Effects of active on-call hours on physicians’ turnover intentions and well-being

Effects of active on-call hours on physicians’ turnover intentions and well-being. Scand J Work Environ Health 2008;34(5):356–363 . Objectives This study examined whether active on-call hours and the co-occurrence of lifestyle risk factors are associated with physicians’ turnover intentions and distress. Methods Cross-sectional survey data on randomly selected female (N=1571) and male (N=1081) physicians, aged 25 to 65 years, from The Finnish Health Care Professionals Study were used. The outcome measures were turnover intentions and distress (general health questionnaire). Smoking, heavy drinking, overweight, and low physical activity were assessed as lifestyle risk factors. Analyses of covariance were used to analyze the data. Results After adjustment for gender, age, employment sector, and job satisfaction, the analyses showed that the physicians who had been on active call more than 40 hours per month reported more distress than the group not on call (P=0.046). The physicians with two or more risk factors also had more distress (P<0.001) than those with no risk factors or only one risk factor. There was an interaction between active on-call hours and lifestyle risk factors for turnover intentions (P=0.002). The physicians with two or more lifestyle risk factors who had been on active call more than 40 hours per month had more turnover intentions than the other physicians. Conclusions On-call duty and the co-occurrence of lifestyle risk factors may both decrease physicians’ well being and increase their intentions to leave their job.

Physician shortage and turnover are major problems worldwide. High physician turnover has many negative consequences, such as substantial financial losses, patient dissatisfaction, and worsened physician morale (1). High demands at work have been found to be associated with physicians' turnover intentions (2). Physicians' work has been shown to be very or extremely stressful (3). For example, burnout (4) and mental health problems (5) seem to be fairly common among physicians.
One important source of stress among physicians is their on-call duties. British general practitioners rated nights on call as one of the top two most stressful aspects of their work (6). Among Finnish anesthetists, being on call was the most frequently mentioned stressor; it correlated strongly with various stress symptoms and was the greatest reason for sleep deprivation (7). What makes on-call work especially strenuous is that it often combines both long workhours and night work. Long workhours have been found to be related to increased stress levels, lower quality of life, and poorer lifestyle (8). Working overtime may lead to work-related injuries (9), health problems (10,11), work-family interference (12), exhaustion (13), job dissatisfaction (14), and risky health behavior (15,16). Overtime work also impairs cognitive performance in the areas of attention and executive functioning (17). Shift work and, particularly, night work can interrupt sleep patterns and lead to problems with physical and mental health (18)(19)(20)(21)(22). For example, long night shifts have been associated with alcohol use among nurses (23).

Heponiemi et al
Long on-call periods tire physicians and affect their work performance. Physicians on 24-hour on-call shifts are at increased risk of occupational injury, making serious or even fatal medical errors, and motor vehicle accidents on the drive home (24). After a 36-hour inhospital call, significant decreases have been found in visual attention, short-term recall, and coding ability (25). Among residents, post-call performance during a heavy call rotation was found to be comparable to or worse than the impairment observed in relation to a 0.04 to 0.05 g% blood alcohol concentration in a test of sustained attention, vigilance, and simulated driving (26). Nighttime on call has been found to deteriorate neuropsychological and cognitive function (27) and to be associated with physical health problems (28). Eliminating interns' extended workshifts and reducing the number of hours worked per week has been found to increase sleep, decrease attentional failures, and decrease serious medical errors (26,29).
On-call work schedules may also deteriorate mental health (30). For example, among general practitioners, the number of on-call duties undertaken and one or more nights on call per week have been found to be associated with anxiety and depression (31,32). Among general practitioners in Leeds, England, the amount of time spent on call was associated with psychological symptoms measured with the general health questionnaire (GHQ) (28). Physicians' mood has been shown to be significantly lower when on call than when off call (33).
The on-call burden may contribute to negative coping, such as an unhealthy lifestyle, which may, in turn, further intensify the harmful effects of the on-call burden. Some previous studies have suggested that there are factors in physicians' work that may lead to unhealthy behavior. For example, heavy drinking has been shown to be a problem among physicians (34). Lifestyle risk factors, such as smoking, heavy drinking, obesity, and physical inactivity, tend to aggregate (35), and high job strain increases the likelihood of such a phenomenon (36). The combined effect of these risk factors may be synergistic rather than additive; for example, the disease risk associated with a combination of risk factors has been found to be higher than the sum of their independent, additive effects (37).
Using the GHQ, we examined the effects of the number of active on-call hours, the co-occurrence of lifestyle risk factors (smoking, heavy drinking, overweight, and low physical activity), and their interaction on turnover intentions and distress in a large sample of Finnish physicians. We adjusted for the effects of gender, age, job satisfaction, and employment sector because previous studies have shown that all of these factors may have a significant effect on intentions to leave or stay in a job (1,(38)(39)(40).

