A quantitative, online survey of physicians and dentists working in Hungary was conducted between 9 May and 15 July, 2013. The survey was performed with the permission of the Hungarian Medical Chamber. Ethical permission was obtained from the Ethical Committee of Semmelweis University, Budapest (No: 60/2013).
A link to an anonymous self-administered questionnaire was sent to all potential participants, namely to the registered members of the Hungarian Medical Chamber with valid e-mail address (n = 42.342). This was followed by four reminder e-mails. Response rate was 16.18 % (n = 5.607). We considered a questionnaire complete in case the participant responded to at least 90 % of the questions. All international studies emphasize that surveys of physicians have lower response rates even in case of traditional, paper-based surveys in comparison with the general population [31, 32]. The response rates among physicians average about 10 % lower than studies with the general population. Studies on physician surveys have shown that lower response rates were the result of using web surveys alone compared to other survey modes [32].
The data were weighted by gender, age, and type of profession (physicians vs. dentists), according to the data on members of the Hungarian Medical Chamber. After three-dimensional weighting (gender, age, and type of profession), the distribution of data concerning the region and type of workplace were compared with the same type of data from the Hungarian Central Statistical Offices [33]. As a result of the comparison, we found no distributional differences by regions (counties) or by type of workplace (general practice, in-patient and out-patient care). Therefore, we considered our survey material representative. Due to the special nature of their profession, dentists were examined as a separate group. In the present analysis, we limited our attention to female medical doctors (n = 2414).
Among mental health indicators, we focused on depressive symptoms, suicidal thoughts and attempts, and different dimensions of sleep disorders. In the analysis, the following work-related variables were used: number of workplaces, working hours per day, changing work schedule (night shifts), satisfaction with work, and frequency of work-related stressful events. In addition, we analyzed indicators of role conflict and burnout.
The first representative study on Hungarian female physicians was conducted in 2003 in which we used random starting point systematic sampling among the registered members of the Hungarian Medical Chamber. Two thirds (62.7 %, n = 408) of the questionnaires completed by the sample chosen were then suitable for statistical analysis. This sample was weighted according to the age and type of workplace dimensions provided by the Hungarian Medical Chamber and the Hungarian Central Statistical Offices. Thus, our sample was considered representative with regard to age and type of workplace, which are the most important dimensions from the perspective of this survey.
Control group
With regard to the questions related to mental health (depression, suicidal thoughts and attempts), we used control groups from the population-based surveys Hungarostudy 2002 (n = 12.640) and Hungarostudy 2013 (n = 2.000). The objectives of these representative surveys were to describe the physical and mental health condition of the Hungarian population and to investigate the association between health indicators and certain environmental, social, and economic background factors. Altogether, 12.640 and 2.000 respondents, respectively, were interviewed about their socio-demographic data, health condition, behaviors and habits, and psychological characteristics. Participants of both surveys represented the adult population in Hungary according to gender, age, and place of residence. Each questionnaire consisted of approximately 200 items, and answers were obtained via interviews and self-administered questionnaires. For the comparison with female physicians, the data of 818 and 146 professional women were used, respectively [34, 35]. Professional women mean women who graduated from college or university. The professional control groups were representative in both time-points; thus, we compared two samples that were adjusted for gender, educational level, and age.
Measurement of mental disorders
The following tools were used for the assessment of mental health factors:
Sleep disorders were assessed by the shortened Hungarian version of the Athens Insomnia Scale (AIS) [36, 37]. The original scale consists of 8 items, e.g., such as “Have you had problems with your sleep during the last month that occurred at least three times per week?” or “Do you have difficulties falling asleep?” or “Do you wake up in the middle of the night?” or “Do you wake up too early in the morning?”; total sleep duration (0: no problem, 1: minor problem, 2: considerable problem, 3: serious problem). The cut-off score of 10 in epidemiological surveys is adequate. In our sample, the Cronbach alpha value of the Hungarian version of AIS is 0.874.
We assessed depression by the shortened version of Beck Depression Inventory (BDI) [38]. Converting the total score of the shortened Beck Depression Inventory to the original BDI, four grades of depression can be distinguished (total score/9 × 21): 0–9: indicates no depressive symptoms; 10–18: mild depressive symptoms; 19–26: moderate depressive symptoms; above 26: severe depressive symptoms. In our study, the Cronbach alpha value of BDI was 0.86.
