Journal of Vocational Behavior 73 (2008) 242–253

Contents lists available at ScienceDirect

Journal of Vocational Behavior journal homepage: www.elsevier.com/locate/jvb

A model of career success: A longitudinal study of emergency physicians q
Sarah Pachulicz, Neal Schmitt *, Goran Kuljanin
Department of Psychology, Michigan State University, 262 Psychology, E. Lansing, Michigan 48824-1116, USA

a r t i c l e

i n f o

a b s t r a c t
Objective and subjective career success were hypothesized to mediate the relationships between sociodemographic variables, human capital indices, individual difference variables, and organizational sponsorship as inputs and a retirement decision and intentions to leave either the specialty of emergency medicine (EM) or medicine as output variables. Objective career success operationalized as the number of leadership positions held did not mediate the relationship, but income change and career satisfaction mediated the relationship between the hours worked and years employed in emergency medicine. Work centrality was significantly related to subjective career success more so for men than women and perceptions of success or self-efficacy were positively related to subjective career success for women, but not for men. The expected pattern of women indicating more difficulties with personal time and family time did not emerge; but women did indicate less perceived support from the organization, fewer EM leadership positions, less perceived control over their work situation and less organizational support than did men. Ó 2008 Elsevier Inc. All rights reserved.

Article history: Received 22 January 2008 Available online 12 June 2008

Keywords: Career success Emergency physicians Salary change Career satisfaction Leisure and career success Work centrality Gender difference in careers

1. Introduction Career success has been of interest to both individuals and organizations. For individuals, a sense of career success has been related to life satisfaction as well as general mental health (Hall, 1976, 2002). For organizations, the prevailing view is that individual career success is to some extent coincident with organizational success as well (Hall, 2002; Judge, Higgins, Thoresen, & Barrick, 1999). Consequently, both the determinants and the outcomes associated with career success have been of interest to researchers. Recently, Ng, Eby, Sorensen, and Feldman (2005) have meta-analyzed the data regarding predictors of career success. They considered career success from two vantage points. Objective career success or mobility was indexed by salary level and promotional advancement. Subjective career success was measured by self-reports of career satisfaction. In this paper, we examine the same issues for individuals who are engaged in a particularly demanding career (i.e., emergency medicine) in which conflict can and does arise between family concerns, personal and leisure activities, and demands related to careers. 1.1. Overview of Ng et al. meta-analysis Ng et al. (2005) considered both subjective and objective indices of career success. Subjective career success is a person’s subjective judgment about her career attainments such as career satisfaction. On the other hand, objective career success is measured by extrinsically observable factors, and includes salary progression and promotions. According to Ng et al. (2005),

q We acknowledge the financial support of the American Board of Emergency Medicine and the considerable effort of Dr. Mary Ann Reinhart and her colleagues in the development of the survey used in this project and for their many helpful comments on our work. * Corresponding author. Fax: +1 517 432 2476. E-mail address: schmitt@msu.edu (N. Schmitt).

0001-8791/$ - see front matter Ó 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.jvb.2008.05.003

S. Pachulicz et al. / Journal of Vocational Behavior 73 (2008) 242–253

243

the components of career success can be influenced by four categories of predictors: Human capital, organizational sponsorship, sociodemographic status, and stable individual differences. Human capital is comprised of an individual’s educational, personal, and professional experiences. Across different occupations, it may be comprised many different indicators including the number of hours worked, work centrality (job involvement or the psychological investment in work or centrality of work for self-identity or self-image), job tenure, organization tenure, work experience (total years in the workforce), willingness to transfer, international work experience, education level, career planning (‘‘I have a strategy for achieving my career goals,” ‘‘I have a plan for my career”), political knowledge and skills, and social capital (quantity and quality of accumulated contacts, quantity of people an employee knows of in other functions or at higher levels, and the extent to which an employee engages in networking activities). Organizational sponsorship includes career sponsorship (the extent to which employees received sponsorship from individuals within the organization, including senior managers and mentors). This included the self-reported career-enhancing functions of being assigned challenging tasks, obtaining exposure and visibility, receiving protection, sponsorship, and coaching, supervisor support (extent to which supervisors provide emotional and work related support), training and skill development opportunities (self-reported perceptions of the extent to which their company provided opportunities for training and skill acquisition), and organizational resources (measured by organization size, number of employees in organization). Sociodemographic predictors include demographic and social background, such as gender, race, marital status, and age. Stable individual differences factors are personality factors including the Big Five factors (i.e., neuroticism, conscientiousness, extroversion, agreeableness, and openness) as well as factors such as proactivity, locus of control, and cognitive ability. Human capital and sociodemographic predictors were found to have stronger relationships with objective success, and organizational sponsorship and stable individual differences had stronger relationships with subjective career success. Gender and time were found to be moderators of some relationships. 1.2. Medical studies of career success In the medical literature, studies of career success have been almost totally focused on career satisfaction and often treated satisfaction not as an outcome variable, but rather as a predictor of another variable such as intention to leave medicine or intention to reduce the hours of one’s practice. However, several studies of medical personnel have addressed the correlates of career success. Number of hours worked was consistently found to be related to satisfaction, such that the higher the number of hours worked, the lower the satisfaction (Frank, McMurray, Linzer, & Elon, 1999; Landon, Reschovsky, & Blumenthal, 2003; Landon, Reschovsky, Pham, & Blumenthal, 2006). However, Williams et al. (2001) found that the dissatisfaction with one’s job and career were not related when individuals reported working under 60 h a week, but dissatisfaction was significantly greater when physicians reported working over 60 h a week. This raises the question of potential curvilinearity between the hours worked and satisfaction. In addition, the effect of number of hours worked could be perceived differently according to specialty; for example, Lepnurm, Danielson, Dobson, and Keegan (2006) found that the number of hours worked per week was a negative predictor of satisfaction for psychiatrists, but a positive predictor of satisfaction for surgeons. Glisson and Durick (1988) measured variables called task identity and task significance, which were similar to Ng et al. (2005) work centrality variable. Statistically significant (p < .01) correlations between task identity, task significance, and career satisfaction were .44 and .47, respectively (both ps < .001). Organization tenure was not found to correlate with career satisfaction (Glisson & Durick, 1988; Lepnurm, Dobson, Backman, & Keegan, 2006), and neither were work experience and educational level (Glisson & Durick, 1988). These relatively objective indicators of human capital were not related to subjective career success. Evidence in the medical literature for the contribution of organizational resources is mixed. Glisson and Durick (1988) used a variable called workgroup budget that did not correlate with satisfaction. Lepnurm, Dobson, Backman, and Keegan (2006b) determined that access to and quality of the health system in which the physicians worked did contribute to career satisfaction. While these two variables appear to be organizational resources, their operationalization is quite different from that used in Ng et al. (2005), where organizational resources was simply defined as size of the organization. A few medical researchers have documented the role of sociodemographic variables. Frank et al. (1999) reported that marital status was not significantly related to satisfaction; but this study only included women physicians. Age was found to be positively related to satisfaction (Frank et al., 1999; Glisson & Durick, 1988), and gender was related to satisfaction such that males generally were more satisfied than females (Glisson & Durick, 1988; Lepnurm et al., 2006b). Locus of control was also consistently found to be related to satisfaction (Frank et al., 1999); for example, primary care physicians who became owners of their practice reported increased satisfaction (Landon et al., 2003); female physicians who perceived more control over their work situation were more satisfied (Robinson, 2004); and individuals who perceived they had a high level of influence over decisions felt more satisfied (Lepnurm et al., 2006b). Some variables that are mentioned frequently in the medical literature were not included in the Ng et al. (2005) metaanalysis. For example, Lepnurm et al. (2006b) found that self-reported health predicted 13% of the variance in career satisfaction for psychiatrists, and 18.9% of the variance in career satisfaction for surgeons. Similarly, Williams et al. (2001) reported that physical health was correlated .26 with satisfaction, and mental health correlated .56 with satisfaction (p < .001).

