Overall, we found rather weak and statistically non-significant associations between adverse psychosocial working conditions and subsequent concerns to have made an important medical error. Poor collaboration was found to be moderately predictive of the concern to have made an important medical error across all time periods examined. This association was mediated by the potential intermediate factors vigor, depression, and anxiety. In addition, a potential trend towards positive associations was observed for higher workload and poorer practice organization with the outcome.
Comparison to prior research
The few earlier prospective epidemiological studies partially found links between psychosocial working conditions and patient safety [
13,
15,
19,
20]. The comparison with the existing literature is limited though as the specific combination of exposure and outcome constructs used in this study of unfavorable psychosocial working conditions with patient safety in terms of medical errors has only been applied in one prospective study [
15]: That study – carried out among nurses in Japan – documented a link between several job stressors (i.e., nursing stress scale e.g., conflicts with supervisor or colleagues, high workload, lack of support), and the subsequent self-reported frequency of near misses and adverse events combined into a medical error risks variable [
15]. However, those job stressors were combined into a single variable thereby limiting comparability with the specific work stressors and resources (e.g., collaboration, workload) analyzed in our study. Therefore, the definitions of psychosocial working conditions and patient safety in terms of medical errors were broadened to include safety culture [
13] and other indicators of the quality of care or overall patient safety [
19,
20] to draw on further prospective studies: A study among hospital physicians in Germany suggested associations of high social stressors (e.g., conflicts with colleagues and supervisors) and time pressure with the physicians’ perception of the impairment of the quality of care they provide due to working conditions [
19]. A study from Switzerland analyzed and confirmed the link between teamwork and clinician-rated overall patient safety [
20]. By contrast, a study among hospital staff in France that used observer-based ratings to assess medical errors (e.g., error of execution or error of planning) and adverse events found only weak and statistically non-significant links between safety culture and medical errors [
13]. Our study adds to the sparse existing literature by providing the first prospective evidence for MAs and thus for health care staff that is mainly employed in the outpatient care sector.
Prior cross-sectional evidence from our study suggested that the ERI components (i.e., high effort, high reward, high ERI ratio) as well as poor collaboration and poor practice organization are strongly and significantly associated with the concerns to have made an important medical error in the last 3 months among MAs [
6]. The results from our current prospective study may support those earlier cross-sectional findings in some instances, i.e., by highlighting poor collaboration and possibly poor practice organization to be predictive of the concern to have made an important medical error among MAs. In our study, the construct collaboration is defined by interpersonal relationships (e.g., conflicts with colleagues or supervisors, unfair treatment), while the processes of cooperation are covered by the factor practice organization (e.g., well-structured work processes, responsibilities). Our results illustrate a rather moderate relationship of poorer collaboration with the concern to have made an important medical error. The notion of a relationship of collaboration with patient safety is in line with the abovementioned prospective studies [
15,
19,
20]: Whereas the study among hospital physicians in Germany found that pronounced social stressors (e.g., conflicting relationship with colleagues and supervisors) are directly related to a lower self-rated quality of care [
19], the study from Switzerland focusing on the interplay between different types of teamwork and physicians’ rated overall patient safety in intensive care units (ICU) found that interpersonal teamwork (i.e., equivalent to collaboration as measured in this study) predicted patient safety only indirectly, that is, through team organizational and coordination behaviors [
20]. The authors of the latter study argued that positive interpersonal teamwork may enhance communication within the team, which in turn increases patient safety. This hypothesis is supported by the longitudinal study among nurses in Japan which found a lack of communication to be associated with job stressors including collaboration, which in turn were associated with the risk of medical errors [
15]. Therefore, communication may potentially explain the link between collaboration and patient safety observed in our study. In addition, we found that poor mental health (i.e., depression and anxiety) and vigor, may partially explain the observed positive relationships between poor collaboration and reported concerns to have made important medical errors. Poor mental health may develop due to poor collaboration [
10] and may in turn contribute to subsequent medical errors [
13,
38]. This is in line with the study by Tanaka et al. (2012) which found depression to be an intermediate factor between job stressors and perceived risk of medical errors. Further, one may speculate that communication is another factor on that pathway as poor mental health is often associated with difficulties in social interactions and thus likely also with poor collaboration [
39]. In terms of positive wellbeing that can emerge from favorable working conditions, good collaboration may lead to work engagement in the form of vigor [
40]. The high work energy and work-related persistence associated with vigor may in turn reduce the risk of medical errors [
29,
30]. Our study may illustrate that poor collaboration contributes to worse patient safety and that this relationship may be partially mediated by depression, anxiety and vigor.
In our study, we observed a potential pattern of positive associations of higher workload and poorer practice organization with an increase in reported concerns to have made an important medical error. The longitudinal study among hospital physicians in Germany found that time pressure was linked to a diminished physician-rated quality of care [
19]. Time pressure, captured by the factor “workload” in this study, was assumed to be a structural deficiency within the care system that prevents health staff from performing their tasks effectively [
19]. An observer-based study among ICU health staff in France found that high workload increased the risk of medical errors by almost 50% [
13]. A high workload in combination with suboptimal staff planning was considered part of the organizational factors hypothesized to increase medical errors [
13]. Inefficient practice organization has been reported to add to the burden of workload in a qualitative study among MAs [
41]. Our results indicate that poorer practice organization may potentially be an independent determinant of reported concerns to have made an important medical error. A cross-sectional study among physicians in primary care from the US measured the perception of the atmosphere at the workplace (i.e., ranging from calm to hectic or chaotic) and found that chaotic workplaces were associated with higher rates of medical errors [
42]. An observer-based study among MAs in Germany found that well-structured tasks and a clear responsibility was relevant for a functioning workflow, as MAs’ work processes were frequently interrupted [
43]. Interruption in turn may lead to a reduced patient safety in terms of medical errors [
44]. Overall, poor practice organization and high workload are processes and structural determinants that may facilitate the occurrence of errors [
45]. It needs to be mentioned again though that the corresponding associations were weak and non-significant in our study. We hope however that our discussion stimulates more research into those potentials determinants of patient safety.
