Vallières et al. Hum Resour Health (2021) 19:73 https://doi.org/10.1186/s12960-021-00617-9 RESEARCH Open Access Determinants of safety climate at primary care level in Ghana, Malawi and Uganda: a cross-sectional study across 138 selected primary healthcare facilities Frédérique Vallières1, Paul Mubiri2, Samuel Agyei Agyemang3, Samuel Amon3, Jana Gerold4,5* , Tim Martineau6, Ann Nolan1, Thomasena O’Byrne1, Lifah Sanudi7, Freddie Sengooba2 and Helen Prytherch4,5 Abstract Background: Safety climate is an essential component of achieving Universal Health Coverage, with several organi- sational, unit or team-level, and individual health worker factors identified as influencing safety climate. Few studies however, have investigated how these factors contribute to safety climate within health care settings in low- and middle-income countries (LMICs). The current study examines the relationship between key organisational, unit and individual-level factors and safety climate across primary health care centres in Ghana, Malawi and Uganda. Methods: A cross-sectional, self-administered survey was conducted across 138 primary health care facilities in nine districts across Uganda, Ghana and Malawi. In total, 760 primary health workers completed the questionnaire. The relationships between individual (sex, job satisfaction), unit (teamwork climate, supportive supervision), organisa- tional-level (district managerial support) and safety climate were tested using structural equation modelling (SEM) procedures. Post hoc analyses were also carried out to explore these relationships within each country. Results: Our model including all countries explained 55% of the variance in safety climate. In this model, safety climate was most strongly associated with teamwork (β = 0.56, p < 0.001), supportive supervision (β = 0.34, p < 0.001), and district managerial support (β = 0.29, p < 0.001). In Ghana, safety climate was positively associated with job satis- faction (β = 0.30, p < 0.05), teamwork (β = 0.46, p < 0.001), and supportive supervision (β = 0.21, p < 0.05), whereby the model explained 43% of the variance in safety climate. In Uganda, the total variance explained by the model was 64%, with teamwork (β = 0.56, p < 0.001), supportive supervision (β = 0.43, p < 0.001), and perceived district managerial sup- port (β = 0.35, p < 0.001) all found to be positively associated with climate. In Malawi, the total variance explained by the model was 63%, with teamwork (β = 0.39, p = 0.005) and supportive supervision (β = 0.27, p = 0.023) significantly and positively associated with safety climate. Discussion/conclusions: Our findings highlight the importance of unit-level factors—and in specific, teamwork and supportive supervision—as particularly important contributors to perceptions of safety climate among primary health workers in LMICs. Implications for practice are discussed. Keywords: Safety climate, Primary health, Low- and middle-income countries *Correspondence: jana.gerold@swisstph.ch 4 Swiss Tropical and Public Health Institute, Socinstrasse 57, 4002 Basel, Switzerland Full list of author information is available at the end of the article © The Author(s) 2021. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licens es/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Vallières et al. Hum Resour Health (2021) 19:73 Page 2 of 11 Background information flows and clinical incidents in Sweden [9], Recent decades have seen progress towards Universal with quality management systems emerging as an impor- Health Coverage (UHC) across many low- or middle- tant correlate of both perceived teamwork and safety income countries (LMICs), mostly through an expansion climate within European hospitals [10]. Within LMICs, of basic health services and strengthening of primary and particularly within decentralised contexts, manage- health care. Changing health needs and growing expec- rial responsibilities tend to be held within district-level tations from residents of LMICs however, suggests that health management teams (DHMTs), who oversee the increasing physical and financial access to, and coverage delivery of services (i.e. planning, budgeting, fundraising, of, health care remains insufficient, with a need to also monitoring); management of health workforce and other improve the quality of existing health systems [1]. Con- resources, including through paying attention to quality sequently, increasing attention has been given to the and safety; as well as stakeholder coordination (partners, importance of balancing efficiency of delivery, with the citizens, patients and other service-users, etc.). Manage- need to deliver high-quality, safe patient care. rial support is therefore considered particularly impor- Safety culture, as one aspect of an organisation’s cul- tant to ensure that primary health care centres have all ture, has been defined as “the product of individual and of the resources required for system performance and the group values, attitudes, perceptions, competencies, and delivery of quality health services in LMICs [11]. patterns of behaviour that determine the commitment to, and the style and proficiency of, an organisation’s health Unit‑level factors and safety management” (p.1) [2]. Furthermore, Sexton At unit or team-level, factors such as supervision, and et  al. [3] indicate that when examining group-level per- more specifically supportive supervision [12, 13], is ceptions, the most appropriate term to use is climate widely recognised as important for the improvement of (e.g., safety climate, or teamwork climate), in reference service quality across a range of primary health care ser- to the more readily measurable aspects of safety cul- vices [14–17]. Previous research suggests a strong posi- ture, and as opposed to other aspects of culture such as tive relationship between safety climate and occupational behaviour and values. In this way, safety climate acts as safety at the unit-level [18, 19], in support of group-level an emergent property; a social construct, characterising factors as an important determinant of climate safety. groups of individuals based on their shared perceptions Likewise, good collaboration with co-workers and an of enacted policies and practices that serve as an indica- environment that encouraged safety reporting were tor of the true priority of safety against other organisa- found to be positively associated with safety