Review: application of the Safety Attitudes Questionnaire (SAQ) in primary care - a systematic synthesis on validity, descriptive and comparative results, and variance across organisational units

A literature search was conducted, with assistance from a librarian, to find studies that used any version of SAQ in primary care. The electronic databases: PubMed, Embase, Cinahl PsycInfo and Web of Science was used. The search was carried out by first author with assistance from the Medical Librarian at Aalborg University Hospital. The search terms used were: “Safety attitude* questionnaire*” AND “Primary health care” OR "primary healthcare" OR "primary care" OR "primary health sector" OR "primary sector" OR "first line care" OR "primary medical care" OR "ambulatory setting*" OR out-patient* OR outpatient*. The search was limited to 2006-2023 due to the fact, that the first SAQ publication was published in 2006. The searches were conducted March 14^th, 2023 and updated November 7^th 2023. In addition, snowballing using citations, and references to other publications on SAQ in primary care was used to search for additional studies.

Studies were excluded if only abstract or poster was available, as the information in abstract and posters was deemed insufficient. Commentaries and nonempirical studies (e.g., study protocols, editorials) were excluded. Only English manuscripts were included. Studies referring to primary care were included.

A synthesis was carried out focusing on a qualitative analysis of the information obtained for each for the four themes. In an analytical framework, classification was used to charting the data by organising concepts into themes to systematically select relevant outcomes to compare properties of SAQ in primary care.

The literature search resulted in 53 publications after deletion of duplicates. The first study was published in 2007. Additional four studies were identified via snowballing. A total of 43 studies were included based on in- and exclusion-criteria. A least two authors screened each record independently to evaluate if the studies met the inclusion criteria. Due to the low number of retrieved studies, no studies were removed before screening and no automation tools were used in the process. Three of the 43 studies were not included in further analyses. The study selection flowchart is illustrated in Fig. 1.

An analytical framework classification charted the data by organising concepts into themes. Relevant outcomes were systematically selected to compare properties of SAQ in primary care (Fig. 2). All authors contributed to the analytical framework and agreed on the synthesis and the relevant outcomes. One German study validated a newly developed and unique safety climate questionnaire, based partly on SAQ-A, for use in German general practices [25]. In comparison with the SAQ-A, only 17 of 30 items and two of six dimensions (team climate and job satisfaction) remained. Although, the study started out with the SAQ-A, the different interpretation and content of factors showed that a brand new questionnaire was developed (FraSiK) [25]. Thus, this study was not included in further analysis. Results from a Slovenian study were presented both internationally and nationally [17, 26]. We decided only to include the international study, thus the national Slovenian study was not included in further analysis [26]. Another study investigated the practice environment of primary care nurses and used the job satisfaction subscale of SAQ-SF to assess job satisfaction [27]. As only one of six dimensions were used, the study was not included in further analysis.

The 40 remaining studies were divided into four analytic categories: 1) validation analysis, N=17, 2) descriptive analysis, N=25, 3) variance assessment, N=3 and 4) intervention evaluation, N=5. Nine studies were included in more than one category.

The analytical key issues and themes were used to synthesize the findings and present results. Figure 3 provides an overview of number of studies included in the four analytic themes.

Seventeen studies from thirteen different countries have validated different versions of SAQ including SAQ-A, SAQ-AV (n=11), SAQ-SF (n=4), a modified Chinese version (CSAQ, n=1) and a non-specified SAQ version (n=1) in a variety of settings and participants in primary care [10, 17, 18, 28‐41] (Table 1). Doctors and or nurses participated in all studies except for one carried out in pharmacies. Sample sizes varied from 211 to 7427 invitees and response rates varied between 8.7% and 99.4% (Table 1). SAQ-AV are derived from the SAQ-A versions and in some studies, SAQ-AV and SAQ-A were used interchangeably. Psychometric properties of the questionnaires in each of the studies are presented in Table 1. Reliability was mostly assessed using measures of internal consistency including Cronbach’s alpha, but acceptable threshold levels varied slightly across studies (0.60-0.70). Seven studies [10, 17, 32, 33, 35, 38, 41], had response rates above the anticipated 60% [15] with the number of completed questionnaires varying from 154 to 4090 These studies were carried out in a variety of settings. All seven studies reported Cronbach’s alphas above 0,6 and different factor structures (between 4 and 6 factors). Factor analyses, both exploratory and confirmatory, were used to validate the constructs of the SAQ questionnaires (Table 1), leading to identification and acceptance of different factor structures across studies, most often a six-factor structure was confirmed.

