Study design and participants
A cross-sectional study was carried out in December 2018. Eight-hundred thirteen health facility workforce were selected, randomly from 32 health facilities located in northwest Ethiopia. The workforce that took integrated LMG in-service capacity building training so far were excluded.
Data were collected using a structured multi-item questionnaire. It comprised the participants’ basic characteristics (Table 1) and items that potentially measured competence to lead, manage, and govern (dependent variable). The measuring items were adapted among activities clustered in the integrated leading, managing, and governing framework(12).
The test stimuli (psychometric properties) of the questionnaire was refined through rigorous debriefing sessions, focused on instrument clarity and validity. In this process, five specialists of health service management, of whom three were from civil service and two from the academic spheres were involved.
All of the measuring items were rated with a five-point Likert scale, ranging from 1 = very low to 5 = very high. The data related to measuring items (Table 2) were checked for inter-correlation of 0.3 and greater, intra-item consistency of 0.7 and greater(19), communality of 0.5 and greater(20, 21), and complex structure that is whether any factor had resided on more than one item with factor loadings of .4 and greater(22) using factor analysis. In this process, six items were removed from the original dataset. Of which, two items: looking for best practices and match deeds to words were removed due to violating the rule of communality, and the other four items: set annual and strategic plan, allocate adequate resources, provide accountability and authority, and provide appropriate feedback were removed due to violating the rule of complex structure. It showed that the dataset was reduced to a 20-item dataset. Note that the six measuring items trimmed from the original dataset due to violating the rules of communality or complex structure were taken as predictors.
The dependent variable was computed from the 20-item dataset. The computed values were leveled into four ordinal categories: low, moderate, high and very high that represented scores of <60, 60-79.99, 80-94.99 and >95 respectively. These scales were taken from the Ethiopian ministry of health workforce performance appraisal guideline (unpublished work).
The relationship between the dependent variable and its predictors was modeled using ordinal logistic regression analysis with logit link function. Model fitting information tested by (-2Log Likelihood) was significant at p<0.001. The consistency of the observed data tested with Pearson chi-square goodness-of-fit was remained satisfactory with p = 1. The explained variance of the dependent variable from the predictors was tested by pseudo-r-squared value (Nagelkerke’s R2 =0.765), which indicated a strong association. The test of parallel lines or testing proportional odds assumption that is testing whether the location parameters (slope coefficients) of predictors were the same across outcome variable categories was tested by (-2Log Likelihood) and became non-significant with p = 0.487. This showed that the slope coefficients were the same across response categories, which told that there had no evidence to reject the parallelism hypothesis. Here, to interpret the impact of individual predictors in a better way, odds ratios with 95% CI were calculated from the odds.
Ethical clearance with a protocol record 090/18-04 was secured from the institutional review board of Bahir Dar University. Each participant provided written consent. The process was strictly anonymous and questionnaires completed were stored in a locked cabinet.