Job satisfaction trends during nurses' early career

Understanding what motivates workers and how this impacts on performance has always interested organisations and managers and different theories have sought to answer this question. Fung-Kam [10] identified four general theories: need/value fulfilment theory; person-environment (P-E) fit theory, the theory of career and the theory of work adjustment. Similarly Adams and Bond [11] classified job satisfaction theories into three groups: discrepancy theories, which examine the extent to which employees needs are satisfied in the work place; equity theories, which highlight social comparisons in the evaluation of job rewards; and expectancy theories which focus on employee motivations. The theory of need/value fulfilment proposes that the discrepancy between individual needs and the extent to which the job meets these needs is negatively related to job satisfaction and expectation gaps [12] have been linked to the violation of the psychological contract between employer and employee [13]. Theory of P-E fit suggests there are person characteristics that suit working environments better than others and working environment characteristics that suit certain individuals better. Drawing on both theories failure to meet expectation has been shown to be related to lower work commitment of graduates and the extent they 'fitted' in was a central motivation to remain with an employer [14]. Connected to P-E fit is the theory of work adjustment. This theory is concerned with the degree of correspondence between individuals and their work environments. Hackman and Lawler [15] believed that the employees' perception of their job rather than the jobs objective characteristics was a more important determinant of job satisfaction. Holland's theory of career [16, 17] suggests that personality and type of work are congruent so similar types of individuals converge on the same occupations.

The two theories that have been important in the development of an understanding of job satisfaction in nursing are Maslow's human needs theory [18] and Herzberg and Mausner's motivation-hygiene theory [19]. Maslow identified two types of needs; deficiency needs (physical, safety and belonging) and growth needs (self-actualization and self esteem) [20]. Herzberg and Mausners' theory consists of intrinsic factors or 'motivators' that promote job satisfaction and extrinsic factors or 'hygiene factors' that cause dissatisfaction [8]. Kramer's reality shock theory [21] is based on the reaction new nurses feel once they enter a work situation that they are unprepared for and has also been used to understand job satisfaction in nurses in early career.

Blegen [6] synthesised findings from 48 studies and identified thirteen variables that were most strongly associated with job satisfaction. These included stress, commitment, communication (with supervisor and peers), autonomy (and locus of control), recognition, routinization, and fairness. A secondary analysis of data [22] from the 1977 Quality of Employment Survey [23] found that task variety, relations with co-workers, financial rewards and age were all positively associated with job satisfaction. Conversely role conflict and tenure had negative effects although the latter finding was not consistent with other literature [22]. Work attitudes (supervisor support, work-group cohesion, variety of work, autonomy, organizational constraint, promotional opportunities, work and family conflict, and distributive justice) were also important in explaining the job satisfaction of registered nurses in the United States [24].

Different managerial styles and practices at the organisational unit level (e.g. ward) can have a direct bearing on nurse satisfaction. The work of Adams and Bond [11] highlighted the importance of interpersonal relationships with nurses and other medical staff, workload and ward cohesiveness. A number of studies [25‐28] have shown a positive association between autonomy and levels of job satisfaction and which has been confirmed amongst nurses [6, 10, 29‐31].

The effect of educational level on job satisfaction has been conflicting. Some studies have found a positive association with job satisfaction [32, 33] and others a negative association [3, 6, 34]. Lower level qualifications impacted positively on job satisfaction based on findings from a survey of NHS nurses [35]. The same study found training had a positive impact on job satisfaction but diminished with the number of training spells. The inverse relationship supports the argument that education raises expectations that subsequently are not met [14] whereas a positive correlation suggests that the greater extrinsic rewards that come with education raises satisfaction. Blegen [6] found that job satisfaction correlated less strongly with age or years of experience, while Shields and Ward [35] found that increasing age, marriage, and children impacted positively on nurses' satisfaction.

In a study of nurses working in the NHS [35], individuals who stressed non-pecuniary reasons (e.g. flexibility of hours, helping others) had significantly higher job satisfaction than those who did not (e.g. attracted by job security, promotion prospects, pay). Not being graded fairly was the largest negative determinant of overall job satisfaction and not having the hours to suit an individual's preference had a negative impact. Absolute and relative levels of pay (compared with other occupational groups) are also important [36].

There have been a number of US studies recently which have researched job experience and satisfaction during the early years after qualification [1, 2, 4, 5, 9]. These have covered the period up to 18 months after qualification and some are longitudinal [2, 4, 9]. Graduate nurses were found to lack confidence in skill performance and had concerns with peer and preceptor relationships, dependence (on others) and becoming an independent practitioner, the work environment, organizational and priority setting and communications with physicians [9]. Low-points in satisfaction and confidence between 6 and 12 months have been reported in the US [9]. Graduates participating in residency programs were particularly vulnerable between entry and 6 months [2]. Dissatisfaction with patient care, scheduling (work-life balance) and pay may precipitate job exit [4]. Low pay satisfaction has been reported in the US [9].

