).•p < .05.

deterioration in physical health over tiine. Femalesocial workers were less healthy than their malecounterparts (ß = .12). More years in the field wasassociated with better initial physical health (ß =—.14). Additional path analyses (not shô wn) wereconducted for each separate physical health problemand showed that burnout was associated with signifi-cant increases in headaches (ß = .23),gastrointestinalproblems (ß = .20), and respiratory infections (ß =.19) but not •with sleep disturbances.

DISCUSSIONThis study revealed that burnout can adversely affectthe physical health of social workers, with higherlevels of burnout leading to more physical healthproblems one year after initial assessment. More im-portant, social •workers with higher levels of burnoutalso experienced a greater decHne in overall physicalhealth over a one-year period. Specifically, •we alsofound that social •workers with higher initial levels ofburnout reported more headaches, gastrointestinalproblems, and respiratory infections a year later.

Given the prevalence of burnout in the socialwork profession, these findings may have seriousimplications for social work practice. Health prob-lems may negatively affect the relationship between

social ^vorkers and their clients; more specifically,such problems may hinder them from developingnurturing alliances •with their chents. In addition,poorer physical health can lead to diminished jobperformance, including absenteeism and turnover(Darr & Johns, 2008) .Turnover among social work-ers has been found to negatively affect the quality,consistency, and stability of services (Mor Barak etal, 2001).

Although the current study focused on the physi-cal health symptoms of social workers, the potentiallinkages between physical and mental health cannotbe ignored. Schat et al. (2005) found that physicalhealth symptoms are significantly associated withpsychological health, as measured by the GeneralHealth Questionnaire (GHQ) (Banks et al., 1980).The GHQ is a widely used measure of mental healthin occupational settings and consists of items relat-ing to depression and self-confidence (Schat et al,2005).This further suggests that burnout can resultin poorer occupational •well-being for social workersin general, thus greatly compromising the qualityof services provided to clients.

Our findings support the need for both burnoutprevention and recovery interventions. To date,burnout intervention studies have primarily focused

264 V 2011

on preventive efforts. For example. Maslach et al.(2001) emphasized educational interventions toenhance the ability of workers to cope with stress.Similarly, S. H.Jones (2007) argued that social workeducation programs should increase awareness ofburnout symptoms and teach students about strat-egies to prevent burnout, such as communicationtechniques and coping skills. Organizations cantake more proactive steps to address the burnoutissue (for example, increasing employee awarenessof burnout, its symptoms, and its effects; developingpreventive strategies).

In particular, managers and supervisors must playa critical role in supporting their staff and preventingburnout. Research has shown that supervisor supportand performance feedback are essential to preventingworker burnout (Bakker, Demerouti, & Euwema,2005). Supervisors who are open and responsiveto the opinions of fronthne social workers abouttheir job-specific problems can help those workersto cope with job demands (Kim & Lee, 2009). In asimilar vein, Himle andjayaratne (1991) found thatinstrumental support from a supervisor helped tobuffer the influence of job stress on burnout. Directsupervisors and managers may play an important"bridging" role between their staff and agency ad-ministrators, helping to identify burnout symptomsamong their staff and communicating these difficul-ties to administrators. Relatedly, effective supervisionhas been identified as a key factor in the retentionof social workers, pubHc child welfare workers, andhuman service workers in general (Chenot,Benton,& Kim, 2009; Mor Barak,Travis, Pyun, & Xie, 2009;Rycraft, 1994).

In addition to supervisor support, support fromcoworkers, peers, and spouses have been shown tobe critical in helping to prevent burnout amongsocial workers. For example, Himle and Jayaratne's(1991) study showed that informational supportamong coworkers softened the effects of role conflictand workload on the risk of emotional exhaustionamong social •workers. Davis-Sacks, Jayaratne, andChess (1985) found that increased spousal supportmay buffer the risk of depersonalization amongfemale child welfare workers. Future research couldalso focus on the potential role of familial andcommunity support in alleviating burnout amongsocial workers.

