Long work hours and shift work increase the risk of both physiological and psychological distress [1-6], including cardiovascular disease [7] and depressive symptoms [8]. Medical staff, especially trainee physicians, work for a large number of hours per week and have a heavy workload. Systematic reviews have reported that such long working shifts and erratic schedules result in numerous adverse consequences in patient care [9]. In 2003, the Accreditation Council for Graduate Medical Education implemented the first work-hour restriction policy for all physicians in training in the United States, limiting resident physicians’ workweeks to 80 hours and shifts to 30 hours [10]. In Taiwan, similar policies by the Ministry of Health and Welfare limited workweeks to 88 hours in 2013 [11]. However, surveying medical interns’ compliance with the 2003 work-hour limits in the United States using a traditional assessment took 2 years; the resulting national survey was published in 2006 [10]. In addition, these self-reports do not reflect the fluctuations of work hours in real time, especially for medical staff with frequent on-call duties.

Nowadays, smartphones offer us an objective and ecological source of measurement that continuously and passively collects data [12]. These reliable, quantitative data could facilitate real-time policy evaluation and target resources to those with the greatest need for them, even in remote and inaccessible regions. As of July 2019, the Google Play Store and iPhone Operating System (iOS) App Store had about 500 mobile apps dealing with the tracking of work hours [13]. However, most of those apps require manual entries or active clocking in and out. Fewer than 20 of those apps are capable of automatic timekeeping using various technologies such as GPS geofencing or Wi-Fi detections. Furthermore, medical personnel’s work hours could not be identified solely by GPS geofencing technology. Physicians often have several on-call duties every month. During their on-call duty, physicians have to be physically present at the hospital or be on-call at home. In addition, some medical staff practice in more than 1 institution, and these patterns of work hours could not be identified by a simple GPS geofencing algorithm.

We herein report the design of the app Staff Hours, which automatically calculates users’ work hours through GPS data. Automatic recording of exact working hours is highly beneficial for people with erratic work schedules. This GPS-based work-hours recorder would not need users to precisely log their work hours by manual input or clocking in and out. The specific aims of this study were to (1) identify the app users’ GPS-defined work hours, (2) examine the overtime work hours from the app-recorded total work hours and the participants’ self-reported scheduled work hours, and (3) compare these app-recorded total work hours among different occupations.

We collected data on 183 office workers from July 30, 2018, to August 25, 2019, using the Staff Hours database. All participants were volunteers interested in their work hours, and workers who lived within a 1-km radius from their workplaces were excluded from this study. Of these participants, 123 were males and 60 were females, and their mean age was 32.98 (SD 6.74) years. Most of the participants (162/183, 88.5%) were medical staff, and their positions were resident physicians (n=89), visiting staff (n=38), medical students (n=10), registered nurses (n=25), and non–health care professionals (non-HCPs; n=21). All clinical investigations were conducted according to the principles expressed in the Declaration of Helsinki.

Staff Hours is a mobile app designed to record work hours automatically using the technology of geofencing. As medical staff, along with most office workers, work at fixed locations, geolocation data are valid indicators of whether one is at work. On the basis of this principle, we developed algorithms to log the exact work hours from GPS data; these hours are called GPS-defined work hours. Furthermore, this app can perform overtime calculations by comparing the regular work schedule defined by the user and the corresponding GPS results.

Staff Hours, which was designed and developed by our team, is now available exclusively in Taiwan through the iOS App Store and the Google Play Store. The app runs in the background and records GPS information every 10 min with low power consumption. For iOS users, the app should be kept in use to function correctly, either in the foreground or background.

We recruited 5 medical doctors and interviewed them regarding their work hours for the past 7 days to examine the accuracy of these (n=5×7=35) app-recorded total work hours by Staff Hours. This structural interview detailed the users’ on-call duties, regular work hours, and overtime work hours for each day. These self-reported total work hours excluded the trainee physicians’ educational activities. Therefore, we used these self-reported total work hours as the putative gold standard to validate the app-recorded total work hours. The sensitivity was 94.6% (242.7/256.6) and the specificity was 93.9% (547.6/583.4) for the app-recorded total work hours. There was a high correlation between the app-recorded total work hours and the self-reported counterparts (r=0.923; P<.001).

A paired t test was used to compare the differences between the total work hours and the scheduled work hours. One-way analysis of variance was used to compare the differences in total work hours among all 5 categories of workers (ie, medical students, visiting staff, resident physicians, registered nurses, and non-HCPs). Fisher least significant difference (LSD) was used for further post hoc tests. In addition, we used independent t tests to examine the differences in the total work hours between the medical staff (ie, medical students, visiting staff, resident physicians, and registered nurses) and non-HCPs. All statistical analyses were 2-tailed, and P<.05 was considered to be statistically significant. Data arrangement and statistical analysis were performed using R software (version 3.5.3).

The app-recorded total work hours (mean 55.69 hours, SD 21.34) of all the 183 participants were significantly higher than the self-reported scheduled work hours (mean 50.67 hours, SD 21.44; t ₁₈₂=2.61; P=.01). All participants showed an average of 4.78 hours (SD 26.12) overtime work per week (ie, total work hours−scheduled work hours). Table 1 demonstrates that resident physicians had a significant positive overtime (mean 7.95 hours, SD 24.25), that is, the residents’ total work hours were significantly higher than the scheduled work hours (t ₈₈=3.14; P=.002).