Before a nurse can practice in the UK they must hold a bachelors degree or diploma in nursing. These educational programmes consist of theory and practice (in community and hospital settings) in roughly equal proportions beginning with a common foundation programme (CFP) lasting 12 months followed by about two years in one of the four branches of nursing: adult(general), mental health, learning disabilities or children's nursing. Either qualification allows a nurse to register with the Nursing and Midwifery Council (formally the United Kingdom Central Council (UKCC)). The diploma qualification is similar in educational level to a US associate degree but training extends over three years and there is a single level of professional registration (equivalent to a US Registered Nurse) irrespective of the training programme. Adult nurses work with adults of all ages and children's nurses with newborns to adolescent (0–16 years). Both provide care to patients with chronic and acute health conditions. Learning disability nurses help, care and develop the skills of people with disabilities in family, community and residential settings, and adult and young peoples' education. Mental health nurses provide care for mentally ill people and their families primarily in the community and less often in a hospital setting. Nurses can specialise further after gaining experience in more general settings post-registration. At the time of the study the pre-registration degree and diploma were the only points of entry. The majority (over 80%) of individuals opted for the diploma programme.

Newly qualified nurses in the UK typically used to start as D Grade nurses and after a minimum of 6 months post-registration experience could become E grade nurses. Nurses at this grade were often encouraged to gain valuable management experience and/or receive further training in a specialty (e.g. accident and emergency). F grade nurses had more of a managerial role and were sometimes left in charge of a ward or other setting. In December 2004 these grades were superseded in the UK by Agenda for Change pay bands [37].

Our research examines how job satisfaction varies amongst newly qualified nurses in the England over time in early career (6 months, 18 months, 3 years) and to what degree trends in factors arising from a factor analysis of a multi-item job satisfaction question [38] vary between specialisms.

The design was longitudinal and correlational with data collected at four time-points (qualification, 6 months, 18 months, 3 years). The 6 month time-point was selected to reflect early experiences, 18 months to reflect established early experiences and first promotion, and three years to reflect consolidation. More frequent follow-up was considered but deemed undesirable due to the potential burden this would place on respondents who were asked to complete large questionnaires at each survey sweep.

The study population was formed from all nurses qualifying in 1997/98 from the diploma programme in England. A mixed approach to sampling was undertaken determined by the population size of each branch. Initial estimates of the children's and learning disability branch nurse populations were small (574 and 246 respectively). Estimates of the adult and mental health branches were much larger (4850 and 1142 respectively). A decision was taken therefore to sample all children's and learning disability nurses qualifying in England. The larger adult and mental health branch nurse populations were sampled using a multistage approach based on the eight regional health authorities (that existed at the time), colleges and intakes. Between three and eight college of nursing (depending on branch) were located in each region. Approximately a half and two-thirds of colleges were sampled from each region for the adult and mental health branches respectively. These sampling fractions were based on information previously collected from the English National Board and our own enquiries to Colleges of Nursing. There was further sub-sampling at the college level of the adult branch when intakes were large. The number of college intakes(classes) for the adult, children's nurses, mental health and learning disability branches was 46, 49, 56 and 34, the total number eligible to be recruited was 2109, 758, 293 and 802 respectively (total 3962) and recruitment rates were 87%, 93%, 90% and 85% respectively (88% overall). A comprehensive report of the sampling strategy is found in Marsland and Murrells [39].

The number and percentage responding at each sweep was as follows: qualification (3009, 80%), 6 months (2524, 64%), 18 months (2118, 53%), three years (1785, 45%). Non-response is common to most longitudinal postal surveys particularly when the point of contact is home address. We asked nurses to inform us of address changes. Questionnaires were always sent to last known address and if no response to second address (typically parents) and finally to the UKCC registered address (written on the envelope at the UKCC premises). Higher non-response rates were noted in parts of the country where nurses were likely to change address more often (e.g. London) and therefore contact was more easily lost. Region of workplace was therefore included in the statistical model to compensate for regional heterogeneity.

