Background

Health governance, leadership, and management are widely recognised as central to health system strengthening efforts and are considered important to ensure the quality and safety of health service delivery, patient satisfaction and increased staff performance and retention [1,2,3,4,5]. This recognition has prompted several calls for research focusing on the nature, quality, and contributions of governance, leadership and management to health systems [6]. Similarly, the World Health Organization (WHO) has identified strengthening governance, leadership, and management as a key strategy to improve health systems within low- and middle-income countries (LMICs) [7], where health systems remain severely under-resourced with regard to management capabilities.

While the terms ‘leadership’ and ‘management’ are often used interchangeably across health policy documents, health systems research, and within health services, others differentiate ‘leadership’– as the act of driving ‘change’ or providing ‘direction, alignment, and commitment’ across multiple actors in a health system–from the concept of ‘management’, which refers to the act of ‘supporting, resourcing, and facilitating day-to-day work’ through the effective mobilisation and utilisation of resources [8,9,10]. Governance, on the other hand, while also used interchangeably with ‘leadership’ or ‘stewardship’, more broadly refers to the set of processes– policies, structures, standards, norms, and practices– that guide decision-making, accountability, and ethical conduct within healthcare organisations, and often implies oversight by boards of directors and regulatory bodies [11]. In this way, management is considered (by some) to be more concerned with operations, general administration, and sufficient availability of both human and material resources. That said, governance, leadership and management are inextricable from quality health service delivery, which, among other factors, depends on leaders with strong managerial skills working to improve performance, accountability and decision-making of health organizations, [12].

While assessing health system governance tends to rely primarily on the use of frameworks [13, 14], several questionnaires and scales have been designed to assess management and leadership [3, 15,16,17,18,19]. The Multi-factor Leadership Questionnaire (MLQ), for example, has long been used as a measure of transformational, transactional, and passive-avoidant forms of leadership [20]. The MLQ, however, is designed to assess leadership styles, and therefore does not take into consideration other factors, such as resource and time management, that can also contribute to one’s perception of leadership and managerial support [9]. Similarly, the Leadership Practices Inventory [21], while inclusive of an ‘Observer’ version that allows others to rate the leadership of others, is only available at a fee, thus limiting its accessibility within more resource-constrained contexts. The Foundational Healthcare Leadership Self-assessment (FHLS) [22]; the International Leadership Scale (ILS) [23]; Evidence-Based Practice Nursing Leadership Scale [24] have also all been used within health care settings. A common critique of common management and leadership scales, however, is that like the MLQ and LPI, they were developed based on a Western, individualistic perspective, which may or may not be appropriate to assess leadership across other cultural contexts.

Which form of governance, leadership, and management is preferable in any given context is highly susceptible to socio-political and cultural norms, and speaks to the need to consider ‘the reciprocal influence actors have upon one another’s interests and priorities, and the enabling environment within the health-eco system’ [25]. Whereas governance strengthening efforts tend to take place at a national level, investment in leadership and management for health are frequently concentrated at a district (i.e., within district health management teams; DHMTs) or sub-district level (i.e., within large, district-based, referral hospitals), with the assumption that improvements will ‘trickle-down’ to primary health facilities [26,27,28]. Consequently, there is a dearth of research focused on health facility based leadership and management– despite its proximal influence within primary care settings and decentralized health systems [29]. Better methods to assess, and therefore regularly monitor, facility-level management are therefore necessary if we consider strengthening facility-level management as an equally important element of the governance, leadership, and management health system building block. The specific objective of this study was thus to develop a Facility-level Management Scale (FMS), and to test its reliability among facility-level health workers within LMICs.

Methods

This study leveraged PERFORM2Scale baseline data, collected through a multi-country survey of health workers. PERFORM2Scale is a multi-country research collaboration whose aim was to strengthen the management of health systems by applying participatory action research within three sub-Saharan African countries [30].