Also critical, but less understood, are burnoutrecovery interventions. Social ^vorkers are argu-ably at a high risk of burnout because of intense

worker—client interactions and unmanageable case-loads. Therefore, to mitigate any potentially adverseeffects of burnout, recovery interventions wouldbe necessary to assist social workers •who have orare experiencing burnout and to help them copewith their current burnout experiences. Again, ad-ministrators and supervisors may play an importantrole in supporting and empowering social workersto sustain their commitment to the job. Self-helpgroups or social networks may also provide socialworkers with spaces in which to share their experi-ences and support each other. Easily implementedscreening tools and interventions could be devel-oped and targeted to social workers in settings thatmay pose the greatest risk of burnout (for example,CPS workers, workers serving severely mentally iUpopulations) .The feasibihty and cost-effectiveness ofpotential burnout recovery programs are importanttopics for fiiture research.

Our results also showed that number of yearsin the field was negatively associated with physicalhealth complaints, suggesting that entry-level orearly-career social •workers are particularly at risk.This is consistent •with earlier studies, which haveshown that human service workers who are begin-ning their careers are more hkely to experienceburnout (Maslach et al., 2001). Moreover, on thebasis of an extensive review on the burnout litera-ture, Schaufeli and Enzmann (1998) concluded thatburnout, if not addressed early on, tends to persist.Similarly, Poulin and Walter (1993b) conducteda longitudinal survey of 879 social workers andfound that burnout is a stable phenomenon amongprofessional social workers. Therefore, preventionand recovery interventions in the early stages of asocial worker's career (including while he or she isa student) could help to improve his or her long-terni health (and, potentially, career) trajectories.Both baccalaureate- and graduate-level social workprograms—via their curriculums, alumni networks,and continuing education programs—could assiststudents and recent graduates in handhng burnoutearly in their careers and, thus, protect their healthin the long term.

Gender emerged as another significant predictorof physical health complaints, with female socialworkers reporting significantly worse physicalhealth than their male counterparts. This genderdifferential is consistent with findings in the lit-erature (Haug, Mykletun, & Dahl, 2004) and mayhave several explanations. For instance, it is possible

•* among Social Workers: A Three-Year Longitudinal Study 265

that women have poorer health and exhibit moresymptoms than do men, resulting in their report-ing more symptoms (Barsky, Goodson, Lane, &Cleary, 1988). Another possible explanation is thehigher prevalence of anxiety and depression foundamong women, which has been shown to be as-sociated with increased reporting of physical healthproblems (Haug et al.,2004;Wool & Barsky, 1994).Considering that the majority of social workers arefemale, this finding has significant imphcations.

It is interesting to note that age •was found tobe a negative determinant of physical health in thebivariate analysis but not in the path analysis. Age hasbeen consistently found to be a negative correlate ofburnout (Brewer & Shapard, 2004), so one possibleexplanation for the negative relationship betweenage and physical health is that younger (and oftenless experienced) workers are more susceptible toburnout and, thus, exhibit poorer physical health.For example, in an extensive reviê w of the litera-ture on burnout and cardiovascular-related events,Melamed et al. (2006) found that the relative riskassociated with burnout was equal to, and sometimeseven exceeded, the risk associated •with classical riskfactors such as age, smoking, and body mass index.

The present study has several key limitations. First,the analysis was based on a sample of registered socialworkers in California, which may Hmit the gener-alizability of the findings to social •workers in otherparts of the United States. Nationally representativestudies could further advance our understanding ofthe association between burnout and physical healthin social work. Second, our findings may also bebiased due to sample attrition. We tried to addressthis concern •with a preliminary attrition analysis,which determined that the attrition •was unrelated toour key study variables. In addition, by using FIMLto estimate our models, •we were able to includeinformation for the entire sample—including thoseparticipants who left the study.Third, we were onlyable to follow social workers for three years, collect-ing information at one-year intervals. It is possiblethat three years is not sufficiently long enough aperiod to assess changes in workers' physical health.Furthermore, the participants' initial physical healthstatus was not collected, so information about theirphysical health was only available for two time points.Finally, the current study did not account for a broadarray of additional factors that may affect a person'shealth, such as health behaviors, life events or Stres-sors, genetic influences, and accidents.

