Key factors influencing motivation among health extension workers and health care professionals in four regions of Ethiopia: A cross-sectional study

Background Although Ethiopia has improved access to health care in recent years, quality of care remains low. Health worker motivation is an important determinant of performance and affects quality of care. Low health care workers motivation can be associated with poor health care quality and client experience, non-attendance, and poor clinical outcome. Objective this study sought to determine the extent and variation of health professionals’ motivation alongside factors associated with motivation. Methods We conducted a facility based cross-sectional study among health extension workers (HEWs) and health care professionals in four regions: Amhara, Oromia, South nations, and nationalities people’s region (SNNPR) and Tigray from April 15 to May 10, 2018. We sampled 401 health system workers: skilled providers including nurses and midwives (n = 110), HEWs (n = 210); and non-patient facing health system staff representing case team leaders, facility and district heads, directors, and officers (n = 81). Participants completed a 30-item Likert scale ranking tool which asked questions across 17 domains. We used exploratory factor analysis to explore latent motivation constructs. Results Of the 397 responses with complete data, 61% (95% CI 56%-66%) self-reported motivation as “very good”or “excellent”. Significant variation in motivation was seen across regions with SNNPR scoring significantly lower on a five-point Likert scale by 0.35 points (P = 0.003). The exploratory factor analysis identified a three-factors: personal and altruistic goals; pride and personal satisfaction; and recognition and support. The personal and altruistic goals factor varied across regions with Oromia and SNNPR being significantly lower by 0.13 (P = 0.018) and 0.12 (P = 0.039) Likert points respectively. The pride and personal satisfaction factor were higher among those aged > = 30 years by 0.14 Likert scale points (P = 0.045) relative to those aged between 19-24years. Conclusions Overall, motivation was high among participants but varied across region, cadre, and age. Workload, leave, and job satisfaction were associated with motivation. We conducted a facility based cross-sectional study among health extension workers (HEWs) and health care professionals in four regions: Amhara, Oromia, South nations, and nationalities people’s region (SNNPR) and Tigray from April 15 to May 10, 2018. We adapted a motivation tool developed and validated among community health workers (CHWs) in Uganda [19], making minor changes to wording to suit the Ethiopian context. The tool consisted of 17 questions. We added 8 additional questions from a health worker motivation evaluation conducted in Tanzania to explore extrinsic motivating factors in more depth [20]. Finally, we included 5 further questions relating to activities related to the quality improvement programme being implemented in our sample. The final tool is shown in Table 1. All items had Likert scale response options where 1 = strongly agree, 2 = agree, 3 = neutral, 4 = disagree, 5 = strongly disagree. We sampled 401 health system workers: skilled providers including nurses and midwives (n = 110), HEWs (n = 210); and non-patient facing health system staff representing case team leaders, facility and district heads, directors, and officers (n = 81). The survey was part of a baseline evaluation of a quality improvement (QI) program delivered by the Institute for Healthcare Improvement (IHI) in partnership with the Ministry of Health Ethiopia (MOH). Although the sampling frame of this study is based on the IHI program, data are from pre-intervention baseline data collection, and we do not expect motivation to have been influenced by the intervention at this point. The IHI program was implemented in 19 districts: 7 in Oromia, 5 in Amhara, 5 in SNNPR, and 2 in Tigray. Using a random number generator, we randomly selected one intervention district from each region (Jimma Town, Wogera, Chena, and Degua Tembien respectively). We added one additional randomly selected district in Amhara because Wogera would not have yielded 50 eligible respondents—our target for each region. We further purposively sampled two additional districts from Oromia and SNNPR (Bunno Bedelle and Chencha respectively) where qualitative evaluative work took place, to triangulate findings in a larger evaluation of IHI’s QI program. Data collection was conducted by seven research assistants who received one week training at the start of the data collection process and then were matched to their home regions where they have experience working in and speak local language to assist with community entry and mitigate language issues. The data quality was assured by using validated tools, trained data collectors, and conducting interviews in the local languages. The survey was piloted out of 19 district health office staff in December 2017. No changes were made to the survey between piloting and the final survey as it was understood well by participants, assessed through debriefing interviews after survey completion. In each district, we mapped the hospital, all health centres and health posts, and approached the district health office for permission letters that was later obtained. In each hospital and health centre, we obtained a list of all eligible health care professionals and HEWs. We then randomly selected participants for interviews. In each district, we interviewed around 50 participants across a range of health worker and management cadres, including the heads or clinical directors of the district, each hospital, and each health centre. We interviewed around four maternal and child health care providers from the hospitals and two from each health centre, and around five HEWs from each health centre. A target sample size of 50 respondents per region was chosen, based on the primary research question of assessing changes in motivation as measured by Likert scale questions, in line with a rule of thumb in exploratory factor analysis that 50 participants per cluster is a reasonable sample size to detect differences across clusters [21]. In each hospital or health centre, we obtained a list of all eligible MNH providers and randomly selected participants for interviews. Their names were written in alphabetical order next to a column of randomly generated numbers and interviewers sequentially chose participants from the smallest random number upwards until the requisite number of participants was reached. If participants were not available, we sought to arrange interviews via phone and returned to the facility up to three times before classifying them as unreachable and selecting the next worker from the list. Data were entered on tablet computers using Open Data Kit software (www.opendatakit.org) and exported to STATA V.13. We categorized the responses according to sociodemographic factors using counts and percentages as appropriate. To explore the underlying correlations and associations and identify factors within the survey items, we first re-coded the survey items from the 5-point Likert scale from poor to fair, good, very good and excellent to a continuous variable from 1 (poor) to 5 (excellent). Next, we used the re-coded items in an exploratory factor analysis. For the exploratory factor analysis, we first removed items from our list of 30 questions which had poor psychometric performance, removing items which had more than 10% missing data, items which were given the same score of over 80% of participants, and items which did not load on any factors up to a level of 0.4 in initial factor analysis. We used a threshold of 0.4 to assume a strong relationship with a factor, and the optimal number of factors was established through a scree test and multiple runs [22, 23]. We used maximum likelihood ProMax oblique rotation to reduce the number of variables with high loadings and to allow factors to be correlated. Construct validity was indicated by loading at least three items per factor and absence of substantive cross-loading. We explored the association of overall motivation and with the motivation factors identified with overall job satisfaction and demographic and structural factors including gender, location, cadre, age, perceived gross salary, work experience, using univariate and multivariate ordinary least squares regression models, and show ordered logit model results in the S1 & S2 Appendices. Variables having p value ≤ 0.2 in the bivariate analysis were fitted into a multivariable regression model to control the effects of confounding. Normality assumptions were checked by Schapiro—Francia W tests, and variance inflation factor estimates were generated for regressors [24]. Average job satisfaction was assessed by re-coding the 5-point Likert scale ranging from least satisfied with their job (1) to most satisfied (5) as a continuous variable. Written informed consent was obtained from all participants. The study was undertaken with ethical approval from the Observational Research Ethics Committee of the London School of Hygiene and Tropical Medicine (Ref: 14429) and a program evaluation waiver from the Ethics Committee of the Ethiopian Public Health Association (Ref: EPHA/OG/2380).