244

S. Pachulicz et al. / Journal of Vocational Behavior 73 (2008) 242–253

Rather than using income as part of career success, Landon et al. (2003) treated it as a correlate (implicitly a determinant) of career satisfaction. For medical specialists, income was not related to satisfaction, but for primary care physicians, satisfaction was significantly related to income. In addition, an increase in the level of patient acuity (perhaps an indirect measure of stress) that physicians were expected to handle without referral was related to decreased satisfaction. In a similar vein, Frank et al. (1999) reported that income was related to higher satisfaction, and that work stress was related to lower satisfaction. Lepnurm, Danielson, Dobson, and Keegan (2006a) used some of the same constructs as included in the Ng et al. (2005) meta-analysis but applied a different analytical approach. They conducted a survey and computed loadings of their survey items on four factors of career satisfaction that they labeled personal, professional, performance, and inherent. Control of work schedule loaded on the personal factor; relationship with administrators, relationship with nurses, authority to get your decisions carried out, and earnings loaded on the professional factor; access to resources to treat patients and ability to keep up with advances in one’s specialty loaded on a performance factor; and interactions with other physicians and career advancement loaded on an inherent factor. All of these variables were considered in past research (Ng et al.), but Lepnurm et al.(2006a, 2006b) did not assess the relationship between these variables and overall career satisfaction or salary or promotion outcomes. Interestingly, personality variables such as the Big Five received no mention in the medical literature at all though they have been the focus of many studies in the broader literature on career success. The only individual difference variable that received some attention was perception of locus of control. The career success literature, especially as it relates to success in medical professions, leaves several unanswered questions, some of which we try to address in this paper. First, we consider both objective and subjective measures of career success and its correlates. Promotions in a medical career are likely different than in a traditional business organization. In medical careers, a promotion likely means some change in one’s job to include more administrative than medical tasks which may actually decrease career satisfaction for those heavily invested in the practice of emergency medicine. On the ‘‘input” side, objective human capital measures likely include specialty certifications of some type since all will have medical degrees. Second, we examine the correlates of career success over a span of 10 years. While almost required conceptually in career success studies, longitudinal studies of career success (Bray, Campbell, & Grant, 1974) especially among medical professions, are rare. Use of cross time measures also reduces the common method explanation of observed relationships when measures are collected by self-report. Third, we examine the role of involvement in leisure activities in career success studies. While considered an effective source of relief or buffer of the stress and time required of people in demanding professions, very few studies have considered the role of leisure activities as correlates of career success. ‘‘Excessive” involvement in leisure activities may actually be a detriment to objective career success indicators while opportunities to engage in leisure activities may contribute to one’s satisfaction in a career. Fourth, like Ng et al. (2005), we examine the degree to which observed relationships are moderated by gender. In this sample of emergency medical physicians, we believe that gender will be more important as a moderator of relationships with career success because of the demanding nature of the job and the fact that responsibilities for family and children often fall disproportionately on women. Fifth, we go beyond the Ng et al. meta-analysis in considering some outcomes of career success (or the lack thereof) such as retirement, leaving emergency medicine, or intending to leave the profession. Finally, we consider some individual difference correlates of career success that have been largely neglected in the medical literature. 1.3. Hypotheses Consistent with the meta-analysis reported above, the medical literature on career success, and the situation faced by the emergency medical physicians that are the subject of our investigation, we propose the following hypotheses. Hypothesis 1. Four human capital indices (number of hours worked, number of years working in emergency medicine, number of special certifications received, and the level of self-reported work excitement) are related positively to objective career success. Measures of these variables were collected at the second of three data collection efforts separated by 5-year intervals since we wanted to assess the degree to which the presence/absence of human capital would lead to potential career-ending decisions as well as career success. Using an earlier measure of human capital seemed inappropriate as there would likely be important changes in human capital in a 5-year period that should be reflected in the relationship. In addition, all human capital variables and the objective career success indicators with the exception of the excitement variable (see description below) are relatively objective and should not be subject to the usual common method bias criticism of self-report variables. Hypothesis 2. Sociodemographic variables (age, race, and marital status) are related positively to objective career success variables. These variables, all collected in the third survey, are unlikely to change with the possible exception of marital status. Since marital status was collected at Time 3, any attributions concerning the direction of causality are unwarranted. Hypothesis 3. Individual difference variables (planning, sociability, self-efficacy or success, health, leisure activity will be related positively to subjective career success. Personal conflict will be negatively related to subjective career satisfaction.