Nonetheless, this study presents compellingevidence of the significant impact of burnout onphysical health in the social work profession. Thefindings also highlight the critical need for socialworkers to pay attention to their own occupationalwell-being and, more broadly, the need for thesocial work profession to focus more resources onunderstanding and addressing the impact of burn-out. To date, our understanding of burnout and itspotential health consequences for social workershas been hmited. Moreover, there is a tremendousneed for research on effective burnout interventions.Future research should explore how physical healthmay affect the job performance of social •workers,including their decision making and job turnover.At the same time, •we need to identify factors thatmay ameliorate the adverse effects of burnout onphysical health among social workers.

Finally, this study highlights the critical need forthe social •work profession and all of its stakehold-ers (for example, members, educational programs,service agencies) to deal proactively •with burnout.Social workers are among the ^vorkers most at riskwhen it comes to experiencing burnout at somepoint in their careers (for example, Priebe, Fakhoury,Hoffmann, & Powell, 2005; Siebert, 2005). As ourfindings suggest, burnout can lead to a worseningof one's health over time, which, in turn, can com-promise the quality of the services one providesto chents.

According to the NASW (2008) Code of Ethics,social workers are called on to promote social justiceand social change by addressing social inequalitythrough the empo^werment of clients, promotion ofcultural diversity, and resistance to social injustice. Inparticular, social •workers are committed to address-ing the needs of vulnerable and underprivilegedchildren and families. However, in the process,social workers find themselves at risk in terms oftheir health and well-being. If action is not taken tocounteract the prevalence of burnout and its impacton the psychological and physical well-being of itsmembers, our profession •wül continue to struggleto sustain a strong and vibrant pool of social work-ers. Therefore, it is critical to build healthy workingenvironments in which the next generation of socialworkers can meet the emerging social challenges ofthe 21st century.

One such current effort is the Dorothy I. HeightandWhitney M.Young,Jr. SocialWork ReinvestmentAct (H.R. 795 and S. 686), which is comprehensive

266 LY 2OII

federal legislation aimed at promoting recruitment,retention, research, and reinvestment in the profes-sion of social work (see NASW, 2009). If passed, itwould help to improve the working conditions ofprofessional social workers. For example, it ^wouldfund demonstration programs on improving theworkplace, addressing, among other conditions, highcaseloads, fair market compensation, social workersafety, supervision, and working conditions—all ofwhich are significant factors in burnout and turnoveramong social workers (Chenot et al., 2009; Kim &Stoner, 2008; Mor Barak et al., 2001,2009; Siebert,2005). If we are unable to find ways to protect theoccupational well-being of the current workforce,worker shortages wül only become more serious.Our hope is that this study will help to inform thedevelopment and the enactment of policy measures,such as the Social Work Reinvestment Act, focusedon securing both federal and state investment in thesocial work profession.

REFERENCESAcker, G. M. (1999).The impact of clients' mental illness

on social workers'job satisfaction and burnout.Health & Social Work, 24, 112-119.

Anderson, D. G. (2000). Coping strategies and burnoutamong veteran child protection workers. Child Abuse& Neglect, 24, 839-848.

ArbuckleJ. L. (1996). Full information estimation in thepresence of incomplete data. In G. A. Marcoulides& R. E. Schumacker (Eds.), Advanced structural equa-tion modeling (pp. 243-277). Mahwah, NJ: LawrenceErlbaum Associates.

Bakker,A. B., Demerouti, E., & Euwema, M. C. (2005).Job resources buffer the impact of job demands onburnout. JoMmu/ of Occupational Health Psychology, 10,170-180.

