Influencing factors on the health of nurses—a regression analysis considering individual and organizational determinants in Germany

Data were collected from June to October 2021. Employees of the participating care organizations were invited to take part in the survey. The inclusion criteria were 1) a minimum age of 18 years and 2) employment in a care organization participating in the BAGGer project (ACH, ICF, or OCS). Exclusion criterion was a professional activity outside of nursing in terms of patient or resident care.

Questionnaires were sent to all 3,484 employees in the participating care organizations with 2,200 working in nursing care (ACH n = 1,028; ICF n = 544; OCS n = 628) as nurses, as nursing assistants or nursing aids (e.g. in the form of a long-term nursing internship). The survey could be completed in each care organization in paper–pencil or digital format over a two-week period. Participation in the survey was voluntary and anonymous (cover sheet/ start page with information for participants about the research topic, objectives, and contact information; implied consent).

The survey comprised questions on sociodemographic aspects, work demands, and employee health.

The following sociodemographic information was assessed: gender (female, male, non-binary), age (18–20 years, 21–30 years, 31–40 years, 41–50 years, 51–60 years, > 60 years), profession (nurse, nursing assistant, nursing aid, other with free-text field) and working time per week (≥ 35 h, 15-34 h, < 15 h).

Work demands were assessed by care setting-specific questionnaires (BGWmiab) developed by the German Institution for Statutory Accident Insurance and Prevention in the Health and Welfare Services. The BGWmiab questionnaire contains 22 questions adapted to the specific care setting rated on a 5-point scale from "no, not at all” to "yes, exactly" [29]. The BGWmiab compromises the following dimensions: Quantitative workload (QUANTI) considering the amount of work in relation to time (ACH = 4, ICS = 5, OCS = 6 items), qualitative workload (QUALI) considering the emotional demands such as coping with the needs of the patients (ACH = 6, ICS = 5; OCS = 3 items), organization of work (ORGA) meaning the perceived job control or the opportunity to shape and influence working hours and content, including break schedules (ACH/ICS = 3; OCS = 2 items) and social work climate (SOCIAL) referring to the social relationships and support at the workplace (ACH/ICS = 6; OCS = 8 items) as well as after work situation (HOME) assessing non-work well-beeing as indicator of the degree of stress (3 items) [29]. The BGWmiab dimensions were classified according to the handbooks into “low", “under usual”, “over usual” and "high" [30‐32]. These questionnaires have a sufficient to good reliability (Cronbach’s α for the five “work demands” = 0.53–0.86) and high construct validity [30‐32].

Beyond, the work demands of experiencing violence on duty were assessed with the “Survey of Violence Experienced by Staff German-Version-Revised” (SOVES-G-R) [33, 34]. In the SOVES-G-R, three forms of violence are distinguished: verbal violence e.g. abusive or insulting language, personal verbal attacks, swearing, obscene or sexually harassing comments, threats e.g., warnings that you will be hurt, (sexual) harassment, physical intimidation, threats with a weapon and physical violence e.g., hitting, pinching, pushing, shoving, spitting, kicking, using a weapon or object as a weapon [33, 34]. For each violence forms a yes/ no question was asked to determine whether violence had been experienced by patients or family members/ visitors during the past 12 months [33, 34]. The SOVES-G-R has a good internal consistency (Cronbach's α 0.87–0.91) [35].

Employee health was determined with two variables: subjective health (SH) and work ability (WA). In order to assess SH, the question "How is your health status in general?" (5-point scale from "very good" to "very poor") was taken from the Robert Koch Institute's "German Health Update" study (GEDA 2014/2015-EHIS) [36]. This question is also part of the Minimum European Health Module and has a good reliability with a correlation coefficient of 0.81 and kappa estimate of 0.74 for self-perceived health [37].

WA was operationalized with the Work Ability Score where the current WA is compared to the best ever WA on a scale of 0–10 [38]. The scores allow classification into: poor (0–5), moderate (6–7), good (8–9) and excellent (10) WA [39]. The single item question is showing a very strong association with the Work Ability Index [40, 41]. The Work Ability Index has satisfactory reliability (Cronbach’s α = 0.78) and a strong and consistent predictive power in a German sample [42].