S. Pachulicz et al. / Journal of Vocational Behavior 73 (2008) 242–253

245

Measures of planning, sociability, and self-efficacy were collected at all three time points. We used the first set of available responses, because these variables were conceived of as stable individual difference variables that should precede and contribute to the subjective appraisals of one’s career. Measures of health, leisure activity, and level of personal conflict were collected simultaneous with the measure of subjective career success at the second data collection since it was felt that these variables would change with time and that the most recent measure of these variables would be most relevant to perceptions of career success. Hypothesis 4. Organizational sponsorship variables including organizational support items (compensation, benefits, job security, a sense of ownership, opportunity for specializations), further educational opportunities, and perceptions of control over one’s work environment are related positively to subjective career success. The input variables described in this hypothesis were measured in the Time 2 survey on the thesis that the effects of such factors would take place over time, but that they would be relatively proximal in their effects on subjective career success. Hypothesis 5. Objective career success is related positively to career outcomes such as leaving medicine, intentions to leave emergency medicine, and retirement. Objective career success (number of academic and emergency medicine leadership positions held, salary change over the 10-year period) and the outcome measures were all assessed at Time 3. We felt the objective career success variables should include all events including recent ones that occurred prior to or simultaneously with outcomes. Since all these items were relatively objective, the likelihood of common response bias should be minimal. The three outcome variables are not conceptually related and, in fact, empirical correlations between the three ranged from .08 to .21. Hence each outcome was treated separately in our analyses rather than considering each as indicators of an ‘‘outcome” construct. Hypothesis 6. Subjective career success is related to career outcomes such as intentions to leave medicine, intentions to leave emergency medicine, and retirement. Subjective career success was measured at Time 2 and outcome variables were taken from the Time 3 survey. Hypothesis 7. Relationships between variables are moderated by gender. Both the Ng et al. (2005) meta-analysis and various medical studies of career success suggest that the hypothesized relationships above are moderated by gender. There is a general lack of specificity as to which of these relationships will be moderated; hence, the examination of this hypothesis is largely exploratory. However, we would surmise that the relationship between the marital status variable and objective career success may vary as a function of gender. Given the usual societal roles associated with marital relationships, we would expect marriage to increase the responsibilities of women and decrease the career success they experience. For men, a spouse may serve a more supportive role and increase their career success. It may also be the case that personal-work conflicts will play a greater role in the career success of women than men. The first six hypothesized relationships are depicted in Fig. 1; Hypothesis 7 suggests the relationships depicted in this figure will vary as a function of gender. 2. Method 2.1. Sample and procedures Details of the sampling procedure were described in Reinhart, Munger, and Rund (1999). Representative emergency physicians (EPs) in four cohorts (as represented by the year they completed a residency from 1979 to 1993) were selected. These EPs were then contacted and invited to become members of the study for their lifetimes. Future cohorts were identified and invited to participate in the 1999 and 2004 survey with the stipulation that all must be graduates of professionally approved residency programs. Participants were solicited based on the recommendations of Dillman (1978). A letter of invitation with a response postcard was followed by a postcard, then a second letter of invitation and response postcard and finally a certified letter. Individuals who declined to participate were asked to specify the reason for their refusal. New invitations were extended until the targeted sample size of 1008 was met. A similar four-step procedure was used to get the surveys returned at all three waves with the result that there were 95% rates of return across the three waves of surveys. To retain the desired number of participants in each cohort, new recruits to the survey were solicited to replace those who refused to answer the 1999 and 2004 surveys. The resulting sample consisted of a total of 1269 EPs that responded to one or more of the three surveys. Most had medical (MD) degrees with about 7% having Doctor of Osteopathy (DO) degrees and approximately 10–12% with other advanced degrees such as law degrees, masters, and doctoral degrees. Over 80% were male. Participant ages at the first survey were primarily between 35 and 49 with substantially smaller numbers in the older age groups (i.e., 55 to 65+). As would be expected, the sample ‘‘aged” so that by the 2004 survey there was a much more equal representation of participants across age with approximately 6% of the sample reporting that they were over 65. The sample was predominantly (89% or greater)

246

S. Pachulicz et al. / Journal of Vocational Behavior 73 (2008) 242–253

Human Capital

No. of hours worked (2) Work excitement (2) No. of years in EM (2) Certifications (2)

H1

H2 SocioDemographic Age (1) Race (1) Marital Status (3)

Obj. career success No. of academic leads (2) No. of EM leads (2) Salary change (3-1) H5

Outcomes (3) (Retirement, LV. EM, LV. Prof.) Planful (1) Social (1) Self efficacy (success) (1) Health (2) Leisure (2) Personal conflict (2)

Individual Difference

H3

H6

Subjective career (2) success Org. Sponsorship Org Support (2) Perc control (2) Further education Opportunities (2) H4

Fig. 1. Hypothesized model of career success and outcomes. Numbers in parentheses represent the time at which data were collected.

white with 5% reporting that they were Asian and only 1–2% reporting that they were Black or Latin American. Over 80% reported that they were married and nearly 80% reported that they had one or more children living with them. Most of the respondents reported income between $100,000 and $300,000. As expected, the portion reporting they were retired increased to 7.3% in the 2004 group. 2.2. Measures With some minor exceptions, the same 38-page survey was administered at all three times. For the purpose of this report, we used items and created scales that were relevant to the career success questions described above and at times in the data collection process that allowed for causal attributions consistent with the model in Fig. 1. We recognize that our data are all correlational, but timing of measurement does allow some weak causal attributions. 2.2.1. Subjective career success Career satisfaction was measured with four items that were considered indicators of a career satisfaction construct. These questions inquired as to the degree to which the EM specialty met their expectations, whether it measured up to the type of career they wanted when they selected it, whether they would select the specialty again, and how satisfied they were overall with their career in EM. Responses were made on four- or five-point Likert-type scales that were appropriate in content to the item stem. This measure like the indices of objective career success was collected at Time 2. 2.2.2. Objective career success Consistent with the Ng et al. (2005) definition of career success, we employed three objective career success indicators. The first of these was the number of academic leadership roles the person reported holding (e.g., academic dean, academic department chair, residency director). The second was the number of leadership roles the individual reported in leading emergency medicine groups (e.g., medical director of emergency department, emergency medical service physician director, or manager of emergency medicine group). The third was a salary progression variable. Income change was computed as the difference between respondents’ self-reported medical income from a variety of sources at Times 1 and 3 and was used to indicate their salary progression over a 10-year period. All three were objective variables self-reported by the participants. The income change variable at each time point was measured on a 10-point scale ranging from 1 (less than $50K per year) to 10 (more than $500K per year). Income change was computed using this 10-point scale.