In the survey, we used the modified question of Paykel’s “Suicidal ideation and suicide attempts” questionnaire [39] to assess suicidal thoughts: “Have you ever been preoccupied with suicidal thoughts?” The answers (1. no suicidal thoughts, 2. suicidal thoughts in the past year, 3. suicidal thoughts in the past 5 years, 4. suicidal thoughts for more than 5 years) were dichotomized as follows: 1. never have had suicidal thoughts, 2. have had suicidal thoughts.
To assess suicide attempts, we used the following question: “Have you ever attempted suicide?” Answers (1. never, 2. suicide attempt in the past year, 3. suicide attempt in the past five years, 4. suicide attempt for more than five years) were dichotomized as follows: 1. have never attempted suicide, 2. have attempted suicide.
Burnout was measured using the Hungarian version of Maslach Burnout Inventory-Human Services Survey (MBI) [40, 41]. This questionnaire of 22 items has three subscales for the measurement of the three different dimensions of burnout: emotional exhaustion subscale (EE), depersonalization subscale (DP), and personal accomplishment subscale (PA). EE consists of nine items and assesses the feelings of being overextended and exhausted in one’s workplace (e.g., “I feel depressed at work”). DP has five items and measures an unfeeling and impersonal response toward recipients of one’s service, care treatment, or instruction (e.g., “I don’t really care what happens to some recipients”). PA consists of eight questions assessing feelings of competence and success in one’s work (e.g., “I have accomplished many worthwhile things in this job”).
The items were rated on a seven-point Likert scale according to the prevalence of certain work-related feelings (from 0 = “never” to 6 = “every day”). By the trisection of the total score reached on each subscale, three levels of burnout could be detected (1. mild, 2. moderate, 3. severe). Among physicians, severe burnout was characterized by the following cut-off scores: EE >27, DP >10, and PA <33. In the present analysis, Cronbach alpha score of the EE, DP, and PA was 0.909, 0.767, and 0.818, respectively.
One questions represented a role conflict [42, 43]: “How often do you feel irritated or dissatisfied because of the impression that you cannot balance between your workplace, family, household, or partnership engagements?”, which the respondents answered on a five-point Likert scale (1. never, 2. rarely, 3. sometimes, 4. often, 5. very often). This variable was dichotomized in certain parts of the analysis: 1. never, rarely or sometimes experience role conflict, 2. often or very often experience role conflict.
We aimed to examine the frequency of work-related stressful events by the question: “How often do you experience stressful situations during work?” This was answered by the respondents on a five-point Likert scale (1. not typical at all, 5. absolutely typical). As the result of dichotomization, two categories emerged: 1. not typical at all, 2. typical/absolutely typical.
The degree of subjective workload was assessed by the following question: “How often do you feel overloaded?” which was answered by the respondents on a five-point Likert scale (1. not typical at all, 5. absolutely typical). As the result of dichotomization, two categories emerged: 1. not typical at all, 2. typical/absolutely typical.
We wanted to detect changing work schedule, with primary focus on shifts by the questions: “Do you work in shifts? How is it scheduled?” The possible answers were dichotomized into two values: 1. do not work in shifts, 2. work in shifts. In the second case, we queried the number of hours being on duty.
The number of workplaces was grouped as follows: 1. only one workplace, 2. two or more workplaces. In case of weekly working hours, we counted the average daily working hours and dichotomized the results: 1. less than 8 h, 2. more than 8 h.
Statistical analysis
During the descriptive analysis, we calculated frequency, mean values, and standard deviation. Deviations of percentages are indicated also. Depending on the type of variables, independent samples t test or chi-squared test was applied. Binary logistic regression analysis was performed to assess the correlation between work-related background factors and the items that became more prevalent among female physicians over a decade (sleep disorders, decreased personal accomplishment). Dependent variables were dichotomized variables of AIS and PA (moderate and severe). Independent variables were age, number of workplaces, amount of working hours, night shifts, work-related satisfaction, work-related stressful situations, and indicators of role conflict.
In the present analysis, we refer to the ratio of valid answers. For the statistical analysis, we used SPSS 15.0.