Study participants were comprised of health workers currently employed within a health facility and offering primary health care services. Participants were located across 138 health care facilities, spanning three districts in Uganda (October 2018), three districts in Malawi (November-December 2018), and three districts in Ghana (February-March 2018). All public primary care facilities within the nine selected districts were included across the three countries, except for private and non-governmental organization (NGO) facilities as these do not fall under the full jurisdiction of local governments.

Data was collected in 2018 using a self-administered paper-based questionnaire to health workers. Sample size varied slightly between countries. In Uganda, non-probability convenient sampling techniques were used, whereby all technical health workers that were present at the health facility and/or hospital on the day of data collection were invited to participate in the study. In Malawi, a two-stage probability sampling approach was used, whereby health workers eligible from the district and government facilities were listed for each district and then sampled proportion to the size of each facility’s workforce (n = 67). In Ghana, sample size was determined based on published sample size table (Israel 2009), with a precision level of ± 5%, confidence level of 95% and adjusted for non-response rate of 5% resulting a sample size of 252 health workers.

In total, 881 health workers completed an initial, self-report, version of the FMS across Ghana (n = 287; 32.6%), Malawi (n = 66; 7.5%) and Uganda (n = 528; 59.9%) as part of PERFORM2Scale’s baseline survey. Of these, n = 119 health workers self-identified as being in a managerial position within their health facility were excluded from the current analysis since the tool was designed to capture health workers perceptions of management. Of the remaining 762 health workers, 69.7% (n = 531) were female. The majority were nursing assistants (24.3%), nurses (19.0%), or community health nurses (9.7%). Written consent was obtained from all study participants and all surveys were completed in English as a language commonly spoken among those of high educational achievement across the three countries.

Development of the initial facility management scale (FMS)

An initial set of nineteen items were developed drawing from extant research on leadership and management for health systems strengthening in LMICs. The original items of the FMS are reported in Table 1. All items were rated using a five-point Likert scale, anchored by Strongly Disagree (= 1) and Strongly Agree (= 5). This original scale was found to have acceptable internal reliability in the current sample (α = 0.88).

Table 1 Exploratory factor analysis using principal component factor, showing only factors loadings > 0.40
Full size table

Missing data

The within-item missing data ranged from 2 to 6%. Twelve observations were dropped due to missing data on all the 19 items of the FMS scale. The pattern of missing data was established using Little’s Missing Completely At Random (MCAR) test [31]. The pattern of missing data was not MCAR (p < 0.001). The Ordered Logistic model was thus used to impute missing values.

Data analysis

The construct validity of the FMS was assessed using a hybrid factor analysis approach, as outlined by M Matsunaga [32]. First, cases were randomly assigned to one of two databases for the purposes of examining the underlying factor structure of the data. This resulted in n = 375 cases being included in each factor analysis, considered a ‘good’ sample size for the purposes of factor analysis [33]. The first n = 375 cases were subjected to Exploratory Factor Analysis (EFA) in order to summarise the data by reducing it to a fewer number of components, or factors [34]. Principal component analysis was carried out to explore the latent structure of the 19-item FMS using an oblique (Promax) rotation. A priori criteria for item retention were then applied. First, items were removed where they yielded a factor loadings of < 0.40 (whereby 16% of the variance is explained by the latent variable) [35]. We then identified and removed cross-loading items, or items that loaded considerably on two or more factors. Finally, those items with residual covariances were also removed.

The number of factors retained was determined based on Eigenvalues larger than 1.0, which were further confirmed by running a Monte Carlo PCA [36]. Where the Eigenvalue value generated by the EFA was larger than the criterion value from the Monte Carlo PCA, it was retained. Where the Eigenvalue from EFA was less than the criterion value, it was rejected. Items within each factor were then examined for communalities, prior to labelling each factor according to the set of highly loading components.