S. Pachulicz et al. / Journal of Vocational Behavior 73 (2008) 242–253

247

2.2.3. Human capital indices The number of hours worked per week on various work activities was summed as a single measure. Three indicators of Work centrality or excitement assessed the respondents’ level of agreement that emergency medicine was ‘‘challenging,” ‘‘exciting,” and overall work is ‘‘fun.” The number of years respondents reported that they spent working in emergency medicine was a single item measure. Investment in educational efforts (referred to as Certifications in Fig. 1) was the number of graduate degrees, residency specialties, fellowship training, and certifications participants reported that they had received. 2.2.4. Organizational sponsorship Respondents indicated which of a set of 11 conditions were available to them on their jobs. These included both extrinsic items (e.g., fair compensation, fringe benefits, job security) and items intrinsic to their work (e.g., exciting work, sense of ownership) as well as items that reflected the degree of perceived control respondents had over their work situation (e.g., autonomy in work, control over working conditions, defined working hours), and further education opportunities (e.g., opportunity to attend conferences, opportunity for subspecialization, and teaching and research opportunities). Three indicators of each of these constructs were formed by summing randomly selected items to form parcels when more than three items were available. 2.2.5. Sociodemographic variables Gender, race, marital status, and age were collected. Gender was proposed as a moderator of the hypothesized relationships. Because of the small number of non-Caucasian respondents, race was coded 1 for Caucasians, 0 for non-Caucasians. Marital status was coded 1 for married, 0 for not married (single, widowed, etc.). 2.2.6. Individual differences Respondents were asked to describe themselves on a group of 13 bipolar adjectives on a scale ranging from 1 to 6. These self-descriptions comprised three scales. A ‘‘success” or general self-efficacy scale consisted of six items including active– inactive, incompetent–competent, successful–unsuccessful, and strong–weak. A ‘‘social” scale consisted of four items including cold–warm, interested in self-interested in others, and open–closed. A final set of three items we labeled ‘‘planful” included conventional–unconventional, dreamer–practical, and impulsive–deliberate. As was true for other constructs in the model, three indicators were computed for each construct. A single item requested their self-appraisal of their health (i.e., how would you describe your current health?). Response options for this scale ranged from one (some serious health concerns) to four (exceptionally healthy for my age). Two items assessed the degree to which enough time for family or personal life were serious problems. Response options for these two items ranged from ‘‘not a problem” (1) to ‘‘serious problem” (5). As they were correlated .90, they were summed to form a measure of personal conflict. Finally, a set of 30 leisure activities was rated on a seven-point scale with anchors ranging from ‘‘Not very enjoyable” to ‘‘Extremely enjoyable.” We formed three leisure activity indicators by randomly assigning items to the three indicators and summing them. Those who responded ‘‘not very enjoyable” or ‘‘do not participate” for a particular activity received a score of zero for that activity, and those who rated the activity ‘‘enjoyable” to ‘‘extremely enjoyable” received a score of one. 2.2.7. Career success outcomes We also considered three potential outcomes of career success. These included having thought of leaving the specialty of emergency medicine in the next year on a scale ranging from ‘‘very unlikely” (1) to ‘‘very likely” (5) and whether or not they hoped to leave medical practice for another career in the next 5 years (yes–no). Finally, respondents were asked if they had retired from medicine (yes–no). For the latter two variables, a yes response was coded 1, no was coded 0. 2.3. Data analysis Because of missing data across the three waves of data collection, we used the SPSS Missing Values Analysis (Version 15.0) to impute missing values for those respondents who failed to respond to all the items. This analysis provided data for 1269 participants. Using a likelihood function that assumes the missing data structure is conditional only on the variables in the model and not unspecified variables external to the model (i.e., data are missing at random; Schafer, 1997), the expectation–maximization algorithm (Dempster, Laird, & Rubin, 1977) in SPSS was used to generate missing values for cases on which we did not have complete data. This data set was used as the basis of the structural equation modeling analyses done with LISREL. It should be noted that the maximum likelihood estimation procedure we used assumes multivariate normality. As is true for many actual data sets (Lei & Lomax, 2005), this assumption was not met by our data. Research on the importance of violations of this assumption has indicated that nonnormality has little effect on the parameter estimates themselves (e.g., Fan & Wang, 1998), but with relatively extreme levels of skewness and kurtosis, the standard errors of the parameters may be underestimated (Finch, West, & MacKinnon, 1997; West, Finch, & Curran, 1995). A similar finding by Lei and Lomax (2005, p. 16) led them to conclude that the usual interpretation of ‘‘SEM parameter estimates can be accepted even under the severe nonnormality conditions” examined in their simulations. Violations of nonnormality inflate the v2 test, but seem to have little effect on fit indices (Lei & Lomax, 2005; West et al., 1995). All of these problems also seem to be greater when sample sizes are much smaller ( .05), but was related significantly to EM leadership positions held (standardized estimate = .24, p < .05), and negatively to income change such that individuals who had worked more years in EM experienced less increase in their income over the 10-year period than newcomers (standardized estimate = À.11, p < .05). The number of special certifications received (an index of educational effort/investment) was not related to any of the three objective career success areas (standardized estimates: academic leadership positions held, .03, ns; EM leadership positions held, À.01, ns; income change: standardized estimate = À.02, ns). As noted earlier, work excitement, as pictured in Fig. 1 (Ng et al., 2005) was misspecified. It was not related to any of the objective career indices, but was related significantly to subjective career success (.66). Hypothesis 2. Sociodemographic variables (age, race, and marital status) are related significantly to objective career success variables. Age was not related to EM leadership positions held (standardized estimate = .05, ns) nor academic leadership positions held (standardized estimate = .06). Income change, however, was related to age (standardized estimate = À.29, p < .05) such that younger individuals reported higher salary increases. Race was not related to academic leadership positions (standardized estimate = À.05, ns). Race was related to EM leadership positions and income change (standardized estimates = À.06 and À.08, respectively, p < .05) such that non-White EPs were more likely to hold EM leadership positions and experienced greater salary changes over the 10-year period of the surveys. Marital status was related only to EM leadership positions (standardized estimate = À.01, p < .05), but unrelated to academic leadership positions (standardized estimate = À.04, ns) or income change (standardized estimate = À.02, ns).