In summary different versions of SAQ have been validated for different settings. However, heterogeneity was high in how extensively the different SAQ adaptations were validated.

Twenty-five studies reported descriptive results of SAQ in primary care [10, 18, 29, 30, 33, 34, 37‐40, 42‐56] (Table 2). The studies came from fourteen different countries. Most studies were conducted in primary health care centres, out-of-hours clinics, nursing homes and general practice. Only one study from Brazil was conducted in home care [44]. The studies used different versions of SAQ including SAQ-SF, SAQ-AV, and a modified Chinese version (CSAQ). Sample sizes of invitees ranged from 140-1974 (Table 2). A variety of participants were included in the studies and response rates ranged from 8.7%-97.8%. The Job satisfaction factor obtained highest scores in some studies [39, 44, 45, 51] and lowest scores in one study from Slovenia [50]. Perception of Management received lowest scores in some studies [37, 39, 44]. Results from the twenty-five studies are presented in Table 2.

Overall differences in SAQ factor scores were related to demographic characteristics as different scores were found when different settings (regions, clinics, practices, and teams), genders, ages, degrees of education, professional groups, time of professional experiences and job types were compared (Table 2).

Three studies compared primary and tertiary health care facilities and found that SAQ-AV results were significantly different between primary and tertiary health care facilities [29, 37, 47]. Two studies found that providers from primary care scored higher than providers from tertiary care [29, 47]. In contrast, another study from Egypt found that tertiary health care workers had higher mean scores of Teamwork climate, Perception of management, Job satisfaction, Working conditions, and Stress recognition’ and the overall CSAQ score [37]. The response rate was lower in tertiary centre than from the primary level of care [29]. Another study compared primary health care centres and home care settings and found that home care professionals gave higher scores than primary care professionals for all domains, except Perception of stress [44].

Several studies reported that males gave higher SAQ-scores than females [11, 37, 38, 44]. However, some studies did not find these differences [39, 43, 51] and others reported higher scores from females in some factors [34]. For example AlMaani et al. reported that the score of Teamwork and Stress recognition was higher among females, whereas Perception of management was higher among males [30].

In general, older employees gave higher scores than younger [11, 37, 38, 45, 48, 51, 52]. However, some studies did not find this association [39, 43] and one study, in contrast found that all factors and the overall score were higher in providers less than 40 years compared to older providers [30]. Another study from Portugal found that in the stress recognition domain, as the age of the respondent increased, the obtained SAQ-SF median score decreased [34].

Some studies investigated the association of SAQ-scores to other variables. For example, one study investigated if SAQ is an indicator for health care providers readiness for reporting quality and found hat providers with the highest SAQ score had higher odds to report a higher readiness on the appropriateness, efficacy, management and personal valence Readiness for Organization Change subscales [46]. Another study investigated safety attitudes of pharmacists and found that higher scores (overall score as well as score for each domain except for stress recognition) correlated negatively with number of reported medication errors [47]. In contrast, Miller et al. studied the relationship between safety culture and voluntary adverse event reporting in a regional ambulatory care group and found that for every 1-percentage-point increase in overall climate score, there was a 1.9% increase in monthly safety reports [54]. Another study investigated the relationship between safety climate and implicit rationing of nursing care and found that higher safety climate was only related to lower rationing in the subscales activities of daily living and caring, rehabilitation and monitoring whereas better safety climate was related to higher rationing in social care [56].