To test, or confirm, whether the latent structure of the FMS was an adequate representation of the observed data, the factor structure generated during the EFA was subsequently subjected to Confirmatory Factor Analysis (CFA). Specifically, we fitted the CFA model using maximum likelihood estimation and model fit was established using Root Mean Square Error Approximation (RMSEA), Standardized Root Mean Square Residual (SRMR), Comparative Fit Index (CFI), Tusker-Lewis Index (TLI), and the raw and modified chi-square (χ2) fit statistics [37]. In line with current conventions, CFI and TLI values greater than 0.90 were considered ‘good’ and values greater than 0.95 were considered ‘excellent’; RMSEA and SRMR values of less than 0.05 were considered to reflect excellent model fit, while values more than 0.08 were considered to reflect poor model fit [38, 39]. The same procedures were repeated on the second random sample to confirm construct validity of FMS scale. All factor analyses were conducted using STATA (Version 17). To evaluate the internal reliability of the FMS scale, Cronbach’s alpha was assessed both for each sub-scale and overall scale.

Convergent validity was evaluated by estimating the Average Variance Extracted (AVE) for each construct against its correlation with the other constructs. If the AVE was larger than the construct correlation with other constructs, the convergent validity was considered to be confirmed. Discriminant validity was established where Maximum Shared Variance (MSV) and Average Shared Squared Variance (ASV) were lower than Average Variance Extracted (AVE) for all the constructs.

Results

Exploratory factor analysis

The results of the EFA for all 19-items are presented in Table 1.

Overall, three items (Management considers gender issues when addressing staffing at the facility; My job description corresponds to the reality of my work; and Management actively seeks feedback from the community about the quality of care) were found to have a factor loading of less than 0.40 on any given factor and were thus removed from the scale. Finally, the results of the Monte Carlo PCA (see Table 2) suggested the retention of the first three factors, with Eigenvalues of 6.25, 1.39, and 1.29, respectively, resulting in the removal of three additional items (Management encourages us to pay attention to differences between women, men, boys, and girls when providing health services; Management assures that we are committed to do quality work; and Management pays attention that we treat patients with respect and dignity irrespective of their financial status).

Table 2 Results of the monte Carlo PCA, compared to Eigenvalues generated by the exploratory factor analysis
Full size table

Ultimately, Factor 1 included items that, when scored highly, describe a supportive style of management, and was thus labeled the ‘Supportive Management’ factor. Factor 2 included items that when scored highly are indicative of effective “resource management”, including the avoidance of stock-outs, and regular repairs of equipment. Finally, Factor 3 included items that, when endorsed, are indicative of supportive time management, including consideration for fair scheduling of duties and annual leave. This was labelled ‘Time Management’. Together, the resulting three-factor solution explained 51% of the variance in perceived facility managerial support.

Confirmatory factor analysis

Results of the CFA on the second half of the data file, carried out with the remaining 13 items (see Fig. 1), yielded good model fit (N = 381; χ2 = 256.8, df = 61, p < 0.001; CFI = 0.94; TLI = 0.92; RMSEA [95% CI] = 0.065 [0.057–0.074]; SRMR = 0.047). The factor loading for the 13 items were all positive, statistically significant, and of a robust magnitude (Table 3). Table 4 presents the FMS’ inter-factor correlations, which were all significant, moderate, and positive.

Fig. 1
figure 1

Three-factor correlated structure of the FMS tested using Confirmatory factor analysis. CFA model. Note Supmgt = Supportive Management, resmgt = Resource Management, timemgt = Time Management

Full size image
Table 3 CFA loading’s
Full size table
Table 4 Inter-factor correlations
Full size table

Discriminant and convergent validity

Table 5 presents the Composite Reliability (CR) and Average Variance Extracted (AVE) of the FMS. CR was found to be greater than AVE, in support of the reduced scale’s convergent validity. The AVE of each construct/factor was then compared to the inter-factor correlations and was found to be greater than its correlation with other constructs, in support of the scale’s convergent validity. Discriminant validity was confirmed by AVE > MSV and AVE > ASV.