S. Pachulicz et al. / Journal of Vocational Behavior 73 (2008) 242–253 Table 1 Correlations and standard deviations of study variables Variables 1. Retired 2. Leave EM 3. Leave Med 4. CarSat1 5. CarSat2 6. CarSat3 7. CarSat4 8. EM Lead 9. AC Lead 10. Inc. Chg. 11. Yrs EM 12. Certs 13. Hrs. Work 14. Race 15. Age 16. MarStat 17. Workex1 18. Workex2 19. Workex3 20. Suc1 21. Suc2 22. Suc3 23. Soc1 24. Soc2 25. Soc3 26. Plan1 27. Plan2 28. Plan3 29. Leis1 30. Leis2 31. Leis3 32. Health 33. Personal 34. OrgSup1 35. OrgSup2 36. OrgSup3 37. Control1 38. Control2 39. Control3 40. FurEd1 41. FurEd2 42. FurEd3 1 .29 À.04 .01 .07 .07 .01 .03 .04 .05 À.04 À.06 À.02 .07 .06 À.01 À.07 .04 À.04 .01 À.03 À.04 À.07 À.00 À.01 .03 À.04 À.02 À.02 À.01 À.02 .01 .04 .06 .07 .08 .09 .07 .06 .08 .04 .09 .04 20 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. Suc1 Suc2 Suc3 Soc1 Soc2 Soc3 Plan1 Plan2 Plan3 Leis1 Leis2 Leis3 Health Personal OrgSp1 OrgSp2 OrgSp3 Control1 Control2 Control3 FurEd1 FurEd2 FurEd3 1.80 .40 .44 .33 .08 .36 À.07 À.01 .01 .16 .15 .12 .32 À.16 .12 .10 .12 .09 .13 .06 .04 .08 .07 2 1.04 .21 À.11 À.10 À.07 À.10 À.04 À.02 À.14 .19 .01 À.07 .03 À.00 .24 À.12 À.07 À.03 À.05 À.05 À.03 À.05 .04 À.05 À.04 .03 .00 .06 .07 .08 À.06 À.02 À.01 .02 À.04 .02 À.04 .02 À.03 À.02 À.04 21 1.81 .49 .26 .04 .32 .02 .08 .05 .06 .08 .10 .21 À.12 .11 .08 .10 .07 .13 .02 .05 .07 .06 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

249

19

.30 À.19 À.13 À.11 À.15 À.04 À.02 À.07 .07 .01 À.03 .04 .06 .09 À.11 À.14 À.12 À.04 À.08 À.14 À.03 À.01 .02 À.07 .02 À.04 .03 .02 .04 À.04 .07 À.04 À.03 À.07 .02 À.06 À.02 À.02 À.02 À.03 22

.85 .66 .63 .73 .12 .17 .09 À.00 .14 .32 À.07 .00 .05 .26 .47 .49 .15 .13 .17 .11 .10 .03 À.06 .04 .04 .28 .25 .29 .19 À.28 .41 .47 .39 .42 .37 .37 .17 .36 .35 23

.67 .72 .79 .13 .23 .02 .14 .10 .33 À.12 À.07 .17 .31 .48 .54 .19 .24 .23 .11 .05 .07 À.04 .05 À.02 .30 .31 .35 .19 À.30 .44 .51 .43 .44 .41 .39 .22 .41 .39 24

.68 .79 .19 .28 .02 .18 .12 .35 À.13 À.04 .23 .28 .50 .57 .20 .21 .20 .10 .05 .07 À.04 .02 .00 .35 .33 .37 .18 À.29 .49 .54 .46 .48 .44 .44 .25 .45 .43 25

1.07 .19 .31 .04 .17 .13 .41 À.09 À.07 .24 .33 .59 .68 .22 .21 .25 .12 .06 .05 À.02 .05 .01 .39 .38 .44 .20 À.36 .55 .62 .51 .55 .49 .50 .27 .50 .47 26

1.20 .10 .04 .14 .05 .21 À.07 À.08 .14 .06 .15 .18 .02 .01 .04 .02 .01 .00 .08 .02 .03 .19 .14 .14 À.01 À.01 .22 .19 .20 .18 .16 .17 .15 .22 .26 27

.26 À.06 .31 .09 .28 À.10 À.06 .28 .10 .13 .19 .09 .07 .06 .03 .05 .02 .07 À.01 .02 .26 .25 .27 .09 À.03 .40 .32 .32 .32 .31 .29 .11 .26 .21 28

2.88 À.36 .08 .11 À.06 À.02 À.36 .08 .05 .02 À.01 .01 .01 .05 .02 .04 .05 À.02 À.03 .00 À.04 À.06 .04 .13 .06 .04 .06 .04 .06 .03 .04 .07 .09 29

1.28 À.17 .09 À.06 À.08 .82 .00 .03 .14 À.03 À.07 À.08 À.17 À.06 À.06 .03 À.07 À.00 .14 .24 .25 À.08 À.13 .13 .18 .14 .18 .12 .21 .08 .16 .09 30

.52 .19 À.02 À.07 À.01 .02 .19 .19 .04 .07 .09 .08 .07 .02 .03 .03 À.02 .24 .20 .24 .09 À.10 .19 .20 .18 .23 .14 .17 .14 .23 .18 31

28.5 À.09 À.11 .13 .12 .31 .36 .05 .05 .05 .03 .06 .02 À.01 .01 À.04 .38 .37 .41 .10 À.11 .49 .50 .45 .53 .32 .47 .22 .48 .48 32

.31 .02 À.05 À.10 À.08 À.03 .01 À.05 .01 .07 .09 À.01 .13 À.03 À.03 À.13 À.12 À.13 À.03 .02 À.12 À.17 À.12 À.14 À.07 À.12 À.02 À.11 À.13 33

10.8 À.05 À.05 À.07 À.09 À.00 .03 À.10 .03 À.02 .04 À.10 À.03 À.03 À.11 À.07 À.10 À.01 .01 À.09 À.11 À.13 À.15 À.10 À.11 À.01 À.09 À.13 34