In summary, differences in SAQ factor scores were related to a variety of factors, that should be considered in future studies.

Only three studies investigated variance in SAQ-scores across organizational units at different levels, one study from the Netherlands and two studies from Norway [18, 22, 23] (Table 3).

Different settings and participants were investigated, but all three studies used SAQ-AV. Sample size varied between 510 and 765 invitees, and they were nested into different numbers of units, wards, and teams. Response rates varied per organisation.

One study found that team level variance was higher than unit level variance [18]. Another study found that Intraclass correlation coefficients (ICC) for variance at nursing home level was zero or less than one % for all factor scores [22]. At ward level ICCs for the factors were 10.2% or higher for the factors Safety climate, Working conditions and Perceptions of management, 2.4% or lower for Teamwork climate, Job satisfaction, and zero for Stress recognition [22]. Another study found that staff perceptions varied considerably at the work site level: ICCs were 12.3% or higher for all factors except for Job satisfaction–the highest ICC value was for Perceptions of management: 15.5%. Although most of the score variance was at the individual level, there was considerable response clustering at work unit level, for the general practitioner practices and out-of-hours clinics [23].

In summary, variances in SAQ-scores across organizational units were found.

Only four studies used SAQ to assess changes in SAQ scored over time during interventions in before-after studies and one study used SAQ to examine association between burnout and other factors among health care workers during COVID-19 in primary care settings (Table 4). The five studies came from Singapore, UK, and USA [16, 57‐60] (Table 4). Different settings and participants were studied. Three studies used the SAQ, one study used SAQ-A and one study conducted interviews based on a framework adapted from the SAQ (Table 4). Sample size varied from 14 (the qualitative study) to 11286 invitees. Response rates varied from 14.5% to 96.2%.

The first interventional study using SAQ in primary care was performed in USA in 2013. The study was conducted to sequentially measure, evaluate, and respond to safety climate and practice safety concerns following electronic medical record implementation in medical group practice (affiliated with an academic medical centre) [16]. It was demonstrated that safety climate improved over the period after implementation of electronic medical record, with statistically significant improvement in all domains except for stress recognition [16].

In 2017 another American study assessed the impact of Comprehensive Unit-based Safety Program (CUSP) on safety climate and teamwork through a before-after comparison of results on the validated SAQ [59]. The study was conducted in general internal medicine, a suburban, academic practice. Twenty-five providers and staff completed the survey three months before CUSP implementation and six months following the kick-off of CUSP. Compared to before, following CUSP implementation, survey respondents were more likely to report knowledge of the proper channels for questions about patient safety, feel encouraged to report safety concerns and believe that the work setting made it easy to learn from the errors of others. However, these differences did not reach statistical significance [59].

Since 2020, three different studies have used SAQ in the evaluation of the impact of COVID-19 on health care workers [57, 58, 60]. One study from Singapore examined burnout and associated factors among health care workers in public hospitals and primary health care services involved in the care of COVID-19 cases [60]. They found that high SAQ scores were significantly associated with lower scores of Oldenburg Burnout Inventory [60]. Another study from Singapore compared health care workers mental well-being in 2021 against the previously published cohort in 2020 [57]. The study included 1475 respondents (response rate 14.5%). For each factor, % positive was a significantly lower in 2021 than in 2020 [57]. However, results for these two studies were not reported for primary and tertiary care separately.