We wanted to measure job satisfaction specifically for nurses in early career. Existing nursing scales suffer from a number of limitations. Some have not been adequately tested for reliability and validity, they are often very long or very short, not contemporary, developed from theory without contextual representation and developed on nurses from different healthcare systems [40]. For this reason we chose to develop a new instrument [40] using the five-step method recommended by Spector [41]. Following in-depth interviews of 30 diploma-qualified nurses a pool of items was generated and the number of interviewees who regarded each item as important was noted. A total of 34 items were identified. A small number of items were added as the study progressed to reflect aspects pertinent to career after qualification. Each item was measured on a five point scale from very satisfied to very dissatisfied. Some items had a 'not applicable' or similar response option. The set of items available for psychometric analysis was reduced because either the item did not apply (e.g. many respondents did not have a family or partner) or the item did not apply across all time points (e.g. content of appraisals). Further analysis was confined to the twenty items applicable to over 90% of respondents. Psychometric analysis was initially confined to the adult branch and included tests of temporal stability across time. Further validation across specialties and over time has now taken place [38]. Factor analysis (not reported here) of twenty selected items that were asked at all three time-points and were applicable to at least 90% of respondents identified two potential seven factor measurement models (Client Care, Staffing, Development, Relationships, Education, Work-Life Interface, Resources) for nurses' job satisfaction in early career that differed on the loading of one item, emotional support from immediate line-manager either on the Relationship(Model I) or the Development (Model II) factor. There was little difference in the fit of the two models. Model I was a better fit for the adult branch and model II for the child and mental health branches. The difference in overall fit however was small (Root Mean Square Error Approximation 0.021 and 0.022 respectively). We did not want to further burden respondents who were asked to complete a large questionnaire (sometimes running to over 60 questions) on four occasions over a three year period) with an additional questionnaire to test inter-rater reliability however internal consistency was good with Cronbach's alphas all above 0.7 except for Work-Life Interface and correlations between factor scores recorded at 6 and 18 months ranged from 0.27 to 0.42. The instrument did not provide a good fit for the learning disability nurses and for this reason we have excluded this branch from the data analysis. Findings were similar for both models and for this reason (and for brevity) we report on the second measurement model (Table 1) which had a better fit for two of the three remaining branches.

Pay which was represented by a single question item was included as a notional eighth factor. The job satisfaction question included two items on pay and grade which were Pay in relation to responsibility and Grade/position in relation to level of responsibility. The second item was asked from 18 months onwards and was therefore excluded from the psychometric analysis on the grounds that it had not been asked across all three time-points.

All participants gave informed consent. An initial request to access students was made by letter to the 'head' of each college [42]. The letter described the project and asked if research team members could meet with students and invite them to participate. The letter emphasised that participation was voluntary. The recruitment meeting consisted of a short presentation of 10 minutes followed by a discussion of up to 30 minutes. During the meeting forms were circulated and completed by those willing to participate. Information requested included current address, a second, more permanent, address (e.g. parents). The form was accompanied by an information sheet. Various strategies were used to recruit non-attendees such as asking group members and course leaders to pass on recruitment packs, writing to non-attendees via the course leader and repeating the visit when there were a large number of non-attendees.

Agreement to participate was high for those who attended these meetings (over 80%) but on the few occasions when face-to-face recruitment was not possible this fell to below 50%. Participants were supplied with change of address slips (and freepost envelopes) so if they moved we were kept informed. Change of address slips were also supplied with each survey questionnaire. Between 18 months and 3 years after qualification lost participants were traced via the UKCC.

Job satisfaction information was collected as part of a much larger postal questionnaire sent at 6 months, 18 months and 3 years after qualification. Baseline demographics and other profile information were collected at qualification. A question was designed specifically to collect career history information and some of this information (e.g. number of previous nursing posts) was used in the modelling.

Although the longitudinal study of nurses qualifying from the pre-registration diploma course predated the requirement of Multi-Centre Research Ethics Committee approval, guidance was followed from staff of the university from which students were recruited as to the internal procedures required for ethical approval.

Factor scores for each nurse were produced by taking the mean of all the non-missing item scores. We applied the more stringent condition that at least half of each factor's items had to be answered otherwise the factor score was set to missing. There is not set guideline on this. Bryman and Cramer [43] used 50% or more as their exclusion criteria in the example they presented. We could have considered multiple imputation [44] but the success of this method depends very much on the correction specification of the non-response model and so this option was not pursued. The average percentage of respondents answering all items of a factor was 94% therefore any resulting biases should be small.

A comparison of previous satisfaction levels of responders and those who did not respond at subsequent time points was conducted in order to determine if non-response was itself related to satisfaction. Response group means were compared statistically using ANOVA. This analysis did not reveal any major differences in job satisfaction scores between response groups nor did an analysis confined to those nurses with complete data across all three time-points differ from nurses who provided data at one or more time-points. On the few occasions where differences did emerge these have been identified in the tables showing factor means.