Study context and study population
The Finnish health care system has been described as one of the most decentralized in the world. Finland has 5.2 million inhabitants, and the country is divided into 416 municipalities, and even the smallest ones are responsible for arranging health care services for their inhabitants. Being on call is an integral part of physicians' work in Finland, especially among those working in the public sector. The organization of on-call services, especially in primary health care, has gone through significant changes in recent years. The number of health centers providing round-the-clock on-call services has decreased from 200 to 45 in the past 20 years (41). On-call services have been centralized to bigger units and also transferred to hospitals. A hospital physician's work period, which includes on-call duty, lasts normally 24 hours, starting from 0800. Recently it has become common in some clinics to split the 24-hour work period between two physicians, especially on weekends. Over half of all Finnish physicians are on call regularly, the proportion being significantly bigger among hospital physicians (74%) and health center physicians (primary health care) (65%). In 2006, 31% of the health center physicians were actively on call once a month, 35% twice a month, and 15% three times a month, whereas among hospital physicians 18% had active on call once a month, 23% had it twice a month, and 31% had it three times a month (42). In 1997, the weekly average of active on-call hours was 5.3 for male physicians and 5.6 for female physicians (43).
We drew a random sample of 5000 Finnish physicians (30% of the total physician population) from the 2006 database of physicians, maintained by the Finnish Medical Association (register covers all licensed physicians in Finland), as a part of the Finnish health care professionals study. Questionnaires were mailed in the autumn of 2006. The nonrespondents were reminded and sent the questionnaire up to two additional times. Responses were received from 2841 (57%) physicians. We excluded 189 respondents who had incomplete information; thus the final sample included 2652 physicians (1571 women and 1081 men) aged 25-65 (mean 53.0, SD 5.18) years. The sample was representative of the eligible population in terms of age, gender, and employment sector (44).

Survey
Active on-call hours were assessed with a question asking how many active on-call hours the respondent had had during the last month. On-call hours were categorized as (i) no on calls, (ii) <40 hours per month, and (iii) >40 hours per month. Turnover intentions were Effects of on-call hours and physicians' turnover intentions assessed with the use of three questions concerning willingness to (i) change to other physician work, (ii) change to a profession other than that of a physician, and (iii) quit the present job. They were coded as 1 (no), 2 (perhaps), or 3 (yes). Psychological distress was measured with the 12-item version of the GHQ with the response options ranging from 1 to 4 (45). The reliability, as measured by Cronbach's alpha, was 0.89 for this sample. We used the mean response score in the analysis.
The examined risk factors were smoking, heavy drinking, overweight, and low physical activity. Smoking was included as a health risk and coded as 1 if a respondent (i) was a current smoker, (ii) had quit regular smoking within 5 years, or (iii) had smoked regularly for >10 years. Otherwise smoking was coded as 0. The cutoff point for overweight was a body mass index (BMI) of ≥25 kg/m 2 (46). The cut-off point for heavy drinking was a weekly consumption of ≥190 g of absolute alcohol for the women and >275 g for the men (47). Physical activity was measured with the mean of three questions assessing (i) the frequency of physical activity "How often do you engage in leisure-time physical activity in which you sweat or become breathless?", (ii) intensity of physical activity "How much breathlessness and sweating do you experience when you engage in physical activity and sports?", and (iii) the average duration of a physical activity session "How many hours do you usually spend in a physical activity session in which you sweat or become breathless?" The answers were coded from 1 (total inactivity) to 6 (vigorous activity). The participant was coded as being low in physical activity if the mean was ≤3, which represented 20.8% of the participants. This cut-off point was chosen on the basis of previous findings showing that the prevalence of physical inactivity was 18% and 27% for Scandinavian men and women (48) and 18% and 22%, respectively, among white Americans (49).
All four risk factors were coded as binary (0/1) for the analysis. Co-occurrence was defined as the number of risk factors for which the individual participant in question had a high risk value. If a participant belonged to the high-risk group for one risk factor, the corresponding co-occurrence score was 1; if the participant belonged to two high-risk groups, the co-occurrence score was 2, and so forth. Because the number of physicians having three or four risk factors was very low (3.5% had three risks and 0.4% had four risks), we collapsed the co-occurrence score into (i) 0, (ii) 1, and (iii) ≥2 risk factors.
Job satisfaction was measured with 10 items from the job diagnostic survey (50). We used scales measuring satisfaction with social relations at work, satisfaction with personal growth at work, and overall satisfaction with the job (7 = very satisfied to 1 = very unsatisfied). The Cronbach's alpha was 0.81 for the reliability in this sample.