First, the data were subjected to a plausibility check by two researchers. Since the data analysis was restricted to employees working in nursing care (see  Setting & participants), the information in the free text field on the profession in particular was critically checked and, for example, persons who indicated interactive or service-providing activities in nursing, e.g. in the form of a long-term nursing internship, were classified as nursing aids.

To describe the sample, frequencies (n) and valid percentages (%) were calculated for gender, age, profession, working time, and experience of violence. Descriptive statistics (means, standard deviation (SD)) were calculated to describe BGWmiab variables of work demands and variables of employee health.

Differences in the variables between care settings were tested by Chi Square Test respectively Kruskal–Wallis-Test. To determine which setting was statistically significant different, a Dunn-Bonferroni test was performed.

To explore the association between care setting-specific work demands and employee health, two binary logistic regression models were calculated. For this purpose, dichotomized SH ((very) good vs. (very) poor/ moderate) and WA (good/ excellent vs. poor/ moderate) were used as dependent variables. Individual sociodemographic variables gender (female vs. male), age (> 40 years vs. ≤ 40 years), profession (nursing assistant/ aid vs. nurse) and working time (≤ 34 h vs. > 35 h) were included in the model as covariates. The organizational work demand variables QUANTI, QUALI, ORGA, SOCIAL and HOME (over usual/ high (high) vs. low/ under usual (low)) as well as verbal violence, threat and physical violence (yes vs. no) were included as independent variables. Care settings (ACH, ICF, OCS) were additionally included as dummy variables to examine the impact of these, with the setting that rated best on the employee health variables used as the reference variable.

The Hosmer–Lemeshow test was used to determine the goodness of fit of the models. The Nagelkerkes coefficient R² and its effect size f² was evaluated to determine the predictive power of the models. According to Cohen the effect size f² = R²/(1- R²) is considered as small (f² ≥ 0.02), medium (f² ≥ 0.15), and large (f² ≥ 0.35) in behavioral science [43]. The Odds Ratios (OR) and 95% confidence intervals (CI) were calculated.

Of 2,200 employees working in nursing care, 528 participated in the survey. This corresponds to a participation rate of 24.0% (ACH = 22.8%; ICF = 27.9%; OCS = 22.6%). The sample consisted of 440 females (84.8%) and 353 participants working as a nurse (66.9%). The largest age group were 142 employees (31.7%) between 51–60 years and 280 (54.2%) worked full-time. A more detailed sample description by care setting can be found in Table 1.

OCS employees were statistically significant older and more females than in ICF or ACH. In terms of profession, more nurses in ACH participated in the survey than in the other settings. In addition, working time varied between settings: in ACH, most full-time employed respondents were present, which was fewer in ICF and even fewer in OCS. Besides differences in individual determinants, there were also statistical significant differences in the organizational variables of work demands between the care settings. The variables QUANTI and HOME were perceived more demanding in ACH than in OCS and also more in OCS than in ICF. Further verbal violence and physical violence were experienced more often in ACH than in ICF and also more frequent in ICF than OCS. It is particularly noticeable that in ICF the QUALI is perceived statistical significantly lower compared to other settings and threats are experienced more often in ACH than in the other settings. Employee health differences between settings were found in both variables. SH was rated in all settings between good (2) and moderate (3), differing statistically significant in ACH and OCS. However, the situation is different for WA, which was rated in all settings between moderate (2) and good (3), in ICF which was statistically significantly higher than in ACH or OCS (see Table 1).

The results of the regression analysis on SH can be found in Table 2. The regression analysis showed that the model on SH was statistically significant (Chi² (14) = 28.97, p = 0.011, n = 410). Age over 40 years was negatively associated with a (very) good SH (OR = 0.650, 95% CI = 0.424—0.996). Experiencing over usual or higher demands in ORGA decreased the likelihood of reporting (very) good SH (OR = 0.595, 95% CI = 0.362—0.978). Working in ICF or OCS decreased the odds of reporting (very) good SH compared to working in ACH but this is not statistically significant. The coefficient of determination is R² = 0.092 which corresponds to Cohens f² = 0.10, a small effect size.