.39 À.02 .10 .22 .04 À.08 À.07 À.13 À.02 À.06 .03 À.02 À.03 .24 .34 .36 À.05 À.21 .19 .26 .19 .26 .18 .28 .08 .23 .12 35

.75 .54 .31 .13 .16 .16 .13 .04 .09 .07 .02 .02 .13 .14 .18 .07 À.08 .19 .21 .16 .15 .15 .15 .09 .14 .18 36 37

.91 .59 .14 .18 .21 .16 .05 .08 .03 .02 .05 .31 .30 .34 .11 À.25 .43 .47 .38 .41 .35 .38 .22 .38 .36 38

.96 .18 .15 .23 .12 .02 .09 À.01 .02 .00 .42 .42 .46 .14 À.34 .53 .57 .49 .54 .48 .46 .28 .49 .46 39

1.36 .40 .15 .27 .00 .09 .11 .12 .10 .11 .26 À.19 .11 .08 .06 .05 .11 .02 .05 .06 .04

1.89 .28 .49 À.06 .03 À.05 .15 .08 .08 .14 À.07 .05 .01 .03 .01 .05 .00 .03 .01 .03

1.04 .16 À.02 .12 .05 .05 .00 .04 .02 .01 .02 À.01 .04 À.00 .03 .02 À.03 À.02 À.01

1.29 À.14 À.05 À.09 .09 .06 .06 .07 À.02 .04 .04 .03 .03 .05 .03 .06 .05 .05

1.35 .25 .21 À.04 À.05 À.02 .02 .00 À.01 À.03 .02 À.01 .02 À.01 À.05 À.03 À.03

1.32 .38 À.06 À.03 .00 .02 À.03 À.02 À.02 À.05 À.03 À.01 À.06 À.06 À.06 .00

1.16 À.07 À.04 À.01 .07 À.04 .02 À.04 À.02 À.05 À.01 À.09 À.05 À.06 À.02

7.82 .80 .78 .13 À.24 .54 .60 .50 .59 .42 .55 .26 .54 .45

7.93 .83 .08 À.27 .52 .58 .47 .59 .37 .54 .20 .49 .43

8.71 .08 À.28 .54 .60 .48 .62 .40 .55 .22 .53 .43

.86 À.08 .11 .10 .11 .08 .06 .09 .01 .08 .06

1.19 À.23 À.32 À.23 À.30 À.25 À.29 À.07 À.21 À.12

1.13 .79 .73 .73 .59 .66 .43 .72 .65

.91 .74 .91 .83 .70 .48 .61 .56 .61 .76 .67 .77 .30 .34 .29 .72 .67 .71 .62 .62 .58 (continued on

.49 .55 .49 .28 .27 .54 .66 .45 .50 next page)

250 Table 1 (continued)

S. Pachulicz et al. / Journal of Vocational Behavior 73 (2008) 242–253

40 40. FurEd1 41. FurEd2 42. FurEd3 .38 .75 .46

41 .70 .68

42

.83

Note: Correlations above .06 are statistically significant, p < .05. Values on the diagonal are standard deviations. CarSat, career satisfaction; Inc. Chg., income change; Yrs EM, years employed as an emergency physician; Certs, certifications; MarStat, marital status; Workex, work excitement; Suc, self-efficacy or success; Soc, social; Plan, planful; Leis, leisure; Personal, personal life-work conflicts; OrgSp, organizational support; and FurEd, further educational opportunities.

Hypothesis 3. Individual difference variables (planning, sociability, self-efficacy or success, health, leisure activity), will be related positively to subjective career success, and Personal conflict will be related negatively to subjective career satisfaction. Three of the six individual difference variables showed a significant positive relationship with subjective career success consistent with Hypothesis 3. Self-efficacy was positively related to subjective career success, such that individuals higher in this trait also reported more subjective career success (standardized estimate = .11, p < .05). Personal conflict or time problems was negatively related to subjective career success, such that individuals who reported having more conflicts for personal and family time also reported lower subjective career success (standardized estimates = À.07, p < .05). Health was positively related to career satisfaction (standardized path estimate = .06, p < .05), which indicated that individuals who reported better health were more satisfied with their careers. Those who were engaged in fewer leisure activities also reported more career satisfaction (standardized path estimate = À.08, p < .05). This relationship is opposite that hypothesized. A post hoc explanation may be that some individuals perceive that their work interferes with their leisure activity. Self-reports of planfulness were not related significantly to career satisfaction and sociability was related significantly, but negatively to career satisfaction (standardized estimates of À.03 and À.09, p < .05). Both of the latter coefficients were inconsistent with Hypothesis 3. Hypothesis 4. Organizational sponsorship variables including organizational support items, perceptions of control over one’s work environment, and further educational opportunities are related positively to subjective career success. This hypothesis was supported for the organizational support variable (standardized estimate = .38, p < .05), indicating that the more support a person perceived, the higher the reported subjective career success. Educational opportunity was related negatively to career success (standardized estimate = À.02, NS) and perceived control was related negatively to career success (standardized path estimate = À.15, NS). Both of the latter negative and nonsignificant relationships were likely a function of suppressor effects as the relationships between the indicators of these constructs and career satisfaction indicators were all positive (see Table 1).
Table 2 Parameter estimates for hypothesized structural model Input Input to mediator relationships Subj carsat Human capital Hours Worked Yrs EM Ed Effort (Certs) Socio-Demogr. Age Race MaritStat Ind diffs Planful Social Self-eff Health Leisure Pers confl Work excitement Org spon OrgSup Perccont Fur ed opp À.03 À.09a .11a .06a À.08a À.07a .66a .38a À.15 À.02 AcadL .18a .07 .03 .06 À.05 À.04 EMLead .24a .24a .01 .05 À.06a À.01 SalaryD .13a À.11a .00 À.29a À.08a À.02 Mediator Acadl EML SalaryD CarSat Mediator to outcome relationships Retired .05 À.01 À.44a À.11a Leave Med À.01 .03 À.08a À.17a Leave EM À.01 .00 À.16a À.10a

Note: In the interest of space, we do not present the parameters associated with the measurement model, but these are available upon request from the second author. a p < .05.