attitudes tional goals [4]. [20]. Consequently, team training and other unit-based The factors that influence safety climate are, by exten- strategies that promote teamwork processes, such as sion, multiple and complex, with an increasing amount of cooperation, open communication, and leadership, are literature focused on identifying its various antecedents, often seen as a key strategy to improve the work environ- moderators, mediators, and outcomes [5]. Yu and Liang ment and patient safety [19]. [6], for example, highlight a number of indicators of safety climate, including working conditions, teamwork Individual factors climate, job satisfaction, stress recognition, and percep- Finally, among individual factors, previous research tions of management. Likewise, Vincent and colleagues report that cadre plays a role, with attitudes towards [7] describe several factors that influence safety and qual- patient safety differing across doctors, nurses, and allied ity in clinical practice according to organisational-level health professionals, while attention is also called to vari- factors such as staffing levels, workplace and manage- ations according to the culture in the country of train- rial support; unit or team-level factors such as teamwork ing [21–23]. Age has been identified as an important and supervision; as well as individual-level factors such determinant of safety climate, with older staff expressing as health worker job satisfaction and perceptions of more favourable safety attitudes [23]. Likewise, individual management. health worker knowledge, motivation [24], and job satis- faction have been associated with health worker percep- Organisational‑level factors tions of safety climate [25]. Alsalem et  al. [8] reinforce that safety climate refers to Taken together, there is broad agreement that the con- the ability of healthcare organisations to learn effectively cepts of both quality and safety need to be investigated from adverse events and implement preventative meas- within the contexts and systems within which errors and ures to reduce related harm to patients. At an organisa- adverse events occur [8]. While patient safety has been tional level, managerial safety practices have been found widely explored within hospital settings however, fewer to mediate the relationship between safety procedures/ studies to date have focused on patient safety within V allières et al. Hum Resour Health (2021) 19:73 Page 3 of 11 primary health care settings [26]. Moreover, most of the Table 1 Primary health care facilities sampled within each research examining determinants of safety climate within country primary care settings to date has taken place in higher Country Districts Health facilities* Total income countries, including the upper-middle-income health countries of Iran [27] and Brazil [28]. Finally, and given facilities the strong cultural component to safety climate, with Ghana Fanteakwa Community health and plan- 11 very practical implications for managers and practition- ning services (CHPS) (n = 6) ers dealing with multi-national organisational contexts, Health centres (n = 3) there is a need for more research examining safety- Faith-based health centre (n = 1) related perceptions across non-Western settings [29]. In Hospital (n = 1) light of these gaps, the current study sought to examine Suhum CHPS (n = 11) 17 the relationship between key organisational, team and Health centres (n = 5) individual-level factors and safety climate across primary Hospitals (n = 1) health care centres in Ghana, Malawi and Uganda. Yilo Krobo Health centres (n = 5) 7Polyclinics (n = 2) Uganda Luwero Level IV (n = 5), 27 Methods Level III (n = 13) The current baseline research took place within the Level II (n = 9) broader remit of the PERFORM2Scale (https:// www. Wakiso Hospital (n = 1) 28 Level IV (n = 5) perfor m2sca le.o rg) programme, as part of regular pro- Level III (n = 22) gramme evaluation. Drawing on previous work [30, Nakaseke Hospital (n = 1) 19 31], PERFORM2Scale facilitates the implementation of Level IV (n = 2) a management strengthening intervention (MSI) pro- Level III (n = 7) Level II (n = 9) cess with district health managers, as those responsible Malawi Dowa Hospital (n = 1) 12 for the managerial oversight of primary care level health Health centres (n = 11) workforce. Ntchisi Hospital (n = 1) 7 Health centres (n = 6) Study design Salima Hospital (n = 1) 10 This study is based on a cross-sectional, self-report sur- Health centres (n = 9) vey conducted in 138 health care facilities across nine Total 138 districts: three districts in Uganda (October 2018), three *In Ghana, CHPS offer community health activities and basic out-patient care districts in Malawi (November–December 2018), and services; health centres are the main providers of primary health care, offering out-patient care services, laboratory services, antenatal care and basic obstetric three districts in Ghana (February–March 2018). All and postpartum services; hospitals and polyclinics offer all services provided by public primary care facilities within the nine selected dis- health centres in addition to comprehensive emergency obstetric and newborn care services; blood transfusion and operative care. In Uganda, Level II health tricts were included across the three countries, with the centres provide basic out-patient care services, Level III provide antenatal care exception of private and non-governmental organisation and basic emergency obstetric care and postpartum services, and Level IV offer all services provided at Level III in addition to operative care and laboratory (NGO) facilities; as these are not under the full jurisdic- services. Hospitals provide comprehensive emergency obstetric and newborn tion of the DHMTs. Table 1 summarises the characteris- care services, blood transfusion, and laboratory services. In Malawi, health tics of the health facilities surveyed within each district. centres provide basic out-patient primary care services, whereas hospitals offer both inpatient and outpatient care, often acting as referral centres for health centres. Staffing wise, more rural health centres would consist of a medical Study population assistant, nurses, health surveillance assistants, and environmental health officers, whereas hospitals would include cadres spanning health surveillance Across all three countries, participants were health work- officers to