Statistical analyses
Analyses of covariance (ANCOVA) were conducted that included turnover intentions as the dependent variable and the categorized on-call variable, the co-occurrence of lifestyle risk factors, and their interaction as independent variables with control for gender, age, employment sector, and job satisfaction. A similar series of ANCOVA were generated in which distress was the dependent variable. Bonferroni-adjusted pairwise posthoc tests were carried out for significant associations. Significant interactions were graphed. Because there were no significant gender differences in the associations between on call and outcomes, the analyses were conducted in the combined sample of men and women. Table 1 shows the characteristics of the study sample. Over half of the participants had not been on active call the preceding month, and approximately 11% had been on active call for >40 hours. Altogether 35% were overweight (BMI ≥25), and the mean BMI was 24.3. A total of 12% were considered heavy drinkers, consuming >275 g (men) or 190 g (women) of absolute alcohol per week. The co-occurrence of lifestyle health risks was not very common; namely, 43% did not have any risks and 20% had two or more health risks. We also examined the responses to the GHQ for caseness, and 22% of the participants scored ≥4, indicating high levels of psychological symptoms (51).

Results
The results of the analyses of covariance are presented in table 2. On-call hours were associated with distress, and pairwise comparisons (P=0.045) highlighted the fact that those who had been on active call >40 hours per month had more distress than those who had not been on call at all. The main effect of on-call hours was unrelated to turnover intentions.
The co-occurrence of lifestyle risk factors was associated with turnover intentions and distress. Pairwise comparisons (P=0.002) highlighted that those who had two or more lifestyle risk factors had more turnover intentions than those without risk factors. In addition, pairwise comparisons showed that those who had two or more lifestyle risk factors had more distress than those who had no risk factors (P<0.001) and those who had one risk factor (P=0.003).
The interaction of active on-call hours and the cooccurrence of lifestyle risk factors was significant for turnover intentions (F=4.14, P=0.002). Figure 1 shows that the physicians who had two or more lifestyle risk factors and had been on active call >40 hours per month had more turnover intentions than the others. There was no interac- Heponiemi et al tion between active on-call hours and the co-occurrence of lifestyle risk factors for distress (F=0.23, P=0.923).
Age and job satisfaction were negatively associated with turnover intentions. Job satisfaction was also negatively associated with distress. The employment sector was significantly associated with turnover intentions, and the pairwise comparisons (P=0.002) showed that the physicians working in health centers had more turnover intentions than those working in hospitals.
We also examined the responses to the GHQ for caseness with a cut-off point of ¾ (≥4 indicates high levels of psychological symptoms) (51) as a dependent variable with binary logistic regression analyses. The results correspond to those with continuous mean response scores. Those who had been on active call >40 hours per month had a higher risk of being over the cut-off point for psychological symptoms than those who had no oncall hours (Wald's c 2 =5.55, P=0.018) or had <40 hours of active on call per month (Wald's c 2 =4.23, P=0.040). Those who had two or more lifestyle risk factors were more likely to be over the cut-off point for psychological symptoms than those who had no risk factors (Wald's c 2 =12.99, P<0.001) or those who had one risk factor (Wald's c 2 =12.77, P<0.001). The interaction between on-call hours and the co-occurrence of lifestyle risk factors was insignificant also in these analyses.

Discussion
Our study examined the effects of on-call burden, the cooccurrence of lifestyle risk factors, and their interaction on turnover intentions and distress among Finnish physicians. Our study found that a high on-call burden was associated with high distress levels and being over the cut-off point for psychological symptoms but not with turnover intentions. In a similar way, previous studies have highlighted the stressfulness of on-call work, especially that of night on calls (6,7,27,52). Moreover, our findings showed that co-occurring lifestyle risk factors were associated with higher levels of turnover intentions and distress. In addition, we found that the physicians who had two or more lifestyle risk factors and a high on-call burden had higher levels of turnover intentions than the other physicians.
Our results imply that a high on-call burden threatens the mental health and well-being of physicians. Numerous previous studies have highlighted the stressfulness of on-call work, especially that of nights on call (6,7,27,28,52). On-call work was found to be a major source of stress and job dissatisfaction among New Zealand general practitioners (53). Similarly, among Irish pediatric hospital physicians, the main reasons for work-related stress were long workhours and subopti-mal on-call conditions (54). Among Japanese female physicians long workhours and engaging in night duty have been found to be associated with psychological symptoms measured with the GHQ (55). In addition to mental health problems, on-call work may also induce sleepiness (52), deteriorate work performance (24)(25)(26)(27), and impair physical health, given that nightshift workers may be at an increased risk of experiencing, for example, cardiovascular, gastrointestinal, and reproductive dysfunctions (22).
According to our results, co-occurring risky lifestyles are associated with mental health problems and turnover intentions. Especially prone to leave seem to be those who simultaneously have a high on-call burden and many lifestyle risks. Previous studies have shown that health behavior is related to mental health and job satisfaction. For example, smoking and drinking have been found to be associated with depression and job dissatisfaction Effects of on-call hours and physicians' turnover intentions (56)(57)(58)(59). Moreover, physical exercise and obesity have been found to be associated with mental health (60,61). Previous studies have also found associations between work conditions and health behavior, such as diet, physical activity, and alcohol consumption (62)(63)(64). We found that younger physicians had more turnover intentions than their older colleagues. No consistent association has been found between age and turnover (1), but our finding is contrary to a previous finding that the strongest predictor of both physician's intentions to leave clinical practice and actual departure was advancing age (65). One reason for the discrepancy in the results may be the fact that the aforementioned study also included retirement intentions in their measure, and it was not included in our study. Studies of the relationship between age and job satisfaction have