The results of the regression analysis on WA can be found in Table 3. The second regression also showed that the model (Chi² (14) = 72.00, p < 0.001, n = 405) was statistically significant. There was a statistical significant negative association of increasing age and “good/ excellent” WA (OR = 0.555, 95% CI = 0.353—0.875). For the work demands variables, the odds of reporting “good/ excellent” WA decreased when QUANTI (OR = 0.565, 95% CI = 0.343—0.930), ORGA (OR = 0.520, 95% CI = 0.315—0.858) and SOCIAL (OR = 0.610, 95% CI = 0.376 – 0.991) were perceived as over usual or high demands. The odds of reporting “good/ excellent” WA were lower in ACH and OCS than in ICF, however this is not statistically significant. The coefficient of determinants is R² = 0.218 and Cohen f² = 0.28, which correspondents to a medium effect size.

German sick leave data on nursing care employees indicate that sickness rates increase with age, being female, assistant staff or part-time employed [2, 3, 44]. In this study, only age was identified as an influencing factor on employee health variables predicting sick leave.

First, sociodemographic data are discussed to determine if the study population differs from the German nursing population. While in the presented study 58.7% of all nursing employees have the age > 40 years, in Germany this age group is represented by 62.0% [45]. Moreover, in Germany 83% of the nursing workforce is female, here it is 84.5% [24]. In regard to profession, the proportion of nursing assistants and nursing aids in Germany is nearly the same in the presented study for ACH 11% (here 11.1%) and ICF 47% (here 47.3%) but lower in OCS 39% (here 44.2%) [24]. It is also known that nursing employees are more likely to work part-time with increasing age [2]. While the respondents in this study are younger than in Germany this is also reflected in the working time per week. In Germany, 57.6% of the nursing care employees work part-time [45], whereas in this sample it is 45.8%. This could support the findings that health is more influenced by age than working time.

Although work ability is a predictor of sick leave, studies also showed the contrary effect that being female gender is positively associated with work ability among nursing assistants [46]. Other authors, on the other hand, conclude that lower professional status leads to more physical complaints [47]. However, it also appears that age and gender have no influence on emotional exhaustion in employees with the profession nurse [48]. If only the subjective health is considered, females in Germany rate it worse than males, but the difference is only statistically significant in the age group of 18 to 29 year-olds [49], which could be one reason why gender does not appear as an influencing factor in this study. Consistent with the findings of this study, younger age is also associated with a higher work ability [46, 50]. Although no association between older age and physical work ability can be found when nurses have low levels of burnout symptoms [50]. Therefore, it seems that in addition to physical fitness, personal coping with stress is also important when it comes to employee health as we age.

All included work demands are known to be associated with burnout [9, 13]. However, in the present study only organization of work, quantitative workload, and social work climate showed an statistical significant effect on employee health variables. In the existing international literature, resources in autonomy, i.e. low demands in organization of work, demands in quantitative workload and resources in various aspects of social work climate are also identified as key issues in health-related outcomes [51]. This is supported by the findings of nurses in Germany, which show that low decision latitude in the organization of work, high quantitative workload, and a weak percieved social work climate are key stressors associated with emotional exhaustion [48].

Organization of work is perceived as particularly demanding when processes are determined by others. One aspect that seems to be responsible for this percieved low job control is the frequent filling in for colleagues, leading to more (weekend) working hours [52, 53]. The burdens of compliance with working time regulations have been further exacerbated by the pandemic in Germany [54]. In this context, it is known that overcommitment, e.g. when filling in for colleagues, and working more than two weekends per month are negatively associated with subjective health in nursing care employees [55]. Moreover, demands in organization of work are characterized by work and break interruptions, which requires a high level of multitasking [53]. In addition, both work and break interruptions are often associated with increasing quantitative workload [56, 57]. Occupational health research in Germany has found that nurses who experience frequent work interruptions and multitasking not only show signs of mental exhaustion, but they also rate their overall work ability lower [57]. It is known, especially among older employees, that the negative stress consequences decrease with additional short rest breaks [58, 59]. Short breaks not only have the potential to reduce stress but also to promote social interaction [60]. This seems to be desirable, as negative effects on employee health due to quantitative workload are buffered when the social work climate is experienced as good [61].