Banks, M.J., Clegg, C.W.,Jackson, P. R., Kemp, N.J.,Stafford,E. M., &Wall,T. D. (l980).The use of theGeneral Health Questionnaire as an indication ofmental health in occupational settings.JoMmal ofOccupational Psychology, 53, 187-194.

Barsky, A. J., Goodson, J. D., Lane, R. S., & Cleary, P. D.(1988).The amplification of somatic symptoms.Psychosomatic Medicine, 50, 510-519.

Brewer, E.W., & Shapard, L. (2004). Employee burnout:Ameta-analysis of the relationship between age or yearsof experience. Human Resource Development Review, 3,102-123.

Browne, M.W, & Cudeck, R. (1993). Alternative waysof assessing model fit. In K. A. Bollen & J. S. Long(Eds.), Teaching structural equation models (pp. 136—162).Newbury Park, CA: Sage Publications.

Burke, R.J., & Mikkelsen,A. (2006). Burnout amongNorwegian police officers: Potential antecedentsand consequences. International Journal of StressManagement, 13, 64—83.

Chenot, D., Benton, A. D., & Kim, H. (2009).The influ-ence of supervisor support, peer support, and organi-zational culture among early career social workers inchüd welfare services. Child Welfare, 88, 129-147.

Darr,W;, & Johns, G. (2008).Work strain, health, andabsenteeism: A meta-analysis. JoHma/ of OccupationalHealth Psychology, 13, 293-318.

Davis-Sacks, M. L.,Jayaratne, S., & Chess,W. A. (1985).Acomparison of the effects of social support on theincidence of burnout. Social Work, 30, 240—244.

Edéll-Gustafsson, U M., Kritz, E.I.K., & Bogren, I. K.(2002). Self-reported sleep quality, strain and health inrelation to perceived working conditions in females.Scandinavian Journal of Caring Sciences, 16, 179—187.

Eriksen, H. R., Svendsrsd, R., Ursin, G., & Ursin, H.(1998). Prevalence of subjective health complaintsin the Nordic European countries in 1993. EuropeanJournal of Public Health, 8, 294-298.

Evans, S., Huxley, P, Gately, C,Webber, M., Mears, A.,Pajak, S., et al. (2006). Mental health, burnout and jobsatisfaction among mental health social workers inEngland and Wales. British Journal of Psychiatry, 188,75-80.

Ferraro, K., & Nuriddin,T. A. (2006). Psychological distressand mortality: Are women more vulnerable? Jowma/ ofHealth and Social Behavior, 47, 227-241.

Gove,W., & Hughes, M. (1979). Possible causes of theapparent sex differences in physical health: An em-pirical investigation. American Sociological Review, 44,126-146.

Grossi, G., Perski, A., Evengard, B., Blomkvist,V., & Orth-Gomer, K. (2003). Physiological correlates of burnoutamong women. Jowma/ of Psychosomatic Research, 55,309-316.

Haug,T.T., Mykletun, A., 6f Dahl, A. A. (2004).The as-sociation between anxiety, depression, and somaticsymptoms in a large population: The HUNT-II study.Psychosomatic Medicine, 66, 841—851.

Himle, D. P., & Jáyaratne, S. (1991). Buffering effects offour social support types on burnout among socialworkers. Social Work Research, 27, 22—27.

Honkonen,T., Ahola, K., Pertovaara, M., Isometa, E.,Kalimo, R., Nykyri, E., et al. (2006).The associationbetween burnout and physical illness in the generalpopulation—Results fi-om the Finnish Health 2000Study. Jowma/ of Psychosomatic Research, 61, 59—66.

Johansson, G.,Johnson,J.V., & Hall, E. M. (1991). Smokingand sedentary behavior as related to work organiza-tion. Social Science & Medicine, 32, 837—846.

Jones, F., & Bright,J. (2001). Stress: Myth, theory and research.London: Prentice Hall.

Jones, S. H. (2007,January). Secondary trauma and burnoutin child protective workers: Implications for preparationof social workers. Paper presented at the 11th AnnualConference of the Society for Social Work andResearch, San Francisco.