specialists ers currently employed within a health facility offer- ing primary health care services. Employment covered 5%, resulting in a sample size of 252 health workers. Ulti- both frontline staff and facility managers. The sampling mately, 241 participants were recruited (n = 182, 75.5% of study participants varied slightly between countries, female; n = 59, 24.5% male). In Uganda, all technical corresponding to the actual numbers of health workers health workers including health facility management that employed in the identified three districts in each country. were present at the health facility and/or hospital on day In Ghana, the sample size was determined based on a of data collection, were invited to participate in the study. published sample size table (Israel, 2009), with an esti- Call-backs were made to facilities with high numbers of mated number of clinical health staff (i.e. professional staff, but poor attendance on the day of data collection. A groups) within the three districts estimated at 600, a pre- total of 466 responses (n = 326, 70% female; n = 140, 30% cision level of ± 5%, confidence level of 95% and degree male) were collected across the three districts in Uganda. of variability of 0.5, and a potential non-response rate of Vallières et al. Hum Resour Health (2021) 19:73 Page 4 of 11 In Malawi, health workers eligible from the district ranging from Strongly Disagree (= 1) to Strongly Agree hospital and government facilities were listed for each (= 5). The PSS was chosen as it has been widely used district and then sampled proportion to size of each facil- across LMICs [13, 36] and was found to have good inter- ity’s health workforce. This resulted in a total number of nal reliability in the current sample (a = 0.87). 67 health workers eligible for the survey drawn from 29 Job satisfaction was measured using Warr et  al.’s facilities across the three districts. To allow for potential (1979) 10-item Job Satisfaction Scale. The job satisfac- non-response, an additional 20% was added, for a total of tion scale was chosen as it has been widely used within 80 health workers. A total of 53 health workers (n = 22, medical practitioner research, and has been validated for 41.5% female; n = 31, 58.5% male) across 29 health use among clinicians [37]. Items on the job satisfaction facilities, including district hospitals, were ultimately scale were scored on a five-point Likert-type scale rang- included. In total, 760 health workers, of which 30.3% ing from Very Dissatisfied (= 1) to Very Satisfied (= 5), (n = 230) were male and 69.7% (n = 530) were female, whereby participants answer in terms of ‘How satisfied completed the questionnaire across all three countries. or dissatisfied they were’ with a number of extrinsic and intrinsic job-related items. The job satisfaction scale was Data collection found to have good internal reliability in the current sam- In-country members of the PERFORM2Scale project dis- ple (α = 0.84). tributed a self-administered, paper-based health worker District managerial support was assessed using a questionnaire to health workers. Written informed con- newly developed set of eight items, whereby participants sent was obtained from all study participants and all sur- answer in terms of ‘how much they agree’ with a num- veys were completed in English. ber of related items. Answers are scored on a five-point Likert-type scale ranging from Strongly Agree (= 1) to The health worker questionnaire Strongly Disagree (= 5). District managerial support was The questionnaire included 50 closed-ended items ask- found to have good internal reliability in the current sam- ing about the health workers’ socio-demographic char- ple (α = 0.77). acteristics, including sex (coded 0 = males, 1 = females), All variables were chosen based on the review of extant country (coded 0 = Uganda, 1 = Ghana, 2 = Malawi), type literature, and guided by the framework put forward by of health cadre (i.e. professional title), health centre level Vincent et  al. [7], ensuring at least one variable at the (where applicable), as well as their perception of safety individual, unit and organisational levels. climate, teamwork climate, supportive supervision, job satisfaction, and district managerial support. The ques- Data analysis tionnaire was piloted in all countries, with each coun- The relationships between organisational, team, indi- try’s version undergoing slight language modifications vidual factors and safety climate, as outlined the study’s to better suit the individual context, based on feedback theoretical framework (see Fig.  1), were tested using received. Those health workers who piloted the question- structural equation modelling (SEM) procedures. SEM naire were not part of the study population. was chosen for its ability to confirm the psychometric Safety Climate and Teamwork Climate were measured properties of the measurements employed as well as the using the respective subscales from Sexton et al.’s (2006) relationships between the latent variables [38], while validated Safety Attitudes Questionnaire—Short Form correcting for measurement error as well as testing the (SAQ). All items were rated using a five-point Likert strength of the model in explaining the observed pattern scale, anchored by Strongly Disagree (= 1) and Strongly of data [39]. Agree (= 5). The six-item Teamwork Climate sub-scale First, confirmatory factor analysis (CFA) was used was designed to capture the perceived quality of col- to assess the factor structure of the individual scales. laboration between health workers, whereas the six-item Optimal fit indicators were sought and therefore items Safety Climate sub-scale was designed to capture percep- demonstrating poor loadings (< 0.3) were removed. tions of a strong and proactive organisational commit- Measurement model goodness of fit was assessed using ment to safety. The SAQ has previously been found to a number of widely recognised fit indices [40, 41] includ- have good psychometric properties [32–34], with both ing: a non-significant Chi-square (χ2), Comparative Fit the teamwork climate and safety climate subscales found Index (CFI:42) and Tucker–Lewis Index (TLI: 43) val- to have acceptable internal reliability in the current sam- ues above 0.95 reflect excellent fit, while values above ple (α = 0.77, α =, α = 0.72, respectively). 