A study from UK investigated GPs' experiences of how UK COVID-19 policies have affected the management and safety of complex elderly patients, who suffer from multimorbidity, at the primary care level [58]. The setting was general practice where fourteen interviews were conducted. The SAQ was used as one of two theoretical frameworks that were the base for drafting a primary interview guide to explore policies’ impact on management and safety. SAQ was not used in full version as the study only included interviews based on themes adapted from five SAQ factors (Work conditions, Safety environment, Perception of management, Teamwork environment and Stress recognition). The study did not make specific conclusions on culture but concluded that the COVID-19 pandemic affected all levels of the health system in the UK, particularly primary care. Based on the GPs’ perspective, changes to clinical practice have offered opportunities to maintain safe healthcare as well as possible drawbacks that should be of concern [58].

In summary, due to lack of available and comparable studies, no firm conclusions can be drawn on the effects of interventions on SAQ-scores.

In this systematic review of 40 studies investigating the application of the SAQ in primary care we synthesized validity, descriptive and comparative results, and variance across organisational units. Seventeen studies reported on validation of different versions of SAQ in a variety of primary care settings across thirteen different countries, different participants. Number of participants, response rates and validation methods including set threshold values for accepting the validation results, greatly varied across studies. Twenty-five descriptive studies demonstrated differences in SAQ dimensional scores found between settings (regions, clinics, practices, and teams), genders, ages, degrees of education, professional groups, time of professional experiences and job types. Three studies investigating variance in SAQ scores across different organizational levels found significant and substantial variance at work group unit level. Lastly, five before-after interventional studies used SAQ in the evaluation of an intervention or event such as introducing the electronic medical record, comprehensive Unit-based Safety Program, or the COVID pandemic. Study results indicate that SAQ can be used for detection of changes in patient safety culture over time or to point at associations between outcome measures.

Strengths of this review include the systematic literature search and systematic methods of study selection and data extraction. To our knowledge, this is the first review of its kind. It adds to knowledge base of using SAQ as a research and quality improvement instrument to assess patient safety culture, and it structures the knowledge into the four themes studied. Other reviews on patient safety instruments have been published, but they cover a variety of assessment instruments [61, 62]. Also, this study covers primary care. Until now most studies on patient safety climate have been conducted in hospital settings [61‐64]. Thus, as expected, we found a very limited number of studies on patient safety culture in primary care settings. Nevertheless, this study synthesises valuable information on patient safety culture, which can be used as a step stone for future studies and application in primary care settings.

The study has some weaknesses. Foremost, this systematic review focused only on articles written in English. Another limitation is that the general definition of primary care is very broad and comprises a variety of different types of healthcare organization including different groups of professional health care providers with highly different tasks, which they pertain to while reporting on perceived patient safety culture. We could have excluded specific primary care settings to create a more uniform group for presentation, but doing so would result in the exclusion of pertinent studies. Our objective was to offer a comprehensive overview of the application of SAQ in primary care practices varying significantly across countries [65], and it would be inappropriate to determine which specific primary care setting is of paramount importance. Patient safety research often involves diverse healthcare settings, patient populations, and methodologies. Including all studies helps account for this variability, enabling a more comprehensive assessment of how patient safety attitudes differ across contexts and patient groups. . Moreover, studies also varied in the methodology and sampling strategies used, which makes synthesis of results across a limited number of studies difficult - as is the case in each of our four themes. However, we accounted for this fact by applying a descriptive synthesis carefully reporting on single studies, as opposed to applying quantitative rating review methodology.

This study did not account for language and cultural disparities predominant in the specific countries in which the reported studies were conducted. Such disparities could possibly introduce bias of a complex and unstudied art, which we cannot know the implications of.

Climate is an emergent property, characterizing groups of individuals. Operationally it is assessed by aggregating individual perceptions to the required unit of analysis. And using the mean to represent the climate for that entity. However, this requires within-unit homogeneity or consensus of perceptions. Without sufficient homogeneity, an aggregate score is not a valid indicator of climate [66]. Thus, not only response rate but also intra-unit variation will affect the results and interpretation. Response rates varied between studies, and not all studies reported precise response rates. Because it has been suggested that response rates below 60%, represent opinions rather than culture and climate, we have assessed results from such studies with caution and not given them weight in the synthesis to minimize the introduction of bias of an unknown kind [15]. We found that some professional backgrounds show higher response rates than others. For example one can speculate if the response rate among medical doctors in general practices is higher than for out-of-hours medical doctor, due to a motivational factor of being an owner/leader and being interested in contributing to the evaluation of their work environment [23]. Additionally, we detected response rate markedly lower for physicians than for non-physicians [67] and the overall response rate was almost twice as high among nurses compared to medical doctors in one study [11]. Moreover, the response rate may vary across units (from 44% to 100%) [67] and settings.