Summary statistics (means, standard errors) were produced, by branch, across the three time-points. A strategy for fitting repeated measures models, similar to the one proposed by Wolfinger and Chang [45] was followed. Three possible covariance structures (compound symmetry, Huynh-Feldt and unstructured) were tested statistically and the best fitting covariance structure was selected. Differences between branches at each time-point were tested statistically using ANOVA. The F-test for the time effect within branch was obtained from a mixed model incorporating the selected covariance structure. In the tables showing factor means we test the equality of branch means at each time-point and show the result in the far right hand column. At the foot of each factor sub-column we test the equality of means across the three time-points within branch. At the end of the same row is a test of equality of means across branch and time as indicated by the four degrees of freedom in the numerator of the F-test. All these tests were computed using SAS Version 8.

The analysis across time within branch was then repeated having accounted for variation, attributable to ten moderating variables. Some were time varying (children, spouse or partner, job grade, region, age, number of previous nursing posts, time in current job) whilst others did not change across time (sex, ethnicity, highest education qualification) (Table 2). All these variables were entered into the model along with a factor representing the three time-points.

The sample profile is shown in Table 2.

Mental health nurses were older at qualification than adult branch nurses by about two and a half years who were themselves older than child branch nurses also by two and a half years. Over 90% of adult and child branches were female compared with about 70% of mental health nurses. Mental health nurses were more likely to belong to an ethnic minority than the other two branches and already have a degree level qualification. Grade progression occurred more rapidly for mental health nurses whereas progression was slowest for adult branch nurses. By 3 years most nurses were onto their second or third post. Nurses tended to stay in post for similar amounts of time and by three years had been in their current post for almost a year.

For Client Care (Table 3), scores were highest for the children's nurses and lowest for mental health nurses. Scores changed little across time in both cases. The time profile was different for the adult branch with a similar score to the mental health branch (3.18 vs. 3.13) at 6 months rising to 3.44 at 3 years. Children's nurses were on the whole happier with Staffing (Table 3) than the other two branches. Satisfaction with Staffing dipped at 18 months for children's nurses and mental health nurses whereas a linear upward trend was observed for adult branch nurses. For these nurses a low score (3.12) was obtained at 6 months and by 3 years adult branch nurses were more satisfied with Staffing than mental health nurses.

Development scores (Table 3) remained reasonably stable across time for all branches. The adult branch had a significantly lower score at 6 months but had caught up children's nurses by 18 months.

The only factor to produce scores in excess of 4 across all branches and time-points was Relationships (Table 3). Trends were stable across time and were significantly higher for children's nurses than the other two branches although in real terms the differences were small.

A common profile for Education scores (Table 4) emerged for all branches starting off low at 6 months followed by a sharp increase at 18 months where it remained, apart from a small increase, for children's and mental health nurses whereas the score continued on an upward trajectory for the adult branch. By 3 years the adult branch nurses had almost caught up the mental health nurses.

Satisfaction with Work-Life Interface (Table 4) progressed positively, although at different rates across all branches. Mental health nurses had established higher levels early on. Greatest progress was observed for the adult branch nurses who started from a significantly lower base at 6 months and had overtaken children's nurses by 3 years.

Satisfaction with Resources (Table 4) remained low throughout for mental health nurses whereas satisfaction improved over time for both adult branch and children's nurses although there was a slight dip at 18 months for children's nurses. The scores for the mental health nurses were comparable to that obtained on Client Care and it would not be unreasonable to assume that there was connection between the two.

The lowest scoring factor of all was Pay (Table 4). Pay satisfaction improved over time for the adult branch and after starting off low (2.44) had almost caught up mental health nurses (2.63 vs. 2.65) by 3 years whereas it fell for children's nurses although this was neither statistically significant nor substantial in real terms.

In summary the strongest trends emerged for the adult branch with increasing job satisfaction from six months onwards for Client Care, Staffing, Education, Work-Life Interface and Resources (Tables 3 and 4). There were also significant increases between six and eighteen months for Development and Pay. Fewer significant trends were apparent for the child and mental health branches. Mean profiles over time were found to vary significantly between branches on the Care, Staffing, Development, Work-Life Interface, Resources and Pay factors. All eight factors differed significantly between branches at six months. Six differed significantly at eighteen months and four at three-years. The rank order of job satisfaction components was very similar for adult and child branch nurses (Figure 1).

Means were then adjusted to account for variation attributable to the ten moderating variables. These variables accounted for a small proportion of the variability and findings remained largely unchanged although there was some change in statistical significance (Adult, Development and Resources no longer significant; Mental health, Resources significant downward trend) and more substantively for Children's nurses, Education remained significant but the direction of trend changed (6 months: 3.51, 18 months: 3.57, 3 years 3.22)