0.90 reflect acceptable fit; Root-Mean-Square Error of Perceived Supervision was measured using the vali- Approximation with 90% confidence intervals (RMSEA dated Perceived Supervision Scale (PSS) [35], a six- 90% CI 44), and Standardised Root-Mean-Square Resid- item scale scored on a five-point Likert-type scale again ual (SRMR: 45) values of 0.06 or less reflect excellent fit Vallières et al. Hum Resour Health (2021) 19:73 Page 5 of 11 Fig. 1 Study’s analytical framework while values less than 0.08 reflect acceptable fit. For the demonstrated acceptable model fit during the measure- models based on Robust Maximum Likelihood estima- ment modelling phase (see Table 3). tor (MLR) estimation, the Bayesian Information Crite- rion (BIC: 46) was used to evaluate and compare models, Structural phase with the smallest value indicating the best fitting model. Results for all countries combined are presented in Second, a structural analysis, was used to determine the Table 4. effects of organisational, team and individual-level fac- In the first step, individual factors explained 26% of the tors on safety climate [47]. Data analyses were carried variance for safety climate. More specifically, job satis- out using STATA (Version 15) and SPSS (Version 26). In faction (β = 0.48, p < 0.001) significantly predicted safety the structural phase, individual factors (job satisfaction, climate and countries differed significantly in their per- country, sex), team (supportive supervision, teamwork), ception of safety climate. and organisational (perceived district managerial sup- Step two of the model, which included unit-level fac- port) were sequentially added into the model to deter- tors, revealed an additional 28% of the variance in safety mine their added contribution in explaining variations in climate. In this step teamwork (β = 0.59, p < 0.001) and safety climate. supportive supervision (β = 0.37, p < 0.001) emerged as the strongest predictors of safety climate. Country and job satisfaction also remained statistically signifi- Results cant (β = 0.19, p < 0.001). The final step, which included Measurement phase organisational factors, revealed an additional 1% of the Table  2 presents the descriptive statistics for each variance in safety climate. Overall, at step three the total variable, across the various countries. All five scales variance explained by the model was 55%. In this model, Vallières et al. Hum Resour Health (2021) 19:73 Page 6 of 11 Table 2 Variable descriptives by country teamwork (β = 0.56, p < 0.001) and supportive supervi- Variable Country N = 760 Mean SD sion (β = 0.34, p < 0.001) remained highly significant con- tributors to safety climate variance. In addition, district Teamwork climate Ghana 239 3.82 0.72 managerial support also emerged as significant (β = 0.29, Malawi 53 3.65 0.84 p < 0.001). Given the noted effect for country, a series Uganda 465 3.75 0.66 of post hoc analyses were carried out for each country, All 757 3.77 0.69 including type of health worker cadre and health centre District managerial support Ghana 235 3.65 0.61 level as additional individual-level and organisational Malawi 51 3.31 0.85 factors in the model, respectively. While a structural Uganda 461 3.59 0.67 equation model was tested for Uganda and Ghana, a All 747 3.59 0.67 hierarchical regression model was applied to the Malawi Supportive supervision Ghana 239 3.89 0.78 data, given the relatively low sample size. The health Malawi 51 3.46 1.11 centre-level variable was also excluded from the Malawi Uganda 464 3.88 0.74 analysis given that there was only one hospital, as one of All 754 3.86 0.79 the two categories (hospital vs. health centre). Results by Job Satisfaction Ghana 235 3.54 0.55 country are presented in Table 5. Malawi 53 3.34 0.83 Uganda 465 3.46 0.67 Ghana All 753 3.48 0.65 In the first step, individual factors explained 46% of the Safety climate Ghana 239 3.70 0.58 variance for safety climate. Job satisfaction (β = 0.65, Malawi 53 3.45 0.69 p < 0.001) and being clinical staff (compared to admin- Uganda 462 3.70 0.59 istrative staff; β = 0.23, p < 0.05) significantly predicted All 754 3.68 0.60 higher levels of safety climate. Step two of the model, which included unit-level factors, explained a reduced 43% of the variance in safety climate. In this step, Table 3 Model fit statistics for each of the study scale Scales No. items χ2 df p CFI TLI RMSEA (90% CI) SRMR BIC Teamwork climate (Sexton et al. 2006) 6 14.1 8 0.080 0.99 0.98 0.032 (0.001–0.058) 0.020 12,139.5 District managerial support 8 415.4 113 0.000 0.93 0.92 0.059 (0.050–0.083) 0.043 31,295.8 Perceived Supervision (Vallières et al. 2018) 5 2.9 4 0.571 1.0 1.0 0.00 (0.000–0.049) 0.007 8057.6 Job satisfaction Scale (Warr et al. 1979) 10 128.3 31 0.000 0.95 0.93 0.064 (0.053–0.076) 0.036 19,550.3 Safety Climate (Sexton et al. 2006) 5 5.6 4 0.230 0.99 0.99 0.023 (0.000–0.063) 0.013 9856.3 χ2 Chi-square goodness of fit statistic, df degrees of freedom, p statistical significance, CFI Comparative Fit Index, TLI Tucker–Lewis Index, RMSEA (90% CI) root-mean- square error of approximation with 90% confidence intervals, SRMR standardised square root mean residual, BIC Bayesian Information Criterion Table 4 Structural model results for all countries combined Covariate Model 1: individual‑level factors Model 2: individual‑level + unit‑ Model 3: individual‑ level factors level + unit + organisational‑ level factors β p‑value β p‑value β p‑value Sex (ref ) − 0.011 0.762 − 0.01 0.786 − 0.01 0.889 Country Uganda (ref ) 1 1 1 Ghana − 0.09 0.014* − 0.08 0.044 − 0.09 0.033 Malawi − 0.17 < 0.001** − 0.13 0.002** − 0.12 0.005** Job satisfaction 0.48 < 0.001** 0.19 < 0.001** 0.09 0.131 Teamwork climate 0.59 < 0.001** 0.56 < 0.001** Supportive supervision 0.37 < 0.001** 0.34 < 0.001** District managerial support 0.29 < 0.001** R2 0.26 0.54 0.55 *p < 0.05, **p < 0.01 denote statistically significant association (bold). β denotes the standardised coefficient V allières et al. Hum Resour Health (2021) 19:73 Page 7 of 11 Table 5 Results by country Covariate Model 1: individual‑level factors Model 2: individual‑ Model 3: individual‑ level + unit‑level factors level + unit + organisational‑ level factors β p‑value β p‑value β p‑value Ghana Male 0.069 0.314 0.012 0.878 0.005 0.947 Job satisfaction 0.65 < 0.001** 0.381 < 0.001** 0.304 0.011* Health worker qualification Administration (Ref ) 1 1 1 Clinical 0.233 0.011* 0.203 0.050 0.115 0.322 Public health staff 0.075 0.439 