Methodological considerations
A strength of this study is its prospective design. Another strength is our comprehensive assessment of psychosocial working conditions of MAs, which relied on both an established generic instrument (e.g., the ERI model) and an instrument specifically developed for MAs.
While these two instruments represent a broad spectrum of psychosocial working conditions that are key working conditions according to MAs [
1], further relevant work stress models such as the job demand-control model or organizational justice [
46,
47] may provide additional insights into the psychosocial working conditions that predict errors [
48,
49]. Unfortunately these concepts were not included in this study.
A limitation of this study is the rather small sample size for prospective analyses. This may have restricted the statistical power, which limits the detection of statistically significant associations. The links of collaboration and the pattern of positive association of workload and practice organization with the outcome may have been random findings due to multiple testing. After Bonferroni correction those estimates were not significant anymore. The frequency of reported concerns to have made an important medical error was low and thus yielded a limited number of cases (i.e., only 45 cases for our primary outcome variable). This limits the feasibility to adjust for a large set of confounders; however we needed to adjust for two confounders only (i.e., age and leadership position).
Selection bias cannot be ruled out. However, firstly, thorough non-responder analysis (see supplementary material Table A
1) did not yield significant differences regarding exposure variables (e.g., psychosocial working conditions). Secondly, our study sample is fairly representative of the German MA population according to the Federal Statistical Office of Germany in terms of sex, age, employment status [
22] as well as comparable in terms of age, work experience, marital status to a previous study among MAs in Germany that claims to be representative [
26].
Our study relied on longitudinal assessments at two time points. However, a three-wave study had been superior to analyze the intermediary position of potential intermediate factors between the relationship of psychosocial working conditions and patient safety, as intermediate factors e.g., depression could also be conceptualized as a shared cause of poor collaboration and medical errors [
14,
39,
50].
A further limitation, which could introduce information bias is that we measured patient safety by self-reported “
concerns to have made an
important medical error”. Participants were not given a definition of an
important medical error. Therefore, the understanding of what an important error constitutes may have differed across MAs. As a consequence, some MAs may have subjectively considered some errors as unimportant and accordingly did no report them. In addition, important errors may have been made, but immediately corrected by the MA or a supervisor and therefore are not recalled as errors. This might have reduced the likelihood to report errors. Minimization of such potential under-reporting bias might not only have improved the validity of our findings, but also the precision of estimations due to a higher number of reported errors. The approach to assess “important medical errors” allowed us to obtain information on all types of errors rather than one specific type of error. In addition, errors were assessed by
concerns and not perceived actual error frequency. Reasons for concealing errors are attributed to fear of individual accountability, judgement of capability, and legal consequences [
51]. The approach to measure
concerns, which has been applied in further studies [
12], therefore may have the advantage of lowering the threshold to report errors as participants feel less exposed to social desirability and possible legal consequences. Nevertheless, pronounced c
oncerns may occur as a symptom of depression and anxiety. Therefore, we ran additional analysis excluding participants with poor mental health (i.e., depression or anxiety) from the primary analysis to rule out confounding. The estimates of the primary analysis were not altered though (data not shown).
In addition, it should be emphasized that many wellbeing concepts - e.g., depression and overall health - can deteriorate independently from work and therefore cannot be conceptualized per se as work-related. We therefore recommend to additionally examine burnout in future research as the development of this syndrome is closely tied to the workplace, which unfortunately was not feasible in this study.
Recommendations for future research and the practice
Further studies are needed to support our findings. Those studies should be prospective and focus further on working conditions as a potential starting point rather than wellbeing as it might act as an intermediate factor [
15]. By identifying the adverse working conditions associated with perceived medical errors, it is possible to intervene at the core structures and processes of medical errors rather than treating the consequences of these deficiencies in terms of impaired wellbeing. Moreover, wellbeing should be analyzed as a potential intermediate factor in greater depth, with a stronger emphasis on salutogenic wellbeing constructs (e.g., work engagement). So far, only pathogenic wellbeing constructs have been tested longitudinally for mediation of the relationship of working conditions with patient safety [
15,
19,
20].
Although the significance of our findings still needs to be substantiated by further prospective studies, our study suggests that poor collaboration may be a promising starting point in order to address patient safety in outpatient care. In practice, this could be addressed through regular team meetings of the entire team as well as solely among the MAs, constructive feedback sessions with the supervising physician(s) to strengthen communication, and involvement of MAs in staff-related decisions and team activities [
52,
53]. As structural processes and delegation of tasks in the workplace are factors contributing to patient safety [
54], interventions should focus on practice organization and workload by efficiently structuring the daily practice routine, e.g., minimizing patients’ stay at the reception to reduce the likelihood of concurrent stressors, relocating answering of the telephone to a separate room and clearly assign work tasks and responsibilities [
52,
53]. Finally, the implementation of error management systems in outpatient care practices is needed to strengthen error reporting and establish clear responsibilities within the team for error management, which could promote communication about errors and subsequent lead to higher patient safety [
55].