Kim, H., & Ji,J. (2009). Factor structure and longitudi-nal invariance of the Maslach Burnout Inventory.Research on Social Work Practice, 19, 325-339.

Kim, H., & Lee, S.Y. (2009). Supervisory communication,burnout, and turnover intention among social work-ers in health care settings. Social Work in Health Care,48, 364-385.

Kim, H., & Stoner, M. (2008). Burnout and turnoverintention among social workers: Effects of role stress,job autonomy and social support. Administration inSocialWork,32{3), 5-25.

Kline, R. B. (1998). Principles and practice of structural equationmodeling. New York: Guilford Press.

Lau, PS.Y,Yuen, M.T, & Chan, R.M.C. (2005). Dodemographic characteristics make a difference toburnout among Hong Kong secondary school teach-ers? Social Indicators Research, 71, 491-516.

Lloyd, C , & King, R. (2004). A survey of burnout amongAustralian mental health occupational therapists

Í-3 among Social Workers: A Three-Year Long.tudinal Study 267

and social workers. Social Psychiatry and PsychiatricEpidemiology, 39, 752-757.

Maslach, C , & Jackson, S. E. (1986). Tiie Masiacii BurnoutInventory-Human Service Survey researcii edition manual.Palo Alto, CA: Consulting Psychologist Press.

Maslach, C , Schaufeh,W. B., & Leiter, M. P. (2001). Jobburnout. Annual Review of Psyciioiogy, 52, 397-422.

Melamed, S., Shirom, A.,Toker, S., Berhner, S., &L Shapira,I. (2006). Burnout and risk of cardiovascular disease:Evidence, possible causal paths, and promising re-search directions. Psyciiological Buiietin, 132, 327-353.

Mohren, D.C.L., Swaen, G.M.H., Kant, I., van Amelsvoort,L.G.P.M., Borm, PJ.A., & Galama,J.M.D. (2003).Common infections and the role of burnout in aDutch working population. Jouriíii/ of PsyciiosomaticResearch, 55, 201-208.

Mor Barak, M. E., Nissly,J. A., & Levin, A. (2001).Antecedents to retention and turnover amongchild welfare, social work, and other human serviceemployees: What can we learn from past research?A review and meta-analysis. Social Service Review, 75,625-662.

Mor Barak, M. E.,Travis, D., Pyun, H., & Xie, B. (2009).The impact of supervision on worker outcomes: Ameta-analysis. Social Service Review, 83, 3—32.

Muthén, L. K., & Muthén, B. (2007). Mpius user's guide(version 5.0). Los Angeles: Author.

Nakamura, H., Hágase, H.,Yoshida, M., & Ogino, K.(1999). Natural kiUer (NK) cell activity and NKcell subsets in workers with a tendency of burnout.Journal of Psyciiosomatic Researcii, 46, 569—578.

National Association of Social Workers. (2008). Codeof ethics of the National Association of Social Workers.Retrieved from http://www.naswdc.org/pubs/code/code.asp

National Association of Social Workers. (2009). DorotliyI. Height and Wiiitney M.Young Jr. Social WoriiReinvestment Act Summary. Retrieved from http://www.sociaWorkreinvestment.org/Content/l 11th/Factsheet2009.pdf

Oktay, J. S. (1992). Burnout in hospital social workers whowork with AIDS patients. Social Work, 31, 432-439.

Peterson, U., Demerouti, E., Bergstrom, G., Samuelsson,M., Asberg, M., & Nygren, A. (2008). Burnout andphysical and lnental health among Swedish healthcareworkers.JoMma/ of Advanced Nursing, 62, 84—95.

Poulin,J., &Walter, C. (1993a). Burnout in gerontologicalsocial work. Social Work, 38, 305-310.

Poulin, J., & Walter, C. (1993b). Social worker burnout:A longitudinal study. Social Work Research & Abstract,29('^),b-n.