0.006 0.951 − 0.119 0.336 Team work 0.453 < 0.001** 0.459 < 0.001** Supportive supervision 0.201 0.049* 0.210 0.043* District managerial support 0.130 0.243 Level of health facility Community based 1 District hospital − 0.168 0.168 Health centre − 0.024 0.828 R2 0.46 0.43 0.43 Uganda Male 0.034 0.524 0.023 0.629 0.027 0.575 Health worker qualification Non-clinical (Ref ) 1 1 1 Clinical − 0.052 0.329 0.006 0.896 0.191 0.693 Job satisfaction 0.343 < 0.001** 0.149 0.005* 0.073 0.208 Team work 0.607 < 0.001** 0.563 < 0.001** Supportive supervision 0.468 < 0.001** 0.432 < 0.001** District managerial support 0.348 < 0.001** Health facility level HC II − 0.002 0.968 HC III 0.095 0.143 HC IV 0.028 0.650 Hospital 1 R2 0.12 0.61 0.64 Malawi Male 0.25 0.847 0.053 0.606 0.041 0.699 Health worker qualification Non-clinical (Ref ) Clinical 0.216 0.109 0.198 0.074 0.205 0.068 Job satisfaction 0.689 < 0.001** 0.260 0.042* 0.223 0.107 Teamwork 0.423 0.002* 0.394 0.005* Supportive supervision 0.292 0.012* 0.272 0.023* District managerial support 0.087 0.491 R2 0.46 0.63 0.63 *p < 0.05, **p < 0.01 denote statistically significant associations (bold). β denotes the standardised coefficient teamwork (β = 0.45, p < 0.001), job satisfaction (β = 0.38, did not explain any additional variance in safety climate. p < 0.001), and supportive supervision (β = 0.20, p < 0.05) Overall, at step three the total variance explained by the emerged as the strongest predictors of safety climate. model was 43%. In this model, job satisfaction (β = 0.30, The final step, which included organisational factors, p < 0.05), teamwork (β = 0.46, p < 0.001) and supportive Vallières et al. Hum Resour Health (2021) 19:73 Page 8 of 11 supervision (β = 0.21, p < 0.05) remained significant con- current samples. This finding is consistent with previ- tributors to safety climate variance. ous studies that emphasise the importance of support- ive supervision [14–17] and teamwork [20] as important Uganda correlates of service quality and patient safety across a In the first step, individual factors explained 12% of the range of primary health care services [19, 48]. Zaheer variance for safety climate, with only job satisfaction et  al. [49], for example, also found that teamwork and (β = 0.34, p < 0.001) found to predict safety climate. Step perceptions of leadership (at supervisory level) were two of the model, explained an increased 61% of the positively associated with perceptions of patient safety variance in safety climate. In this step, job satisfaction among Canadian nurses and allied health professionals. (β = 0.15, p = 0.005), teamwork (β = 0.61, p < 0.001), and Likewise, Kristensen et  al. found positive associations supportive supervision (β = 0.47, p < 0.001) emerging as between the implementation of quality management strong predictors of safety climate. The final step, which systems and both teamwork and safety climate among included organisational factors, explained an additional over 8500 clinical leaders and frontline clinicians sam- 3% of variance in safety climate. Overall, at step three pled across seven European countries [10]. A posi- the total variance explained by the model was 64%. In tive association between safety climate and teamwork this model, teamwork (β = 0.56, p < 0.001) and supportive was also reported in a sample of Jordanian nurses [50]. supervision (β = 0.43, p < 0.001) remained highly signifi- More recently, teamwork and organisational learning cant contributors to safety climate variance. In addition, was also highlighted in a facility based study in Ethiopia, district managerial support (β = 0.35, p < 0.001) was also where patient safety culture was significantly associated found to predict safety climate. with reporting of adverse events including an exchange of feedback about errors [51]. Accordingly, one possi- Malawi ble mechanism through which teamwork may facilitate Each step of the hierarchical regression model was sig- safety climate is through the adoption of practices such nificant (p < 0.001). In the first step, individual factors as quality improvement. significantly contributed to the model F [3, 45] = 12.51, The important role of supportive supervision, including p < 0.001. Individual factors explained 46% of the variance a joint problem-solving focus, the sense of joint responsi- for safety climate, with only job satisfaction (β = 0.69, bilities and teamwork, cross-learning and skill sharing, as p < 0.001) associated with safety climate. Step two of the well as the facilitating and coaching role of the supervisor, model, which included unit-level factors, explained up has been reported among health workers in other Afri- to 63% of the variance in safety climate [F (5, 43) = 17.52, can contexts [13]. As part of a supportive environment p < 0.001]. In this step, job satisfaction (β = 0.26, that fosters strong teamwork and supportive approaches, p = 0.042), teamwork (β = 0.42, p = 0.002), and supportive supervision is likely more conducive to health workers supervision (β = 0.29, p = 0.012) emerged as strong pre- learning from their mistakes, allowing for course cor- dictors of safety climate. The final step, which included rection, and reducing repeated errors. In contrast, more organisational factors did not explain any additional vari- punitive systems—or supervisory approaches that focus ance in safety climate. Overall, at step three the total vari- on fault-finding, inspection, and control [52]—increases ance explained by the model was 63%, F (5, 43) = 14.5, the risk that blame might be apportioned, thus incenti- p < 0.001). In this model, teamwork (β = 0.39, p = 0.005) vising health workers to hide, cover-up or not admit to and supportive supervision (β = 0.27, p = 0.023) remained errors or mistakes [53, 54]. significant contributors to safety climate variance. Dis- Individual country analyses suggest variations in trict managerial support was not found to predict safety whether or not perceived district-level support was climate. associated with perceived safety climate across con- texts. Where no association was found, it is possible that Discussion support at facility level compensated for the absence of The current study sought to examine the relationship more senior (i.e. district level) managerial support [49]. between key organisational, team and individual-level In Malawi, for example, supportive supervision is some- factors and safety