Hospital doctors' working hours and their sleep quality: analysis of actual work schedules and survey responses

The study participants were hospital physicians from four hospital districts participating in the Finnish Public Sector (FPS) study survey in 2015 (N = 11,274, response rate 69%, physicians n = 1018). Physicians were identified from other occupational groups by working time regime coding in registry-based working hour data. Similarly to our previous study [27], physicians’ regular working hour data was retrieved from the working time scheduling program Titania® and the on-call working hour data from Titania for physicians® software (CGI Finland Ltd., Helsinki, Finland). Altogether 728 physicians had combined data from survey and the two working time scheduling software.

The working hour data included realized payroll-based starting and ending times for each work period and absences including days off, sickness absences and other leaves. The working hour characteristics were created from 3 months (91 days) preceding the survey and the data included both regular working hours and all on-call hours at the workplace. The register-data included also starting and ending times for each period of on-call work at home, but this part of data was utilized separately, as these duties are carried out more commonly by senior physicians [6]. The cut-points in the studied working hour characteristics were based on the European Working Time directive [28], the Finnish Institute of Occupational Health (FIOH) recommendations on ergonomic working times [29] and, if no regulatory or research-based cut-points were available, to the distribution of the data utilized.

Average total weekly working hours were asked with a question “On average, how much time do you spend performing your work tasks (normal working time, overtime at workplace and at home and including subsidiary employment)?” and answered as hours per week.

Average 24-hour sleep duration was surveyed with the question “How many hours do you normally sleep during 24 hours?” with options from 5 h or less to 10 h or more in 30-minute intervals [30]. Short sleep was dichotomized as ≤6.5 hours [31]. The insufficient sleep item was derived from the question “Do you sleep enough?” with alternatives “yes, always or nearly always”, “yes, often”, and “seldom or never”. Sleep difficulties (difficulties to fall asleep, waking up several times per night, difficulties in staying asleep, and feeling tired and worn out after waking up after usual amount of sleep; from now on ‘non-restorative sleep‘, for brevity) during the last 4 weeks were enquired with the Jenkins Sleep Scale ranging from “not at all” to “every day” [32]. The six categorized answer options (not at all - 1-3 nights/month, approx. 1 night/wk., 2–4 nights/wk., 5–6 nights/wk., almost every night) were dichotomized as having the sleep difficulties if the frequency was 2–4 times per week or more often.

In addition to age, sex, number of children and hospital district, the analyses were adjusted for control over scheduling of shifts and overall stressfulness of life situation. The first was assessed with the item “How much control do you have over scheduling of work shifts?” [33] with five options from “very much” to “very little”. The latter was used as a 6-point Likert-type scale, from “easy” to “extremely burdensome”.

All the hospital districts gave written permission to FIOH to use the employers´ working time registers for research. All data was pseudonymized, and international ethical standards were conformed. The coordinating ethics committee of the Hospital District of Helsinki and Uusimaa (HUS) have approved the FPS study (HUS 1210/2016).

The statistical analyses were conducted using IBM SPSS Statistics 27 (IBM Corp., Armonk, NY, USA). The associations of the studied working hour characteristics and sleep variables (duration, sufficiency, sleep difficulties) were studied with multinomial regression analysis. After crude analysis, we adjusted the models first with age, gender and hospital district as covariates, followed by models with age, sex, marital status, having children under 18 years, overall stressfulness of life situation, and hospital district as covariates and finally, the fully adjusted models with age, sex, marital status, having children under 18 years, overall stressfulness of life situation, control over scheduling of shifts, and hospital district. As there were missing values in the responses of having children under 18 years, the missing values comprised their own class.

The physicians’ mean age was 43.4 years (SD 9.9, range 26–68 years), and 62% (n = 453) of them were female. The physicians had worked in their current workplace (i.e., hospital district) on average for 9.5 years (SD 8.3, range 0–38 years) and in their current position on average for 5.1 years (SD 5.4, range 0–33 years).

In the 3 months preceding the survey, the average realized weekly working hours were approximately 38 hours per week when realized regular working time and on-call work at workplace were combined. However, in the FPS survey, the physicians had estimated their total weekly working hours at 45.4 h (SD 9.3 h) when also on-call at home and possible second jobs were included. According to the realized working hour data, the physicians had five night duties on average, 2.5 weekend shifts and seven short shift intervals (≤11 h) during the 3 months. The mean length of on-call shifts was 14.5 hours, however only 44 physicians had at least one 24-hour or longer on call-shift. Approximately 45% (n = 369) of the physicians had on-call work at home with a mean of 24.5 hours per week. (Table 1).

