The effects of midwives’ job satisfaction on burnout, intention to quit and turnover: a longitudinal study in Senegal

Background: Despite working in a challenging environment plagued by persistent personnel shortages, public sector midwives in Senegal play a key role in tackling maternal mortality. A better understanding of how they are experiencing their work and how it is affecting them is needed in order to better address their needs and incite them to remain in their posts. This study aims to explore their job satisfaction and its effects on their burnout, intention to quit and professional mobility.Methods: A cohort of 226 midwives from 22 hospitals across Senegal participated in this longitudinal study. Their job satisfaction was measured from December 2007 to February 2008 using a multifaceted instrument developed in West Africa. Three expected effects were measured two years later: burnout, intention to quit and turnover. Descriptive statistics were reported for the midwives who stayed and left their posts during the study period. A series of multiple regressions investigated the correlations between the nine facets of job satisfaction and each effect variable, while controlling for individual and institutional characteristics.Results: Despite nearly two thirds (58.9%) of midwives reporting the intention to quit within a year (mainly to pursue new professional training), only 9% annual turnover was found in the study (41/226 over 2 years). Departures were largely voluntary (92%) and entirely domestic. Overall the midwives reported themselves moderately satisfied; least contented with their ” remuneration” and ” work environment” and most satisfied with the ” morale” and ” job security” facets of their work. On the three dimensions of the Maslach Burnout Inventory, very high levels of emotional exhaustion (80.0%) and depersonalization (57.8%) were reported, while levels of diminished personal accomplishment were low (12.4%). Burnout was identified in more than half of the sample (55%). Experiencing emotional exhaustion was inversely associated with ” remuneration” and ” task” satisfaction, actively job searching was associated with being dissatisfied with job ” security” and voluntary quitting was associated with dissatisfaction with ” continuing education” .Conclusions: This study found that although midwives seem to be experiencing burnout and unhappiness with their working conditions, they retain a strong sense of confidence and accomplishment in their work. It also suggests that strategies to retain them in their positions and in the profession should emphasize continuing education. © 2012 Rouleau et al.; licensee BioMed Central Ltd. The present study is a secondary investigation within the larger study on maternal mortality, the QUARITE cluster-randomized trial [66], which seeks to evaluate the efficacy of the Gesta International Program in improving maternal health outcomes in both Senegal and Mali (2007–2011). Data on the individual characteristics and job satisfaction of maternity personnel were collected as part of the QUARITE study from December 2007 to February 2008 (see additional file 2 for the questionnaire). The second data collection phase on burnout, intention to quit and turnover of midwives took place during subsequent site visits over the period from December 2009 to February 2010 (questionnaire available upon request). The two-year time lapse allowed for a large turnover sample and was not expected to eliminate the associations between job satisfaction and its effects. Hospital inclusion criteria: the 24 health facilities included in this study (from the QUARITE trial) were selected from 26 eligible public sector reference hospitals in Senegal based on four criteria: having functional operating rooms, carrying out more than 800 deliveries annually, having written consent to participate in QUARITE provided by local authorities and not having a pre-existing structured program for carrying out maternal death audits. A total of six capital, ten regional and eight district level hospitals were retained. This sample is considered representative of the variety of the contexts (urban versus rural) and of the levels of care (primary versus secondary referral health facilities) found in Senegal (See additional file 3 for a map of the hospital sites included in this study). Midwife inclusion criteria: the cohort of midwives in this study included all those who were employed in the targeted public hospital maternity wards at the time of the data collection on job satisfaction (n = 235), with the exception of those who declined the invitation to participate in either phase of the investigation (T0, n = 6; T1, n = 2) and those who left more than 5% of their questions blank in the burnout questionnaire (n = 1). Midwives who left their jobs for non-voluntary reasons (death and retirement, n = 3) were excluded from the subsequent analyses involving turnover (voluntary turnover n = 38). Therefore, a total of 226 midwives were followed for the 2 years of this study, and 185 remained in their jobs to participate in the second phase. Recruitment was made possible by collaborating with the data collection team of QUARITE and the authorities of the hospitals and maternity units, who facilitated access to the health care personnel and made the initial contact with the eligible midwives to inform them of the aims of the study and invite them to participate. In general, very high levels of participation in both phases of the investigation (T0: 229/235 (97.4%), T1: 185/188 (98.4%)) were made possible without monetary compensation because of the lasting and functional collaboration between the parties and the familiarity of the staff with the aims and procedures of the QUARITE project. One on one interviews were also made more convenient for midwives as they were done either at work or at their homes, depending on their preference. The socio-demographic characteristics in the sample documented were: age (years), tenure (years), number of years in the profession (years), education level (middle school, high school or higher), job status (permanent civil servant or contractual), level in the professional hierarchy (head midwife/unit chief or midwife), as well as information about the individual’s perception of the availability of attractive job alternatives (unlikely, possible or certain), the type of hospital where they work (capital, regional or district) and the person doing the interview (D.R. or B.M.). Job satisfaction can be evaluated by a variety of instruments, either global or multi-facetted, though no measurement