climate, as central to ensuring quality times associated with development partners, rather health care and achieving UHC, across 138 selected pri- than DHMTs, which may explain why managerial sup- mary health care centres in Ghana, Malawi and Uganda. port is not necessarily associated to the district. Moreo- Overall, results across all three countries suggest that ver, in Ghana for example, while DHMTs offer technical unit-level factors, and more specifically, teamwork cli- and administrative support to primary healthcare units, mate and supportive supervision, emerge as factors that the DHMT maintain narrow decision-space for human best account for the variance in safety climate in the resource and fiscal decentralisation [55]. Increased V allières et al. Hum Resour Health (2021) 19:73 Page 9 of 11 decision-making power of sub-districts or unit heads Finally, the different sampling methods used across the regarding task shifting and task sharing could enhance three different study locations pose a challenge to com- teamwork, thereby improving safety climate for PHC ser- paring results between countries. vice delivery in Ghana [56]. Additionally, positive asso- ciations may be explained by closer interactions between district managers and primary care facilities, as condu- Conclusion cive to shaping a good working environment and condi- Together, our findings resonate with previous studies tions within primary health care facilities. Specifically, conducted by Yu and Liang [6] and Vincent et al.’s frame- health workers receiving supportive district managerial work [7], both of which highlight the importance of support may have more avenues to voice their grievances teamwork climate, supervision and perceptions of man- and challenges they face, and may receive more support agement/managerial support as important contributors in terms of supervision, resources and equipment, all of to safety climate. In addition, our findings highlight the which are necessary for greater workplace safety climate importance of unit-level factors (teamwork and support- [48, 57, 58]. Indeed, the responsibility for inspiring team- ive supervision) as particularly important contributors work, motivation, providing supportive supervision, and to perceptions of safety climate among primary health fostering positive staff attitudes is still widely seen as fall- workers in LMICs. Initiatives aiming to improve per- ing under the remit of the DHMTs [59]. For example, in ceived safety climate within primary health care centres, Malawi, the integrated supportive supervision system including planned initiatives within PERFORM2Scale, (ISS) is widely used by DHMTs across the country, as may want to consider paying particular attention to part of their Service Delivery Integration-Systems (SSDI- teamwork and improving supportive supervision prac- Systems). In Ghana, the observation that job satisfaction tices as key correlates of safety climate. remained positively associated with safety climate, while also accounting for the unit-level factors of teamwork and supportive supervision, is consistent with previous AbbreviationsBIC: Bayesian Information Criterion; CFI: Comparative Fit Index; DHMT: District studies, demonstrating that perceptions of safety climate health management team; ISS: Integrated supportive supervision; LMIC: can also influence job satisfaction, whereby employees Lower-middle income countries; PSS: Perceived supervision scale; RMSEA: who report high levels of perceived safety climate also Root-mean-square error of approximation; SAQ: Safety attitudes questionnaire; SEM: Structural equation modelling; SRMR: Standardised root-mean-square report high levels of job satisfaction [60]. residual; SSDI-Systems: Service delivery integration-systems; TLI: Tucker–Lewis The current study is not without limitations. Firstly, the Index; UHC: Universal Health Coverage. cross-sectional nature of our study design does not allow Acknowledgements for inferences of causality. While teamwork and sup- The authors wish to thank the study participants for their time and willingness portive supervision may contribute towards better safety to take part in the study. climate, it is also likely that within health facilities with Authors’ contributions positive or favourable safety climate, health workers are FV, HP, and PM conceptualised the study. All authors contributed to the study more likely to work as a team and have supervision mech- design. SSA, SA, TOB, LS, and JG conducted the data collection. FV and PM anisms in place encouraging them to perform their duties conducted the data analysis. FV and HP led the manuscript preparation, with signification contributions from PM, JG, TM, AN and FS. All authors contributed to a high standard. More likely however, the relationships to manuscript revisions. All authors read and approved the final manuscript. between teamwork, supportive supervision, and climate safety are likely multidirectional and mutually reinforc- FundingThis paper is a product of PERFORM2Scale (2017–2021), a H2020 programme ing, rather than unidirectional in nature. Second, and as to strengthen management at district level in Ghana, Malawi and Uganda, in safety climate was measured using a self-report meas- support of the achievement of Universal Health Coverage. The programme ure, as an indicator of perceived safety climate, we can- is funded by the European Union’s Horizon 2020 research and innovation programme (Reference Number: 733360). not reliably ascertain whether higher levels of unit-level factors are associated with more objective accounts of Availability of data and materials safety climate. Third, other known correlates of climate Data are available, upon reasonable request, from the first (FV) or second author (PM). safety, such as work engagement, safety behaviour, health worker knowledge and motivation [24], availability of Declarations resources and equipment, and interpersonal interactions, the latter of which showed the strongest association with Ethics approval and consent to participate safety climate in a meta-analysis also conceptually build- Ethical clearance was obtained from the Research Ethics Committee of the Liv-erpool School of Traditional Medicine (LSTM; ID No.: 17-046), the Ghana Health ing on Zohar’s model [61], were not included as part of Service Ethical Review Committee (ID No.: GHS-ERC: 009/12/17), the Commit- the original study