Priebe, S., Fakhoury,W.K.H., Hofihiann, K., & Powell,R. A. (2005). Morale and job perception of com-munity mental health professionals in Berlin andLondon. Social Psychiatrj' and Psychiatric Epidemiology,40, 223-232.

Rycraft,J. R. (1994).The party isn't over:The agency rolein the retention of public child welfare caseworkers.SocialWork,39,lb-?,0.

Schat, A.C.H., Kelloway, E. K., & Desmarais, S. (2005).ThePhysical Health Questionnaire (PHQ): Constructvalidation of a self-report scale of somatic symptoms.Journal of Occupational Health Psychology, 10, 363—381.doi:10.1037/1076-8998.10.4.363

Schaufeh,W B., & Enzmann, D. (1998). The burnout com-panion to study and practice: A critical analysis. Loridon:Taylor & Francis.

Schwartz, B. S., Stewart,W. F., Simon, D., & Lipton, R. B.(1998). Epidemiology of tension-type headache.JAMA, 219, 381-383.

Siebert, D. C. (2004). Depression in North Carolina socialworkers: Implications for practice and research. SocialWork Research, 28, 30-40.

Siebert, D. C. (2005). Personal and occupational fac-tors in burnout among practicing social workers:Implications for researchers, practitioners, and man-gers.JoMma/ of Social Service Research, 32(2), 25—44.

Soai-es,JJ.H, Grossi, G., & Sundin, Ö. (2007). Burnoutamong women: Associations with demographic/socio-economic, work, hfe-style and health factors.Archives of Women's Mental Health, 10, 61-71.

Söderström, M., Ekstedt, M., Âkerstedt,T., Nilsson, J., &Axelsson,J. (2004). Sleepiness in young individualswith high burnout scores. Sieep, 21, 1369-1377.

Stanley, N., Manthorpe, J., & White, M. (2007). Depressionin the profession: Social workers' experiencesand perceptions. British Journal of Social Work, 31,281-298.

Stanley,T. L. (2004). Burnout: A manager's worst night-mare. Supervision, 65(b), 11—13.

Sveinsdóttir, H., Gunnarsdóttir, H., & Riksdóttir, H.(2007). Self-assessed occupational health and work-ing environment of female nurses, cabin crew andteachers. Scandinavian Journal of Caring Sciences, 21,262-273.

Toker, S., Shirom, A., Shapira, I., Berliner, S., & Melamed, S.(2005).The association between burnout, depression,anxiety, and inflammation biomarkers: C-reactiveprotein and fibrinogen in men and v/omen. Journal ofOccupational Health Psychology, 10, 344—362.

Whitaker,T,Weismiller,T., & Clark, E. (2006). Assuringtite sufficiency of a frontline workforce: A national study oflicensed social workers: Executive summary. Washington,DC: National Association of Social Workers.

Wool, CA. , & Barsky,A.J. (1994). Do women somatizemore than men? Gender difference in somatization.Psychosomatics, 35, 445-452.

Hansung Kim, PhD, is assistant professor. Department ofSociology, Hanyang University, 222 Wangsimni-ro, Seongdong-gu, Seoul 133-191, Korea; e-mail: [email protected], PhD, is assistant professor. School of Social Work, SyracuseUniversity, Syracuse, NY. Dennis Kao, PhD, is assistantprofessor. College of Social Work, University of Houston. Thisresearch was supported in part by a research award from theHamovitch Center for Science in the Human Services, an Albertand Frances Feldman Endowed Fellowship from the Univer-sity of Southern California School Social Work, and a facultydevelopment grant from California State University, FuUerton.The authors thank all of the social workers who participated inthis longitudinal study for their patience and support. An earlierversion of this article was presented at the Í2th annual confer-ence of the Society for Social Work and Research, January 19,2008, Washington, DC.

Original manuscript received January 5, 2009Final revision received October 13, 2009Accepted March 25, 2010

268 JLY 2OII

Copyright of Social Work is the property of National Association of Social Workers and its content may not be

copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written

permission. However, users may print, download, or email articles for individual use.