A high variation in response rates across studies was found. It has been discussed that the large number of nursing home employees working part-time may have a higher degree of uncertainty about patient safety [41]. This could possibly affect the willingness to participate in studies on patient safety [41] or introduce information bias. In contrast it could also be speculated that employees working part-time may have more personal resources and are less prone to stress and therefore may be more willing to participate in such studies. It has also been discussed that response rate is highly dependent on the method used for distribution of questionnaires. For example, it was demonstrated that the response rate was much higher for questionnaires distributed in meetings (96%) than for those distributed through the mailing system (50%) [67]. Thus, in summary, big variations in response rates indicate that it may be difficult to achieve response rates above 60% in all settings. Yet, it is highly important to investigate all settings in primary care, and account for low response rates when assessing study validity and reporting results and conclusions.

Synthesised, we found that direct comparison between the included studies in each of our four themes is challenging and should be done with caution. This is also the reason why it was not possible to conduct a systematic review or present quantitative analysis.

Most studies confirmed validation of different versions of SAQ for use in primary care. However, it should be considered that of the 17 validity studies, seven studies had response rates above the anticipated 60% [15] with the number of completed questionnaires varying from 154 to 4090 and they were carried out in a variety of settings. In six of the studies nurses and or doctors participated, whereas only pharmacists took part in the 7^th study. None of the seven studies investigated other health care worker groups of lower-level education that are often represented in primary care, e.g. social and health care assistants, nursing assistant etc. Heterogeneity was high in how extensively the different SAQ adaptations had been validated, thus it was not possible to use a single rating scale for all studies. As an alternative, we found it relevant to show how well the tools had been validated related to reliability, internal consistency, and construct validity (presented in Table 1). All seven studies reported Cronbach’s alpha above 0.6, and different four to six factor structures were accepted following exploratory factor analysis for these seven studies. Five of the studies confirmed factors by confirmatory analysis.

Each clinical area possesses a unique social fabric of culture, potentially leading respondents who work within the same clinical area to respond more similarly than respondents who are members of different clinical areas [6]. Thus, the structural differences in design of the questionnaire versions may reflect variation in the organization of primary health care systems [65], or it may mean that item wordings trigger different connotations in different languages or with different clinical tasks/responsibilities. For these reasons a wise option might be to compare countries’ SAQ results at the item level rather than at the dimensional level [17]. Moreover, given differences in healthcare systems and culture between countries, factors loading differently is not necessarily a problem and may simply reflect the local context of healthcare and the associated variation [68].

It was discussed that the SAQ-AV should be adapted for support staff that have direct patient contact as 25% of the items did not apply to them and that the SAQ-AV needs to be tested in other outpatient settings [10]. One study evaluated the psychometric properties of the SAQ-SF in employees with a leadership role in community health centres [17]. However, leaders and managers tend to rate safety culture higher than frontline workers [69]. Thus, results from studies based exclusively on leaders’ attitudes should be interpreted with caution as they will not provide precise measures of safety climate in each setting. This will inhibit the actions needed in terms of quality improvement strategies.

Each study adapted the SAQ to the setting investigated. The risk of missing items may increase with increased length of the questionnaire which was also discussed in one study arguing that the length of the 62-item SAQ-AV may have dissuaded participants from completing and returning the questionnaire. It would be desired to have a shorter version of the measure for easy administration so long as the shortened version is valid and reliable” [29].