design. It is possible that these other tee on Research in Social Sciences and Humanities in Malawi (NO.P.12/17/232), factors may act as stronger correlates of safety climate. and the Uganda National Council for Science and Technology (SS 4492). Written consent was obtained from all study participants. Vallières et al. Hum Resour Health (2021) 19:73 Page 10 of 11 Consent for publication 17. Hyrkäs K, Appelqvist-Schmidlechner K, Haataja R. Efficacy of clinical Not applicable. supervision: influence on job satisfaction, burnout and quality of care. J Adv Nurs. 2006;55(4):521–35. Competing interests 18. Pousette A, Larsman P, Eklöf M, Törner M. The relationship between The authors declare that they have no competing interests. patient safety climate and occupational safety climate in healthcare—a multi-level investigation. J Safety Res. 2017;61:187–98. Author details 19. Smits M, Wagner C, Spreeuwenberg P, van der Wal G, Groenewegen 1 Trinity Centre for Global Health, Trinity College Dublin, 7-9 Leinster Street pp. . Measuring patient safety culture: an assessment of the clustering South, Dublin 2, Ireland. 2 School of Public Health, Makerere University, Kam- of responses at unit level and hospital level. Quality Safety Health Care. pala, Uganda. 3 School of Public Health, College of Health Sciences, University 2009;18(4):292. of Ghana, Legon, P. O. Box LG13, Accra, Ghana. 4 Swiss Tropical and Public 20. Lee W-C, Wung H-Y, Liao H-H, Lo C-M, Chang F-L, Wang P-C, et al. Hospital Health Institute, Socinstrasse 57, 4002 Basel, Switzerland. 5 University of Basel, safety culture in Taiwan: a nationwide survey using Chinese version safety 4003 Basel, Switzerland. 6 Liverpool School of Tropical Medicine, Liverpool, attitude questionnaire. BMC Health Serv Res. 2010;10(1):234. United Kingdom. 7 Research for Equity and Community Health Trust (REACH 21. Singer SJ, Gaba DM, Falwell A, Lin S, Hayes J, Baker L. Patient safety Trust), Lilongwe, Malawi. climate in 92 US hospitals: differences by work area and discipline. Med Care. 2009;47(1):23–31. Received: 16 March 2021 Accepted: 28 May 2021 22. Travaglia JF, Westbrook MT, Braithwaite J. Implementation of a patient safety incident management system as viewed by doctors, nurses and allied health professionals. Health (London). 2009;13(3):277–96. 23. Gallego B, Westbrook MT, Dunn AG, Braithwaite J. Investigating patient safety culture across a health system: multilevel modelling of differences References associated with service types and staff demographics. Int J Qual Health 1. Kruk ME, Gage AD, Arsenault C, Jordan K, Leslie HH, Roder-DeWan S, et al. Care. 2012;24(4):311–20. High-quality health systems in the Sustainable Development Goals era: 24. Neal A, Griffin MA. Safety climate and safety at work. In: Barling J, Frone time for a revolution. Lancet Glob Health. 2018;6(11):e1196–252. MR, editors. The psychology of workplace safety: American Psychological 2. Sorra JS, Nieva VF. Hospital survey on patient safety culture. 2004. Association; 2004. p. 15–34. 3. Sexton JB, Helmreich RL, Neilands TB, Rowan K, Vella K, Boyden J, et al. The 25. Zarei E, Khakzad N, Reniers G, Akbari R. On the relationship between Safety Attitudes Questionnaire: psychometric properties, benchmarking safety climate and occupational burnout in healthcare organizations. Saf data, and emerging research. BMC Health Serv Res. 2006;6:44. Sci. 2016;89:1–10. 4. Zohar D. The influence of leadership and climate on occupational health 26. Lawati MHAL, Dennis S, Short SD, Abdulhadi NN. Patient safety and and safety. In: Hofmann DA, Tetrick LE, editors. Health and safety in safety culture in primary health care: a systematic review. BMC Fam Pract. organizations: a multilevel perspective. San Francisco: Jossey-Bass; 2003. 2018;19(1):104. p. 201–30. 27. Khammar A, Poursadeghiyan M, Marioryad H, Nabi Amjad R, Alimoham- 5. Zohar D. Thirty years of safety climate research: reflections and future madi M, Khandan M. Patient safety climate and its affecting factors directions. Accid Anal Prev. 2010;42(5):1517–22. among rehabilitation health care staff of hospitals and rehabilitation 6. Yu S, Liang HY, editors. Nurses perceived emotional labour, safety climate, centers in Iran-Tehran. Iran Rehabil J. 2019;17(1):39–48. and their relationships with health status. 5th International Conference 28. Cunha MRPD, Guirardello EB. Patient Safety Climate in Healthcare and Exhibition on Occupational Health & Safety; 2016. Organizations: translation and adaptation to the Brazilian culture. Revista 7. Vincent C, Taylor-Adams S, Stanhope N. Framework for analysing risk and Gaúcha de Enfermagem. 2018;39. safety in clinical medicine. BMJ. 1998;316(7138):1154–7. 29. Griffin MA, Neal A. Perceptions of safety at work: a framework for linking 8. Alsalem G, Bowie P, Morrison J. Assessing safety climate in acute hospital safety climate to safety performance, knowledge, and motivation. J settings: a systematic review of the adequacy of the psychometric prop- Occup Health Psychol. 2000;5(3):347–58. erties of survey measurement tools. BMC Health Serv Res. 2018;18(1):353. 30. Khan K, Ramachandran S. Conceptual framework for performance 9. Teuma Custo P, Teuma Custo R, Buttigieg S. The relationship between assessment: competency, competence and performance in the context safety climate and performance in intensive care units: the mediating of assessments in healthcare–deciphering the terminology. Med Teach. role of managerial safety practices and priority of safety. Front Public 2012;34(11):920–8. Health. 2019;7:302. 31. Leonard KL, Masatu MC. Professionalism and the know-do gap: exploring 10. Kristensen S, Hammer A, Bartels P, Suñol R, Groene O, Thompson CA, et al. intrinsic motivation among health workers in Tanzania. Health Econ. Quality management and perceptions of teamwork and safety climate in 2010;19(12):1461–77. European hospitals. Int J Qual Health Care. 2015;27(6):499–506. 32. Nguyen G, Gambashidze N, Ilyas SA, Pascu D. Validation of the safety 11. Fetene N, Canavan ME, Megentta A, Linnander E, Tan AX, Nadew K, et al. attitudes questionnaire (short form 2006) in Italian in hospitals in the District-level health management and health system performance. PLoS northeast of Italy. BMC Health Serv Res. 2015;15(1):284. ONE. 2019;14(2):e0210624. 