One fourth (26%, n = 187) of the participants had short (≤6.5 hours) self-reported sleep duration, male physicians more often than female physicians (30% vs. 23%, p < 0.03). High numbers of night work shifts (≥7 shifts/3 months) and estimated long (> 48 h) weekly total working hours were associated with higher odds for insufficient sleep (OR 1.87, 95%CI 1.23–2.83; OR 2.09, 95%CI 1.25–3.94, respectively), whereas having less than an average amount of on-call work at home was associated with lower odds for insufficient sleep compared to no on-call work at home (OR 0.64, 95%CI 0.41–0.99). (Table 2).

Approximately one third (32%) of the physicians reported insufficient sleep, without gender difference (p = 0.51). Of those who reported having insufficient sleep, 62% had a short self-reported sleep duration (≤6.5 hours, p < 0.001). Realized and estimated long (> 48 h) weekly working hours were both associated with insufficient sleep (OR 1.78 95%CI 1.15–2.76; OR 2.12 95%CI 1.30–3.45, respectively). The other working hour characteristics with higher odds for insufficient sleep were high (≥13/3 months) number of on-call shifts (OR 2.00 95%CI 1.08–3.72), frequent (≥7/3 months) night work (OR 1.60 95%CI 1.09–2.37), and high number (> 12/3 months) of max 11-hour shift-intervals (OR 1.65 95%CI 1.01–2.69). (Table 3).

Nearly half of the physicians (48%, n = 347) experienced often at least one sleep difficulty, female physicians more often than males (52% vs. 43%, p < 0.02). Of those who reported insufficient sleep, 84% reported having also at least one sleep difficulty (p < 0.001). However, none of the studied working hour characteristics were associated with having at least one sleep difficulty (Table 4). However, when the odds for each sleep difficulty with regards to the working hour characteristics were studied separately, frequent night work increased odds for difficulties in initiating sleep (OR 2.43 95%CI 1.04–5.69) (Table 5).

In the crude models having frequent (≥5/3 months) weekend work increased odds for waking up several times per night (OR 2.40 95%CI 1.06–5.45) (Table 6) and having frequent (≥7/3 months) night work increased odds for having difficulties in staying asleep (OR 1.64 95%CI 1.00–2.68) (Table 7). There were no associations between the studied working hour characteristics and non-restorative sleep (Table 8).

The mean realized weekly working time of the hospital physicians was approximately 38 hours per week. The main challenge in comparison with earlier research is that previously the working hours have been self-reported, and it is known that self-reported hours may deviate from actual hours worked [3]. In our study, the total estimated weekly working hours were over 45 hours when also on-call work at home and possible second jobs were included in the estimations. It is likely that when answering a survey, a respondent thinks of a typical work week and does not calculate reductions for sickness or other absences from work. However, the self-reported weekly hours in this study were at the same level than previous national [34] or Nordic figs [3] and somewhat below European comparative estimated figures around 50 hours per week [1]. Compared to previous survey studies, the mean frequency of on-call shifts [6, 8, 9, 35, 36] and short shift intervals [37] were somewhat lower in this payroll data. On-call work at home was more common among physicians included in our study than in a previous study [38].

In this study, approximately one fourth of the physicians reported short (≤6.5 h) habitual sleep duration. In previous studies, physicians’ average sleep durations have been at least 7 hours [12], but not when having had night-time duties [39]. Of the studied working time characteristics, only frequent night work was associated with short sleep. Previous studies [37, 40] on physicians’ night-time on-call duties have shown parallel results. For example, the connection between reduced sleep and consecutive [37, 39] or otherwise frequent [40] night duties as well as long working times [9]. Contradictory to a recent review [23], having on-call work at home for less than 20 hours per week was associated with lower odds for short sleep compared to those with no on-call work at home at all. As on-call work at home often takes place during night-time, we conducted an additional analysis with the amount of night work as a covariate, but this did not change the results. We speculate that that result may be related to the different seniority position of those having a limited amount of on-call work at home or could derive from less frequent calls to doctors than in an earlier study [10].

Overall, we observed only a few associations between working hour characteristics and sleep duration. It has been reported that although physicians´ longer work hours are associated with shorter sleep, very short sleep (< 5 h) is associated with different amounts of work [25]. However, previous studies have suggested that physicians prioritize other activities than sleep during their free time also when working fewer hours [4, 9].