Heponiemi et al
yielded inconsistent results as well. Earlier research has supported negative linear functions, positive linear functions, U-shaped functions, J-shaped functions, and no relationships between these aspects (66)(67)(68). Work conditions may have an influence on this relationship. One study found support that the association would be negative in unfavorable work conditions and positive or some other form in other settings (69).
In accordance with previous studies, our results suggest that job satisfaction is important for the well-being and retention of physicians (1,28). Thus promoting job satisfaction seems to be of the utmost importance in health care settings. However, a recent study implies that, although job satisfaction among physicians is associated with intentions to quit their jobs, it is not associated with actual exit (65).
Because our study was cross-sectional, we cannot draw any causal inferences. It is possible that the studied independent factors cause increased intention to quit but also other factors, such as differences in the nature of the work, may affect both dependent and independent variables in the analyses. For example, in certain specialties, it may be that physicians are required to work more oncall hours, and, at the same time, the nature of the work in that specialty may be intrinsically more stressful, and may therefore lead to a greater use of avoidant coping strategies, such as alcohol abuse, and greater turnover intentions. In addition, when our results are interpreted, it is important to keep in mind that physicians' intentions to leave their jobs do not necessarily need to translate into action. However, intention to leave has gained support as an important predictor of actual turnover (typical correlation 0.50) (70). Buchbinder et al (71) found that physicians who reported that they were very likely to leave their current practice were 2.4 times more likely to have done so 4 years later than those who reported that they were very unlikely to leave.
The lifestyle risk factors were assessed from self-reports. It is well known that people tend to underestimate their weight and alcohol consumption. In addition, the dichotomization of risk factors may have influenced the results and perhaps reduced the statistical power. Lower cut-off points would have increased the proportion of high-risk persons. Moreover, the cut-off point for physical activity was arbitrary; we labeled the 20% with the least physical activity as being low in physical activity. We did not include all of the important lifestyle risk factors; for example, we did not have information on dietary habits. In addition, although we controlled for several confounders, we could not rule out the possibility of residual confounding.
Further research is needed to confirm that our findings are generalizable. In our present sample of Finnish physicians, 43% had no risks, 38% had one risk, and 19% had two or more co-occurring risk factors, whereas the corresponding figures for Finnish publicsector employees have been found to be 35%, 40%, and 25% (72), respectively, and, for the members of a health maintenance organization in the United States, the corresponding figures were 31%, 21%, and 48% (35), respectively. It seems that especially heavy drinking is a common risk among Finnish physicians; this finding was also confirmed in previous studies (34). Higher rates of alcohol use than in the general population have been reported, for example, among physicians in the United States and Turkey (73,74). However, there are also results supporting the view that physicians are not more likely to abuse alcohol than other professionals (75,76). There can be many factors, such as country, age, gender, and occupational group, that can influence results regarding alcohol use among physicians, and these factors should be better taken into account in future studies. Our present study did not include drugs of any kind, but physicians are thought to be especially vulnerable to prescription drug abuse (77), for example, due to their good access and culture-favoring pharmacological responses to stress (78).
In conclusion, physician shortage is a worldwide problem, and physician turnover is a source of many adverse consequences (1). Our results show that Finnish physicians who have a concurrent high on-call burden and several lifestyle risks seem to be more willing to leave their jobs than other physicians. In addition, a high on-call burden was found to be associated with higher levels of distress. Similarly having many lifestyle risk factors seemed to be harmful for the physicians' mental health and intent to leave their practice. As job satisfaction, lifestyle risk factors, and on-call duties affect physicians' mental well-being and intentions to leave their practice, it would be important for health care organizations to focus some attention to these aspects. For example, promoting coworker support may help, because coworker support is directly related to physician well-being, and the two buffer the negative effects of work demands. In other words, physicians can feel less stressed when they are part of a good team (79). Coworker support might also be of importance in the case of an accumulation of lifestyle risk factors and overburden. In addition, organizations should pay attention to not overloading some physicians with too much on-call burden. Management should pay extra attention to those who have the highest workload and on-call burden and who seem to have several lifestyle risk factors.