Table 1 also shows the comparisons of total work hours and scheduled work hours among the 5 categories. There were significant differences among the 5 categories on total work hours (F_4,183=3.047; P=.02). Post hoc comparisons using LSD tests further revealed that the resident physicians had significantly longer total work hours (mean 60.38 hours, SD 18.67) than the visiting staff (mean 51.42 hours, SD 20.33; P=.03) and non-HCPs (mean 45.48 hours, SD 20.08; P=.004). In addition, medical staff (ie, resident physicians, registered nurses, visiting staff, and medical students) had significantly longer total work hours (mean 57.01 hours, SD 21.20) than non-HCPs (mean 45.48 hours, SD 20.08; P=.02).

Our findings showed that medical staff had longer work hours than non-HCPs, and resident physicians worked for longer hours than visiting staff in hospitals. These findings are consistent with previous reports that showed that resident physicians typically work the greatest number of hours per week among all positions in hospitals [14]. Similarly, it is feasible to compare work hours among different hospitals, departments, or divisions using aggregate data collected from the Staff Hours’ database. In this way, we can find out the hot spots where office workers have the longest total work hours or overtime. From September 2019, the Labor Standards Act in Taiwan will include regulations on the working hours for resident physicians. With the implementation of government policy, Staff Hours can serve as a real-time monitor for compliance with these work-hour limits. The app might also be useful in labor negotiations, balancing the unequal information between employers and employees. Although employers hold official records on work hours, the automatically recorded data in mobile apps provide evidence for employees when they claim overtime pay. In this study, each occupational group comprised greater than or equal to 10 people, which strengthens user privacy. Further analysis of the comparisons among different clusters should take individual privacy into account and make sure that there are not less than 10 people in each group.

The algorithms behind Staff Hours include not only the location-based work hours recording but also several critical designs for a better user experience. First, detached working periods are linked back together in the GPS-defined work hours because of the unique duty-based characteristics of health care work. Take a physician at an academic hospital as an example. She may have to leave the hospital for lectures on the university campus or walk a couple of blocks in the neighborhood to grab lunch in the middle of the scheduled working time. If the physician comes back to the hospital in less than 8 hours, the app still records the period that she is absent from the GPS range of the hospital as GPS-defined work hours. Second, the update interval of GPS location data in this app is set to be 10 min to improve battery performances; this is much longer than the geofencing responsiveness interval of approximately 2 min in the background location limits introduced in Android 8.0 (Application Program Interface level 26) to reduce undesired battery drain [15]. Third, users can adjust the work-hour logs manually only within 7 days. This restriction was established to avoid recall bias, increasing the accuracy of records [16]. Another associated advantage is that it may assure data integrity and prevent future problems of data fraud, which would be crucial in labor inspections.

In addition, this app performs functions that specifically target medical staff, such as automatic calculations of overtime and on-call duty hours. Our findings demonstrated that all the participants on average worked 4.78 more hours per week than scheduled. The lengthy overtime may be from a selection bias in our participants, most of whom (89/183, 48.6%) were resident physicians whose total work hours were always higher than they should be. However, we observed that the scheduled ones were shorter than expected because few residents (58/89, 65%) logged their on-call dates in the system, which might result in an underestimation of the scheduled work hours. Moreover, this app could not differentiate the activities of working and learning in the institution for residents and medical students, which might result in an overestimation of total work hours. Together, they may lead to an overestimation of overtime and thus, we observed significant differences between the total and scheduled work hours in this study. Unsure of the validity of the scheduled and overtime work-hours data, we analyzed only the total work hours of the various occupations.

Several methodological limitations should be noted when interpreting our findings. First, users who live within a 1-km geofencing radius from their workplaces are limited to using the algorithm of GPS-defined work hours. Furthermore, inaccuracies in GPS location tracking may occur when users work at smaller companies or clinics as the app transforms workplace addresses into latitudes and longitudes. Second, if only a few medical staff log their on-call duties in the app, the produced average overtime hours could be higher than the actual overtime hours. Third, the app could not identify breaks within the geofencing range, which might result in an overestimation of total work hours. Finally, there may be a selection bias in the participants as medical staff or other users with longer work hours may have a higher tendency to install Staff Hours.

Staff Hours is the first app designed specifically for medical staff, with built-in functions for tracking on-call duties in hospitals. To the best of our knowledge, no prior study has enabled medical staff to use their personal smartphones, either Apple or Android models, to monitor their work hours automatically. The high accuracy of GPS-defined work-hours data without manual input and the integration between multiple work locations show the app’s ease of use. Moreover, Staff Hours is the first app to perform comparisons on work hours among different categories of medical staff and also between medical staff and other occupations. For staff having excessive work hours, this app could serve as a smoke alarm, providing early signals of overwork or occupational burnout. We designed this app focusing on medical staff’s work-hour patterns as they may have the most complicated work-hour patterns, including frequent on-call duties and multiple workplaces. The power-saving GPS data collection with optimal sampling rate and work periods integration in our algorithm overcame these challenges.

Staff Hours is the first automatic GPS location–based app designed for medical staff to track work hours and calculate overtime. For medical staff, this app could keep complete and accurate records of work hours in real time, reduce bias, and allow for better complying with labor regulations.