33. Bondevik GT, Hofoss D, Husebø BS, Deilkås ECT. The safety attitudes 12. Marquez L, Kean L. Making supervision supportive and sustainable: new questionnaire—ambulatory version: psychometric properties of the Nor- approaches to old problems. Washington: USAID: Maximizing Access and wegian version for nursing homes. BMC Health Serv Res. 2019;19(1):423. Quality (MAQ) Initiative; 2002. 34. Carvalho REFLd, Cassiani SHDB. Cross-cultural adaptation of the Safety 13. Kok MC, Vallières F, Tulloch O, Kumar MB, Kea AZ, Karuga R, et al. Does Attitudes Questionnaire—Short Form 2006 for Brazil. Revista Latino- supportive supervision enhance community health worker motivation? Americana de Enfermagem. 2012;20:575–82. A mixed-methods study in four African countries. Health Policy Planning. 35. Vallières F, Hyland P, McAuliffe E, Mahmud I, Tulloch O, Walker P, et al. A 2018:czy082-czy. new tool to measure approaches to supervision from the perspective of 14. Bosch-Capblanch X, Marceau C. Training, supervision and quality of care community health workers: a prospective, longitudinal, validation study in selected integrated community case management (iCCM) pro- in seven countries. BMC Health Serv Res. 2018;18(1):806. grammes: a scoping review of programmatic evidence. J Global Health. 36. Aldamman K, Tamrakar T, Dinesen C, Wiedemann N, Murphy J, Hansen 2014;4(2):020403. M, et al. Caring for the mental health of humanitarian volunteers in 15. Suh S, Moreira P, Ly M. Improving quality of reproductive health care in traumatic contexts: the importance of organisational support. Eur J Senegal through formative supervision: results from four districts. Hum Psychotraumatol. 2019;10(1):1694811. Resour Health. 2007;5(1):26. 37. Hills D, Joyce C, Humphreys J. Validation of a job satisfaction scale in the 16. Djibuti M, Gotsadze G, Zoidze A, Mataradze G, Esmail LC, Kohler JC. The Australian clinical medical workforce. Eval Health Prof. 2012;35(1):47–76. role of supportive supervision on immunization program outcome— 38. Byrne BM. Structural equation modeling with Mplus: basic concepts, a randomized field trial from Georgia. BMC Int Health Hum Rights. applications, and programming. Hove, East Sussex: Routledge Taylor & 2009;9(Suppl 1):S11. Francis Group; 2012. V allières et al. Hum Resour Health (2021) 19:73 Page 11 of 11 39. Preacher KJ, Hayes A. Asymptotic and resampling strategies for assessing Southeast Ethiopia: an institutional based cross-sectional study. Drug and comparing indirect effects in multiple mediator models. Behav Res Healthc Patient Saf. 2020;12:1–14. Methods. 2008;40(3):879–91. 52. Clements CJ, Streefland PH, Malau C. Supervision in primary health 40. Hu L-t, Bentler PM. Fit indices in covariance structure modeling: sensitiv- care—can it be carried out effectively in developing countries? Curr Drug ity to underparameterized model misspecification. Psychol Methods. Saf. 2007;2(1):19–23. 1998;3(4):424. 53. Sharma R, Webster P, Bhattacharyya S. Factors affecting the performance 41. Hu Lt, Bentler PM. Cutoff criteria for fit indexes in covariance structure of community health workers in India: a multi-stakeholder perspective. analysis: conventional criteria versus new alternatives. Struct Equ Mod- Glob Health Action. 2014;7:25352. eling Multidiscip J. 1999;6(1):1–55. 54. Hancox K, Lynch L, Happell B, Biondo S. An evaluation of an edu- 42. Bentler PM. Comparative fit indexes in structural models. Psychol Bull. cational program for clinical supervision. Int J Ment Health Nurs. 1990;107(2):238. 2004;13(3):198–203. 43. Tucker LR, Lewis C. A reliability coefficient for maximum likelihood factor 55. Kwamie A, van Dijk H, Ansah EK, Agyepong IA. The path depend- analysis. Psychometrika. 1973;38(1):1–10. ence of district manager decision-space in Ghana. Health Policy Plan. 44. Steiger JH, editor Statistically based tests for the number of common 2016;31(3):356–66. factors1980. 56. Avortri GS, Nabukalu JB, Nabyonga-Orem J. Supportive supervision to 45. Jöreskog KG, Sörbom D. LISREL 8: User’s reference guide: Scientific Soft- improve service delivery in low-income countries: is there a concep- ware International; 1996. tual problem or a strategy problem? BMJ Glob Health. 2019;4(Suppl 46. Schwarz G. Estimating the dimension of a model. Ann Stat. 9):e001151. 1978;6(2):461–4. 57. Flin R, Burns C, Mearns K, Yule S, Robertson EM. Measuring safety climate 47. Morrison TG, Morrison MA, McCutcheon JM. Best practice recommenda- in health care. Qual Saf Health Care. 2006;15(2):109–15. tions for using structural equation modelling in psychological research. 58. McCaughey D, Halbesleben JR, Savage GT, Simons T, McGhan GE. Safety Psychology. 2017;8(09):1326. leadership: extending workplace safety climate best practices across 48. Rosen MA, DiazGranados D, Dietz AS, Benishek LE, Thompson D, Pro- health care workforces. Adv Health Care Manag. 2013;14:189–217. novost PJ, et al. Teamwork in healthcare: Key discoveries enabling safer, 59. Gilson L, Lehmann U, Schneider H. Practicing governance towards equity high-quality care. Am Psychol. 2018;73(4):433–50. in health systems: LMIC perspectives and experience. Int J Equity Health. 49. Zaheer S, Ginsburg LR, Wong HJ, Thomson K, Bain L. Importance of safety 2017;16(1):171. climate, teamwork climate and demographics: understanding nurses, 60. Hofmann DA, Mark B. An investigation of the relationship between allied health professionals and clerical staff perceptions of patient safety. safety climate and medication errors as well as other nurse and patient BMJ Open Qual. 2018;7(4):e000433. outcomes. Pers Psychol. 2006;59(4):847–69. 50. AbuAlRub RF, Gharaibeh HF, Bashayreh AEI. The relationships between 61. He Y, Wang Y, Payne SC. How is safety climate formed? A meta- safety climate, teamwork, and intent to stay at work among Jordanian analysis of the antecedents of safety climate. Organ Psychol Rev. hospital nurses. Nurs Forum. 2012;47(1):65–75. 2019;9(2–3):124–56. 51. Kumbi M, Hussen A, Lette A, Nuriye S, Morka G. Patient safety culture and associated factors among health care providers in bale Zone Hospitals, Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. Ready to submit your research ? Choose BMC and benefit from: • fast, convenient online submission • thorough peer review by experienced rese archers in your field • rapid publication on acceptance • support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations • maximum visibility for your research: over 100M website views per year At BMC, research is always in progress. Learn more biomedcentral.com/submissions