Table 5 Convergent and discriminant validity of the FMS (13 item scale)
Full size table

Discussion

Health governance, leadership and management are a key component of any health system and play an important role in strengthening health systems and improving service performance [12, 40]. Within LMICs, health leadership and management strengthening efforts are most often concentrated at district level. Consequently, how health management is experienced or perceived at the level of the health facility is less well understood. Overall, results of the exploratory phase of this study suggest that health-facility-level management is best conceptualised as a three-factor, correlated model characterised by “supportive management”, “time management”, and “resource management”. The emergence of a “supportive management” as a component of health facility management is consistent with more translational approaches focused on building a relationship of trust and confidence between health workers and their managers [41, 42]. Similarly, supportive approaches are consistent with a growing literature which promotes encouraging teamwork and relationships and collective problem solving as key elements of supportive supervision [43,44,45]. Supportive management thus stands in contrast with more ‘top-down’ (i.e., authoritarian) approaches, or styles based on a system of rewards and punishment (e.g., transactional approaches), which are thought to be less effective for fostering health worker engagement and motivation [12]. Likewise, the emergence of the “time-management” and “resource-management” factors is consistent with the concept of ‘management’ as being more concerned with the day-to-day operations and the sufficient and timely available of human and material resources [8,9,10]. Taken together, the inclusion of all three factors within the revised version of the FMS is consistent with the idea of ‘leadership’ and ‘management’ as distinct, but related, concepts and that ‘good’ health managers are commonly perceived as those ‘managers who lead’ [46].

Results from the confirmatory phase of the study further suggest that the three-factor structure had acceptable construct, convergent, and discriminant validity, as well as good reliability. Taken together, the development of the FMS, as an easy to use, open-access, short, validated measure of health facility managerial support that can be easily transferred, translated, applied across cadres and within resource constrained settings, and therefore has the potential to improve our understanding and monitoring of staff perceptions of management and leadership at the level of the health facility. Practically, the sub-scales of the FMS can be examined independently, or the FMS items can be totalled as a sum score, whereby higher scores are associated with perceptions of stronger health facility management. Furthermore, the FMS could serve as a useful tool for higher-level management (i.e., district health management teams), helping them detect problematic managerial practices (i.e., around staffing or resource management) and strengthen facility management training programmes, with the ultimate goal to improve staff retention and the quality and safety of service delivery [1,2,3,4]. The development of the FMS further addresses calls for more rigorous approaches to scale development within human resource for health programming [47].

Like other measures of managerial practices (e.g., MLQ, LPI), the FMS can be completed by multiple raters, used across cadres, and with multiple subordinates, providing a more reliable view of a manager. The FMS, developed and validated with a sample of healthcare staff across three sub-Saharan African countries, however, further adds to our understanding of how managerial support is perceived within these contexts.

Limitations

The current study is not without limitations. First, the final FMS may not represent all possible domains of health workers perceptions about leadership and management. Other factors– some of which are measured within some of other scales– are not assessed within the FMS. For example, the ability to delegate, manage change, and create a positive work environment, are not captured by the items contained within the FMS. Second, the public health facilities across the three countries represent varying levels of care and may not be generalizable to all health workers; it is possible that had the scale refinement process been conducted in a different setting, a different set of indicators may have been retained. Thirdly, limited within-country sample sizes, specifically in Malawi, meant we could not ascertain measurement invariance across each context. We therefore strongly encourage that further validation of the FMS take place across other health cadres and across health workers working in both private and not-for-profit health facilities globally.

Conclusion

In comparison to current tools (e.g., MLQ, LPI), largely developed within Western settings, the FMS allows for the subjective measurement of perceived managerial support, validated across a sample of health workers from three different African countries. Simple and quick to administer, and available open-access (available from Appendix 1 and https://www.perform2scale.org/), the validated FMS has the potential to contribute towards a more accurate understanding of perceptions of managerial support, as a critical determinant of health system performance.