Working hour characteristics which were associated with perceived insufficient sleep included both realized and estimated long weekly working hours and a high number of on-call shifts, night work, and over 12 short shift intervals (< 11 h) during the preceding 3 months. The association was strongest between a high number of on-call duties and insufficient sleep. There are few previous studies that would have studied physicians’ working hour characteristics and sufficiency of sleep, but previous results confirm that insufficient/non-restorative sleep is commonly reported by physicians [8, 41] or physicians working both night-time and overtime duties [42]. However, straightforward conclusions regarding physicians’ preceding insufficient sleep and negative patient outcomes should not be drawn [41].

Nearly half of the physicians (48%) in this study reported having had sleep difficulties, which is a slightly higher [43, 44] or similar [42] proportion than in previous studies. Higher prevalence among female physicians has also been previously reported [43]. However, none of the studied working hour characteristics were associated with sleep difficulties. When the sleep difficulties were analysed separately, frequent night work increased odds for difficulties in initiating sleep, which is in line with previous studies on longer sleep latency in connection with night duties [42]. It can be speculated whether physicians pay enough attention to their sleep quality, as there are results showing that physicians may not use recommended good sleeping strategies, such as engaging in regular physical activity or seeking sleep specialist advice to improve sleep [42].

This study used physicians´ register-based working hour data to calculate accurate working hour characteristics for both regular and on-call hours. The use of register-based working hour data is the main strength of this study. Previous research has been conducted for the most part using self-reported working hour data which is prone to recall bias [26]. However, comparison to previous studies is limited due to methodological differences in working hour data collection (payroll data vs. survey data on working hours) and the sleep scale utilized. Another strength of this study was the relatively large sample covering both university and regional hospitals, and the comprehensive survey data that enabled multiple adjustments in the statistical models. Previous research has also often included junior doctors only [14, 25, 37, 45], whereas this data included physicians in different career stages.

There are limitations to be noted also. The working hour data was collected from all the physicians included in the payroll data in the studied four hospital districts, but the survey response was available for only less than 40% of these physicians. The physicians´ mean age and gender distribution were similar to the realized working hour data of all hospital physicians, but other sources of selection cannot be ruled out. On the other hand, 67% of the physicians with survey data had complete working hour data and were thus included to the final sample. The physicians without working hour data were for example chief physicians in administrative positions. The larger amount of on-call work at home in this study may indicate that these respondents had more senior positions than physicians in some of the previous studies [38]. There was a discrepancy between realized and self-estimated weekly working hours, which is likely being explained by the respondents´ tendency of not to subtract sickness or other absences from work from self-estimated average hours. Moreover, due to questionnaire wording, it is likely that also on-call work at home and work hours in possible second jobs were included.

Importantly, the employers’ register data did not allow us to differentiate between different medical specialties, and there may be differences in the amount and arrangement of on-call schedules. We used self-reported sleep data, which is subject to recall bias. The categorical variables for the outcomes limited the statistical analysis and the multinomial analysis we used may lack sensitivity to distinguish small differences. As there is variation in both legislation limiting, for example, the physicians’ maximum number of weekly working hours, and organization of health care services in different countries [46] these results are likely best generalizable to other European Union member states that comply with the European Working Time Directive [28].

As on-call duties are concentrated on certain medical specialities, such as anaesthesiologists [12, 17] and surgeons [6], it would be justified to study further on-call working hour characteristics and sleep among these specialists. Similarly to other occupations that require working in irregular shifts, studying sleep quality directly in connection with different shift or on-call characteristics would be merited. Extending precise data collection on active working hours and rest opportunities while on-call would be of great value as well.

The coordinating ethics committee of the Hospital District of Helsinki and Uusimaa (HUS) have approved this study as part of the FPS study ethical approval (HUS 1210/2016). All data was pseudonymized, and international ethical standards were conformed. Answering to the FPS survey was totally voluntary and the completed questionnaire acted as an informed consent. In accordance with national Data Protection Act (1050/2018, Section 4) the employer-owned working-hour register data processing for research of public interest did not require informed consent from the individual employees.

Not applicable.

The authors declare that they have no competing interests.

The research data are not publicly available due to privacy and ethical restrictions. Research data are not available on request from the authors or FIOH due to written agreements with the register data owners (the hospital districts) not to forward any data to third parties.