S. Pachulicz et al. / Journal of Vocational Behavior 73 (2008) 242–253

251

3.2. Mediator–outcome relationships The career success variables were each hypothesized correlates of the three outcome variables and tests of those relationships constituted a test of Hypothesis 5. Hypothesis 5. Objective career success is related negatively to career outcomes such as intentions to leave medicine, intentions to leave emergency medicine, and retirement. The parameter estimate for income change was statistically significant, such that the more respondents’ income increased, the less likely the person was to have retired or report that they intended to leave medicine or the EM specialty (standardized estimates = À.44, À.08, and À.16, respectively, p < .05, respectively). None of the relationships between leadership positions held and the three outcome variables were significant statistically. Hypothesis 6. Subjective career success is related significantly to career outcomes such as intentions to leave medicine, intentions to leave emergency medicine, and retirement. Significant (p < .05) standardized parameters were À.11, À.17, and À.10 for retirement status, intent to leave medicine, and intent to leave emergency medicine, respectively. In all cases, EPs were more likely to indicate that they would leave their practice when their career satisfaction was low. Hypothesis 7. Relationships between variables are moderated by gender. A multigroup analysis of the model was attempted but did not converge. Separate models of covariance matrices represented by male and female responses were then evaluated. The model fit the male responses well (v2 (673) = 2892.01, p < .01, RMSEA = .05, SRMR = .06, CFI = .96, NNFI = .95), but fit of the female covariance matrix was not as good (v2 (673) = 1970.23, p < .01, RMSEA = .08, SRMR = .08, CFI = .90, NNFI = .87). Because the nonconvergence of the multigroup model precluded the usual v2 difference tests across gender groups, we computed t-tests of the difference between corresponding male and female parameters. Parameter estimates for the two gender groups are presented in Table 3. Six of the 40 structural parameter estimates were statistically significantly (p < .05) different. The work excitement variable was more strongly related to men’s career satisfaction than it was for women, whereas social and self-efficacy variables were more strongly related to female (as opposed to male) career satisfaction levels. Men’s ages were more strongly and negatively related to income changes than were women’s ages. Race to income change relationships differed significantly in direction for men and women. For women, it was the case that income changes were greater for Caucasians than Blacks. For men, the reverse was true in that income changes were greatest for minority respondents as opposed to Caucasian males. For women, additional educational and professional certificates actually resulted in smaller salary changes, whereas the relationship for men was nonsignificant.

Table 3 Parameter estimates for hypothesized model for men and women Input Input to mediator relationships Subj carsat M Human Capital Hours Worked Yrs EM Ed Effort (Certs) Socio-Demogr Age Race MaritStat Ind diffs Planful Social Self-eff Health Leisure Pers confl Work excitement Org spon OrgSup Perccont Fur ed opp .01 À.02 .02 .06a À.07 À.05a .75a .49a À.31 À.03 .02 À.70a .97a À.05 À.31 À.01 .46a À.93 1.74 À.41 F AcadL M .19a .09 .02 .06 À.06a À.04 F .21a .19 À.09 À.11 .04 À.01 EMLead M .23a .29a .00 .00 À.04 .03 F .13a .18 .14a .22 À.14a À.10 SalaryD M .14a À.17a .05 À.29a À.09a À.01 F .06 À.11 À.19a À.07 .14a À.10 Acadl EML SalaryD CarSat Mediator to outcome relationships Mediator Retired M .05 À.04 À.45a À.10a F .00 .12 À.39a À.04 Leave Med M .00 .04 À.10a À.18a F À.01 À.01 À.14a .09 Leave EM M À.01 .00 À.19a À.10a F À.04 À.01 À.04 À.09

Note: M and F refer to male and female. Bolded estimates are significantly different, p < .05. a p < .05.

252

S. Pachulicz et al. / Journal of Vocational Behavior 73 (2008) 242–253

4. Discussion Perhaps the most important contribution of our study was the overall confirmation of the model of career success based on the meta-analytic work of Ng et al. (2005). Each of the hypothesized links in their model was confirmed, though not for all variables in each of their sets of career success determinants. Human capital in the form of hours worked and years as an EP was related to leadership positions and salary changes held though the number of educational or professional certifications was not. The latter finding may be unique to the EM profession. The number and variance of these certifications was not high limiting the possibility of large correlations with other variables. Of the demographic variables, age and race were negatively related to salary change though the latter relationship was moderated by gender as noted above. With the exception of a relatively small negative relationship between race and EM leadership positions, there was little evidence of demographic correlates of the number of leadership positions that respondents held. Individual difference constructs were significantly related to subjective career success with the exception of the planning variable though the social variable was negatively related to career satisfaction. Self-efficacy or perceptions of success were the most significant predictor of career satisfaction 5 years later, but tests of gender differences indicated this finding was particularly strong for women and nonsignificant for men. The perception that job responsibilities led to conflicts with one’s personal life was negatively related to career satisfaction as expected. Inconsistent with the Ng et al. (2005) model, however, was the finding that engagement in leisure activities was negatively related to career satisfaction. This finding may be consistent with the personal conflict–career satisfaction relationship. Perceptions that one’s organization was supportive in various ways was the only one of three organizational sponsorship variables that were related to career satisfaction. This set of variables was highly intercorrelated and the findings here may represent a suppressor effect rather than differences in the constructs measured in this set of predictors. Finally, work excitement was not related to objective career satisfaction as hypothesized, but it was strongly related to subjective satisfaction, more so for men than women. In terms of the mediator–outcome relationships examined, it is clear that leadership positions held played a minimal role in reported intentions to leave medicine or EM or retire. On the other hand, changes in one’s income were particularly strongly related to the decision to retire, and less so for intentions to leave EM and medicine. Career satisfaction was also related to all three outcomes in the expected negative direction. These mediator–outcome relationships are particularly impressive as the outcomes were collected 5 years after the mediator variables. A second major contribution of the study was the fact that the relationships suggested by Ng et al. (2005) were evaluated and at least partly confirmed using longitudinal data. Ng et al. based their model on meta-analyses of the correlates of career success, but most of the primary studies in this area are cross-sectional. Our data shows that, in general, these relationships hold over a relatively long period of time. Future research should explore over what period of time these relationships exert an impact. Surveys in this study were separated by 5 years, but no theoretical rationale exists to support this period of time or any other. It is also the case, however, that the data were correlational. Hence the temptation to derive strong causal attributions should be avoided. Second, sampling of respondents from the existing EP population was done carefully, but we do not have evidence of the generalizability of our results to members of other professional groups is limited. Because of space considerations, we did not report the means of variables for male and female subgroups. That analysis did reveal the expected pattern of women indicating more difficulties with personal time and family time and that women did indicate less perceived support from the organization, fewer EM leadership positions, and less perceived control over their work situation than did men. These are all problems that apparently continue to confront women in modern organizations; awareness of this situation must also be accompanied by appropriate interventions if the contributions of women are to be fully realized. The results also provide a basis for some practical suggestions as well. Change in salary levels appears to be an important signal that an individual either will leave or intends to leave this occupation. This may be a way in which organizations signal to people that they are no longer as supportive of their role or that these individuals are performing at a level that no longer merits significant raises. It also means that if organizations want to keep valued experienced talent, it should take steps to ensure desired compensation levels are maintained. Second, a sense that the organization is supportive of employees’ efforts appears to be a central factor in their career satisfaction, and indirectly, in their long term retention. Perhaps self-evident, on the individual level, those who report being excited and challenged by their work are likely to report career satisfaction and remain with the occupation. Finally, for employees themselves, it appears that investments in one’s job in the form of hours and years worked do pay dividends in terms of the objective indicators of career success. 4.1. Limitations Perhaps the most significant limitation of the study was the reliance on archival data and the fact that this meant we were forced to use existing measures to operationalize some of the intended constructs. For example, planful is certainly quite similar in concept to conscientiousness in the broader literature on personality, but the items used in this study may not appear on any existing measure of conscientiousness. The social items are similar in concept to extraversion and the selfefficacy or success items reflect a similar construct in the personality literature. Similar statements can be made about the work excitement items and the organizational support and perceptions of control measures. However, in spite of the fact that we used ad hoc items and measures, the results were consistent with the general model.