In some studies, stress recognition was not confirmed as a factor [18, 40]. Similar findings were found in the hospital setting where it has been argued, that it should not be included in the overall safety attitude construct, which the SAQ intends to reflect [70]. It is stated that stress recognition is a dissonant sub-scale of the safety climate construct and that the other subscales refer to the perspectives of respondents on their work areas or broader organisational units; stress recognition is about individual perspectives on abilities [70]. This was supported by e.g. Almaani et al. who also found that stress recognition had the lowest mean indicating that the acceptance of how work is affected by stressors is least recognized among all the sub-scales [30].

One study suggested that Communication and Psychological safety were perceived important safety climate factors since they emerged as individual factors without the study seeking to map them explicitly [31]. Thus, it was suggested that Communication and Psychological safety could be considered possible independent factors in future safety climate surveys [31].

The 25 descriptive studies demonstrated that differences in SAQ dimensional scores were found between settings (regions, clinics, practices, and teams), genders, ages, degrees of education, professional groups, time of professional experiences, job types and rationing of nursing care. However, direct comparison between studies is complex as the studies came from twelve different countries with different clinical cultures and different health care organization/settings. To that end, it has been discussed that the structure differences of the SAQ versions (used in different countries) may reflect cross-national variation in the nature and structure of primary care, or mean that item wordings trigger different connotations in the different languages [28].

As safety climate scores are likely dependent on educational level, they may, be less comparable among healthcare settings that differ in average educational level [18]. Additionally, clear differences in how healthcare professionals conceptualise ‘patient safety and quality have been demonstrated [71]. Understanding this variance may enable more effective targeting of interprofessional improvement strategies [71]. Reporting errors and safety awareness in hospital setting, gender and demographics, work experience, and staffing levels have also been identified as essential factors [61]. It could also be hypothesised that the likely dependency between safety climate scores and educational level is introduced in the curriculum of different professions’ education or in the political, legal, or structural ties and demands of patient safety at national level [72].

Significant variations in patient safety attitudes are related to age [45]. Higher age is directly related to longer life experience and for most people also working experience which may influence patient safety attitudes. Holden et al. discussed that age differences crossed professional lines and may explain why there can be differences among age groups and no substantial differences among professional groups [52]. Thus, Holden concluded that in terms of policy development, those in leadership positions who are concerned with enhancing team-work may be well served to develop strategies and interventions that target the younger professional staff [52]. This was supported by our study as several studies found that younger providers gave lower SAQ-scores.

Only one study was conducted in home care, although the home care setting is highly relevant regarding patient safety culture [44]. A great patient safety challenge in primary care is medication safety. The situation of drug-related problems in the home care setting has not been well- characterized [73]. A systematic review found that home care patients were predominantly elderly. Not surprisingly, polypharmacy was common in patients ≥ 65 years of age and it was found that patients received 5–14 medications per day [73]. Thus, the home care setting is highly relevant regarding patient safety culture and more studies are warranted.

Safety climates vary between work sites in hospitals, and predict where to find risk related to tasks, work environment, staff behaviour, and patient results [23]. This may provide opportunity to direct leadership efforts to where improvement is most needed. One purpose with SAQ is to elucidate variation between organizational units, to be able to use the information to direct improvement efforts to the units with highest need. Thus, it is important to study variance between organizational units in the settings to which the SAQ is adapted [74, 75]. In this study we use data that explicitly reveal variance according to organizational level in primary care to explore if safety culture measurements there provide an equal opportunity. In one of our reviewed studies ICC for variance at nursing home level was zero or less than one percent for all factor scores [22]. The lack of variance across nursing homes was supported by another study that found more variance between work units than between institutions [15]. This means that safety culture measurements also in primary care may provide opportunity to direct leadership efforts to work units where improvement is most needed. It is therefore critical to assess safety culture across all work units in primary care institutions, to uncover worksites that may be more promising candidates for patient safety improvement interventions than others.