Unlike research on mental health as an outcome, qualitative workload, after work situation and violence showed no statistical significant association with employee health variables. This could be related to the fact that qualitative workload is perceived as demanding by caregivers especially when exposed to it alone without peer support [53]. Especially since the beginning of the pandemic, social support and trust among colleagues and with superiors have become even more important for nursing care employees in Germany [52]. In this context it is known that a high level of social support is capable of being a mediator of mental strain even since the outbreak of the pandemic [62]. In this study sample, demands in social work climate were rated as under usual in all care settings. Therefore, it could be that the positively rated social work climate buffered the effects of qualitative workload in this study sample. The questions on after work situation are indicators of the degree of stress [29]. Although after work situation was found over all settings as over usual demanding no statistical significant association was found. Among other things, this could be due to the fact that German nurses who find it problematic to switch off after work, or even experience mental, physical and social burdens, nevertheless seem to consider their health as good [53].

Furthermore, violence was experienced far less frequently in this study than in Germany, which could be one reason why no statistical significant association was found. Within one year, almost every nurse in Germany experiences verbal violence (84–97%), 21–55% experience threats, and 61–77% experience physical violence [11, 12]. In comparison, 52.3% experienced verbal violence, 16.2% threats, and 27.4% physical violence in this study. One reason for this could be that caregivers do not perceive assaults as violence if they associate them with the illness of the person in need of care, even if they are experienced on a daily basis [53]. It is known that nurses who have received training on violence are more likely to perceive assaults as violence, thus increasing the number of incidents [63]. Therefore, it might be beneficial to further raise awareness of this issue among nursing care employees.

It is known that absenteeism due to illness varies between care settings in Germany: ICF has the highest followed by OCS and in ACH these are the lowest [1]. In contrast, the present study has shown that the care setting made no statistically significant contribution to employee health. Supporting this, no statistical significant differences in subjective health between settings are found in Germany [64]. In contrast, statistical significant differences in physical health were found between employees of ACH and ICF or OCS in Germany [65]. Other authors further describe differences in work ability between German care settings [66]. However, it should be noted that the data in those studies are purely descriptive and not adjusted to individual or organizational determinants.

Therefore, it seems that individual and organizational factors are of considerable importance for the employee health of nursing staff in Germany, regardless of the care setting.

The aim of the study was to find approaches for health promotion programs in nursing. Thus, a strength of this study is that employee health was not limited to mental health. In addition, the data were generated in different nursing settings, which means that the statements on the influencing factors relate to a broader target group of nursing staff.

However, the results of this study must be considered with some limitations.

One issue is that the data were collected in organizations participating in a health promotion project which could indicate a possible selection bias. In the participating organizations, it is particularly in ICF and OCS that there is a high level of health promoting capactiy, which depends on health promoting willingness and structured health promoting management [67]. Therefore, it could be assumed that employees in these organizations experience fewer work demands or that these are already being addressed in a preventive manner.

Furthermore, the response rate of the 2,200 employees working in nursing care was only 24%. It remains unclear whether individuals with lower or higher employee health participated in the survey. In terms of individual determinants, it is also not known whether the respondents have private care duties or financial responsibilities. Although it was known that work-family conflicts and lack of formal rewards could have an impact on employee health [9, 51], they were not fully captured by the chosen instrument. This could be considered in future studies.

In the data analysis, the variables were dichotomized and employee health was accessed with single item questions, which can lead to bias. The regression models in this study showed a small effect size for SH and a medium effect size for WA, so the results should be interpreted with some caution.