S. Pachulicz et al. / Journal of Vocational Behavior 73 (2008) 242–253

253

A second possible limitation relates to the timing of the collection of the data. As mentioned above, there is no theory to guide the timing of measurements in longitudinal research and the 5-year interval used in this study was guided primarily by the resources of the organization that sponsored the research. Organizational research, and possibly much psychological research, would benefit greatly from some attention to estimating the time required for various processes to exert an impact on behavioral or attitudinal outcomes. A final limitation may be the relatively small sample size of the female group in our study. While proportionately representative of the population of EPs, the sample size of 222 meant that some of the parameters estimated were likely unreliable and would be very different in a new sample. The model fit the female sample less well than the male sample also. We examined modification indices and residuals and tried various model modifications, but none increased fit substantially nor were they convincing conceptually so only results for the one model are presented. As in any modeling effort, though, there are certainly other models that fit the data equally well. 4.2. Conclusions Overall, the results provide support for the hypotheses formulated in the introduction to the paper. This study does confirm previous work on career success and satisfaction among physicians that indicates that hours worked was positively related to career satisfaction, but also to a number of other outcomes. However, one previous study (Williams et al., 2001) suggested that satisfaction may decrease when the number of hours worked went over 60 h a week. Mean number of hours in our study was small partly because some individuals were no longer working as EMs (for those working, the mean number of hours working in EM was 30 a week). Consistent with past research, organizational sponsorship variables played an important role in subjective career success and a role in intentions to leave the profession. Work excitement/centrality was important to subjective career success rather than objective career success. Results of past studies on income have indicated that income plays a minimal role in career satisfaction, but the results of the present study indicate that income change plays a significant role in intentions to leave EM, intentions to leave medicine, and retire. References
Bray, D. W., Campbell, R. J., & Grant, D. L. (1974). Formative years in business. New York: Wiley. Dempster, A., Laird, N., & Rubin, D. (1977). Maximum likelihood from incomplete data via the EM algorithm. Journal of the Royal Statistical Society, 39, 1–38. Dillman, D. (1978). Mail and telephone surveys. New York: Wiley. Du Toit, M., & Du Toit, S. (2001). Interactive LISREL: User’s Guide. Lincolnwood, IL: Scientific Software, International. Fan, X., & Wang, L. (1998). Effects of potential confounding factors on fit indices and parameter estimates for true and misspecified models. Structural Equation Modeling, 5, 701–735. Finch, J. F., West, S. G., & MacKinnon, D. P. (1997). Effects of sample size and nonnormality on the estimation of mediated effects in latent variable models. Structural Equation Modeling, 4, 87–105. Frank, E., McMurray, J. E., Linzer, M., & Elon, L. (1999). Career satisfaction of U.S. women physicians: Results from the Women Physician’s Health Study. Archives of Internal Medicine, 159, 1417–1426. Glisson, C., & Durick, M. (1988). Predictors of job satisfaction and organizational commitment in human service organizations. Administrative Science Quarterly, 33(1), 61–81. Hall, D. T. (1976). Careers in organizations. Pacific Palisades. CA: Goodyear. Hall, D. T. (2002). Careers in and out of organizations. Thousand Oaks. CA: Sage. Hu, L., & Bentler, P. M. (1998). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55. Judge, T. A., Higgins, C. A., Thoresen, C. J., & Barrick, M. R. (1999). The Big Five personality traits, general mental ability, and career success across the life span. Personnel Psychology, 52, 621–651. Landon, B. E., Reschovsky, J., & Blumenthal, D. (2003). Changes in career satisfaction among primary care and specialist physicians, 1997–2001. Journal of the American Medical Association, 289(4), 442–450. Landon, B. E., Reschovsky, J. D., Pham, H. H., & Blumenthal, D. (2006). Leaving medicine: The consequences of physician dissatisfaction. Medical Care, 44(3), 234–242. Lei, M., & Lomax, R. G. (2005). The effect of varying degrees of nonnormality in structural equation modeling. Structural Equation Modeling, 12, 1–27. Lepnurm, R., Danielson, D., Dobson, R., & Keegan, D. (2006a). Cornerstones of career satisfaction in medicine. Canadian Journal of Psychiatry, 51(8), 512–522. Lepnurm, R., Dobson, R., Backman, A., & Keegan, D. (2006b). Factors explaining career satisfaction among psychiatrists and surgeons in Canada. Canadian Journal of Psychiatry, 51(4), 243–257. Ng, T. W. H., Eby, L. T., Sorensen, K. L., & Feldman, D. C. (2005). Predictors of objective and subjective career success: A meta-analysis. Personnel Psychology, 58, 367–408. Reinhart, M. A., Munger, B. S., & Rund, D. A. (1999). American board of emergency medicine longitudinal study of emergency physicians. Annals of Emergency Medicine, 33, 22–32. Robinson, G. E. (2004). Career satisfaction in female physicians. Journal of the American Medical Association, 295, 635. Schafer, J. L. (1997). Analysis of incomplete multivariate data. Monographs on statistics and applied probability 72, Chapman and Hall/CRC. West, S. G., Finch, J. F., & Curran, P. J. (1995). Structural equation models with nonnormal models. In R. H. Hoyle (Ed.), Structural equation modeling. Thousand Oaks. Williams, E. S., Konrad, T. R., Scheckler, W. E., Pathman, M. L., McMurray, M. G., & Schwartz, M. (2001). Understanding physician’s intentions to withdraw from practice: The role of job satisfaction, job stress, mental and physical health. Health Care Management Review, 26(1), 7–19.