We found only five studies using SAQ to assess the effects of interventions in before-after studies in primary care settings. Only one of the studies used a version of SAQ, that has been validated for use in primary care (SAQ-A) [16] and this was the only study finding that safety climate improved over the period after implementation of electronic medical record, with statistically significant improvement in all domains except for stress recognition [16]. The other studies did either not reach statistical significance, did not report results for primary and tertiary care separately or only used SAQ as a framework for qualitative interviews [57‐59]. Moreover, none of the interventional studies were conducted in nursing and residential homes or home care. Thus, no firm conclusions can be drawn from this review.

It has been suggested that interventions that aim to improve one aspect of the safety climate are likely to positively influence other aspects too [18]. Nursing and residential homes should, therefore, not feel obliged to invest in extensive programmes that focus on all safety dimensions at the same time [18]. Additionally it has been suggested that patient safety improvement work in general practices and out-of-hours clinics should not only address all work sites in the same way, but focus on site specific challenges at places with lower scores on specific patient safety climate factors [23]. It has been suggested that differences between safety specialists' and workforce groups' beliefs about how to improve patient safety may impede the successful implementation of patient-safety programmes [76]. Now that SAQ has been validated for use in various settings in primary care it is time to investigate and evaluate interventions that may improve patient safety culture.

Our analysis found that SAQ has been validated and recommend for use in several countries, descriptive and comparative data have been published, and SAQ has been successfully applied to measure differences in score over time, association to other measures as well as variance across settings in primary care. Thus, SAQ could be implemented in primary care settings to a higher degree than it has been until now.

Our study showed that by only considering variance by ICC at institutional level in primary care, potential for safety culture improvement at work unit level may be masked. An alternative strategy to benchmark institutions is to assess them according to the variance between their subordinate work units. That may be done by assessing institutions by the percentage of work units where 60% or more staff respond positively (with scale score >=75). At this level the safety climate is commonly referred to as “mature”. At the level where 80% of staff in a work unit respond positively (with scale score >=75) the safety climate is commonly referred to as “good”. This terminology is currently used in assessing variance in safety culture between Norwegian hospitals [77]. Such a measure could also be implemented in primary care. It could motivate leaders of institutions and municipalities to engage in direct dialogue with subordinate work units to increase mutual understanding of how staff struggle with their safety culture, and how to improve it.

Variance in SAQ-scores across organizational units were demonstrated. However, understanding the sources of variation in safety culture is still a glaring gap in our knowledge of cultural assessment and a key to deeper understanding and practical application [15]. Thus, more studies on variance in safety climates and possible explanatory factors are warranted.

Although multiple studies have found a correlation between safety culture measured by SAQ and patient harm, this relation has yet to be further explored in primary care [78, 79]. Some studies found associations between patient safety culture and adverse event reports. One study found that more events and near misses were reported when there was a strong culture of safety [54]. This supports the widely held belief that low rates of incidence reporting are linked to poor safety climate and higher levels of patient risk. However, another study found that fewer numbers of medication errors were reported with higher scores of Teamwork climate, Safety climate, Job satisfaction, Perception of management, Working condition and overall safety culture [47]. Only for the dimension Stress recognition, greater numbers of medication errors were reported as Stress recognition increased [47]. Due to few inconsistent results, exploring adverse event rates in primary care and how rates in specific work sites correlate with safety culture measurements is warranted and can also motivate improvement effort.

With the increasing need for care and nursing in an aging population, studies supporting the practical application of patient safety measures in home care and nursing homes are highly welcome in the future. However, the concept of safety climate may be relevant for all organizations that operate with risk [80]. Variance in safety climate is therefore relevant to assess in other public sectors that operate with risk or care for vulnerable users, like for example child protection services [81]. It could also be relevant for social services, the police sector, and services for unemployed and for refugees.