Predictors of job satisfaction and intention to stay in the job among health-care providers in Uganda and Zambia

Background: A shortage of competent health-care providers is a major contributor to poor quality health care in sub-Saharan Africa. To increase the retention of skilled health-care providers, we need to understand which factors make them feel satisfied with their work and want to stay in their job. This study investigates the relative contribution of provider, facility and contextual factors to job satisfaction and intention to stay on the job among health-care providers who performed obstetric care in Uganda and Zambia. Methods: This study was a secondary analysis of data from a maternal and newborn health program implementation evaluation in Uganda and Zambia. Using a Likert scale, providers rated their job satisfaction and intention to stay in their job. Predictors included gender, cadre, satisfaction with various facility resources and country. We used the Shapley and Owen decomposition of R2 method to estimate the variance explained by individual factors and groups of factors, adjusting for covariates at the facility and provider levels. Results: Of the 1134 providers included in the study, 68.3% were female, 32.4% were nurses and 77.1% worked in the public sector. Slightly more than half (52.3%) of providers were strongly satisfied with their job and 42.8% strongly agreed that they would continue to work at their facility for some time. A group of variables related to facility management explained most of the variance in both job satisfaction (37.6%) and intention to stay (43.1%). Among these, the most important individual variables were satisfaction with pay (20.57%) for job satisfaction and opinions being respected in the workplace (17.52%) for intention to stay. Doctors reported lower intention to stay than nurses. Provider demographics and facility level and ownership (public/private) were not associated with either outcome. There were also differences in job satisfaction and intention to stay between Ugandan and Zambian health-care providers. Conclusion: Our study suggests that managers play a crucial role in retaining a sufficient number of satisfied health-care providers providing obstetric care in two sub-Saharan African countries, Uganda and Zambia. Prioritizing and investing in health management systems and health managers are essential foundations for high-quality health systems. This study uses data from the evaluation of the Saving Mothers and Giving Life (SMGL) intervention [17]. The SMGL intervention was implemented in Uganda and Zambia from January to June 2012. This intervention was implemented in four districts—Kabarole, Kamwenge, Kibaale and Kyenjojo in Uganda and Mansa, Lundazi, Nyimba and Kalomo in Zambia. These districts were primarily rural, with a largely agricultural workforce. The districts were selected based on a high maternal mortality ratio, a low facility delivery rate and a high health-care provider shortage. The purpose of the intervention was to improve the quality of care during labor and delivery. The main program activities included are as follows: (i) providing new equipment and supplies, (ii) mentoring providers with experienced clinicians, (iii) upgrading the maternity wards and (iv) building new operating theaters to perform cesarean sections in Uganda. The core inputs and activities of the SMGL have been published elsewhere [18]. The design of the evaluation study was a quasi-random post-test-only comparison group; details of the evaluation have been previously published [19]. The data for the evaluation were collected from May 2013 to July 2013. The evaluators selected two comparison districts per country—Masinid and Kiryandongo in Uganda and Kapiri Mposhi and Kabwe in Zambia—based on their similarity with the intervention districts in terms of geography, health system infrastructure, health system utilization, and morbidity and mortality. Health facilities were included based on their delivery volume, urban/rural location, and whether they offer comprehensive emergency obstetric and newborn care. During the SMGL evaluation, three questionnaires were administered to maternal health-care workers: an obstetric knowledge test, a confidence questionnaire (gauging their reported confidence in performing 26 common obstetric tasks) and a job satisfaction questionnaire. The three instruments were pilot tested in non-study districts among providers in Uganda and Zambia prior to the start of data collection and revised accordingly. All clinicians available in the facility on the days of the study were invited to complete the survey. Eligible health-care providers included all clinicians (i.e. doctors, nurses, midwives, nurse assistants and clinical officers) who provided obstetric care. To maximize the response rate, interviewers returned to the facility on different days to accommodate health-care providers’ schedules. The job satisfaction questionnaire measured the workers’ satisfaction with various aspects of their work, including the providers’ opinions related to their work environment (such as availability of functional equipment and adequate clinical supervision), as well as overall satisfaction about their current work and their intention to stay in their current health facility for some time (Supplementary Appendix A). The survey was administered in person in English. SurveyCTO software was used to enter the data on Galaxy Nexus t computers. Information on providers’ demographics and characteristics of facilities were also collected. Results from the two other instruments—knowledge test and confidence assessment—have been published elsewhere [20]. We selected two provider outcomes: job satisfaction and intention to stay. Job satisfaction was assessed using the question: ‘In general, I am satisfied with this job.’ Intention to stay in the job was assessed using the question: ‘If it were up to me, I would continue to work for this hospital/clinic for quite some time.’ Both questions were measured on a 4-point Likert scale—strongly disagree, somewhat disagree, somewhat agree and strongly agree. We treated them as continuous variables for our analysis ranging from 1 for strongly disagree to 4 for strongly agree. We included six categories of covariates at the provider and facility levels known to influence provider intention to stay and job satisfaction: (i) demographics, (ii) cadre, (iii) facility characteristics, (iv) perceptions on the work environment related to inputs, (v) perceptions on the work environment related to management and (vi) context of the facility. Demographics included age, gender and on-site training. Age was treated as a continuous variable. To account for non-linearity, a quadratic term for age was also included. Amount of training received in the past year was measured as the total number of days during which providers reported receiving on-site trainings. Cadre included seven types of providers based on the length of training: (i) nurse assistant, (ii) enrolled nurse, (iii) enrolled midwife, (iv) registered nurse, (v) registered midwife, (vi) clinical officer and (vii) general doctor, doctor specialist and medical licentiate. Nurse assistants are trained for about 6 months and exist only in Uganda. Enrolled nurses and enrolled midwives are trained for 2–3 years. Enrolled nurses are similar to licensed practical nurses in the United States. Registered nurses and registered midwives are trained for 3–4.5 years. Clinical officers receive 3 years of training. Doctors are typically trained for 5–7 years. Medical licentiates are clinical officers who received additional training so that they can perform several tasks that a doctor would typically perform [21]. Given the similar years of training, medical licentiates were included in the same group as doctors. Facility characteristics included facility type and public vs. private ownership. Facility type was based on availability of services and included basic emergency obstetric and neonatal care (BEmONC) facilities and comprehensive obstetric and neonatal care (CEmONC) facilities. BEmONC facilities perform seven basic functions: (i) administration of parenteral antibiotics, (ii) administration of uterotonic drugs for active management of the third stage of labor and prevention of postpartum hemorrhage, (iii) use of parenteral anticonvulsants for the management of preeclampsia/eclampsia, (iv) manual removal of placenta, (v) removal of retained products, (vi) assistance of vaginal delivery and (vii) basic neonatal resuscitation [22]. CEmONC facilities must perform the seven basic BEmONC functions in addition to cesarean sections and blood transfusion [22]. Private ownership included both for-profit and not-for-profit facilities. Two categories of covariates related to the work environment—inputs and management—were included. The input category grouped variables related to infrastructure, staffing and pay and included three variables measuring providers’ opinion on the functioning equipment and infrastructure to perform their duties, the level of staffing and whether they were satisfied with their current pay compared to similar jobs in other organizations. The management category included four variables measuring provider’s opinion on clinical supervision, whether their workload was manageable, whether their facilities provided adequate in-service (continuing) education to improve their clinical skills and whether they felt like their opinions were respected at work. Questions related to inputs and management of work environment were measured in a 4-point Likert scale—strongly disagree, somewhat disagree, somewhat agree and strongly agree. Providers were assigned 0 point for strongly disagree and somewhat disagree and 1 point for strongly agree and somewhat agree. Context covariates included urban/rural location, intervention/control districts and country (Uganda or Zambia). In order to estimate the proportion of the variance in provider satisfaction and intention to stay explained by each covariate and group of covariates, we used the Shapley and Owen decomposition of R2 method using the rego command in Stata. This method decomposes the R2 (the ratio of the explained sum of squares to the total sum of squares) of an ordinary least square model indicated by Shapley and Owen values [23]. These values are equivalent to partial R2 and allow us to assess the explanatory power of individual regressors or groups of regressors in addition to the full model R2. The unit of analysis was the provider, and the standard errors were adjusted for clustering at the facility level. We also looked at unadjusted correlations between the two outcome variables and the work environment covariates using Pearson’s correlation coefficient. All analyses were conducted in November 2019 using Stata SE version 16.0 (StataCorp, College Station, TX). Since this study was a secondary analysis of de-identified data, it was deemed to be non-human subjects research under the Harvard T.H. Chan School of Public Health Institutional Review Board (IRB) policy. The original study was approved by IRB at Columbia University in the United States, Makerere University School of Public Health and the National Council for Science Technology in Uganda, and ERES (Excellence in Research Ethics and Science) Converge Research Ethics Committee and Ministry of Health in Zambia [19]. Consent was obtained from the health-care provider interviewed, and survey was completed in private rooms to ensure privacy. Data used for the study were stored in a secured folder with limited access.