tool is considered a “gold standard” [67]. In this study, it was measured with a multifaceted instrument developed by previous work with healthcare professionals in Mali and Burkina Faso [68,69]. An instrument composed of 42 items (with Likert scales ranging from 1 (very unhappy) to 5 (very happy)) grouped into six job facets was used for the data collection. The instrument was then restructured through several steps of content and reliability validation into a format with 29 items divided into nine job facets: “remuneration” (two items), “work environment” (five items), “workload” (four items), “tasks” (three items), “working relationships” (five items), “continuing education” (two items), “management” (four items), “morale” (two items) and “job security” (two items) (See additional file 4 for a description of the instrument). The 9 facet model showed satisfactory content validity according to criteria set out by Van Saane et al. [67]. The Maslach Burnout Inventory is considered the “gold standard” in burnout research [36]. The questionnaire [35] consists of 22 items divided into three sub-scales: emotional exhaustion (EE), depersonalisation (DP) and diminished personal accomplishment (DPA). Scores on the three burnout dimension were then categorised into “high,” “average” and “low” according to cut-off levels indicated in Maslach and Jackson’s MBI Manual [35], which are equivalent to the tertiles of a normative distribution of scores from a large sample of North American nurses and doctors. The French-language version of the MBI used in this study was translated and validated by Dion and Tessier [70] using Quebec samples of 260 day-care workers and 123 nurses. Their results gave a positive assessment of the psychometric qualities of their translation, including good internal consistency, long-range stability, factor validity and convergent validity (high correlations with measures of depression, anxiety and stress). Although the instrument has never been tested in a Francophone West African setting, evidence of its psychometric robustness, the cross-national validity of the original instrument as well as discussions with key informants during the piloting phase gave us assurances that it would appropriate for use in this study. Midwives’ intention to quit was assessed by a question about whether they were actively searching for alternatives with three answer options: not searching, intention to search, and actively searching. To provide additional information, questions were also included about the immediacy of their intention to quit and their preferred job alternative. Finally, turnover was measured using health personnel registries and informal interviews with remaining midwives, which included specific information about the timing and destination of the departure. Turnover was then classified as voluntary or involuntary, as well as differentiated among intra-sector (movements within the public health system), inter-sector (movements outside of the public health system) and international movements [71]. As a complementary study included in the “satisfaction” component of the QUARITE trial, this study benefited from ethical approval by both the research ethics committee of the Sainte-Justine University Hospital Center in Montreal (renewed annually, protocol # 2425) and the national ethics committee of Senegal. Participants in the hospitals were selected on the basis of written consent provided by local authorities at the hospital level, individual signed consent at the time of the first phase of the study and confirmatory oral consent during the second phase. Eligible midwives were recruited on a voluntary basis and informed at each step of the option to withdraw from the study without consequences. For the data collected in phase two, missing values were replaced with the mode of the relevant questions and individuals with more than 5% of their questions unanswered were eliminated (n = 1). For burnout, and intention to quit, separate principal component analyses (PCA) were carried out to select the dimensions of each variable that best discriminate individuals amongst one another (significant discrimination along the first two components). All were found to have sufficient factor loading (≥ 0.3) and were retained for further analyses. To prevent collinearity between the co-variables in the regressions, “number of years in the profession” was removed due to its strong correlation with “age” (Pearson’s correlation R2 = 0.61, P ≤ 0.0001). In multiple linear and logistical regressions, collinerarity was verified by variance inflation factor (VIF > 5), tolerance (> 30) indices and correlation matrices respectively. The socio-demographic characteristics of the midwives who left and those who stayed were compared using ANOVAs or Chi-square tests (likelihood ratio). Since age and years in current position variables failed to normality, non-parametric ANOVAs (Wilcoxon/Kruskall-Wallis test) were used. To assess the relationship between burnout (response variable) and job satisfaction (explanatory variable), a series of multiple linear regressions were performed between each dimension of burnout and facets of job satisfaction separately (3*9 =27). These regressions were controlled by the following co-variables: age, tenure, type of institution, educational attainment, rank, employee status and interviewer. For each burnout dimension, the job satisfaction facets that showed a P value less than 0.10 were then entered simultaneously into a single multiple regression controlling for the same co-variables. A similar two-step procedure was performed between job search, turnover (response variables) and job satisfaction (explanatory variable), using a multiple logistical regression. The controlling variables were the same as above, except that “perception of alternatives” was also controlled in the regressions with job search. To allow for logistical regressions, the job satisfaction facet variables were categorised into the lower 25th percentile (least satisfied) versus the rest. Similarly, the three possible answer options of the job search activities item were merged into two modalities, in accordance with their meaning: not searching or intent to search versus actively searching. The two-step procedure for the multiple regressions was chosen because of the exploratory aims of the study. In all analyses, outliers were eliminated on the basis of the Cook’s D Influence test (coefficient greater than 0.2). All analyses were performed using SPSS 17 and/or JMP7 statistical software, and with a significance level of alpha = 0.05.