This study comprises two phases: (1) development of an integrated fatigue measurement system for construction workers using both subjective and objective real-time data and (2) evaluation of feasibility and usability based on the feedback of construction workers in the field.
Phase 1: development of an integrated fatigue measurement system
This measurement system includes a smartwatch (Galaxy Watch Active 2; Samsung Electronics Co., Ltd., Republic of Korea), an application for sensor data collection, LASoR (LK2 Consulting, Republic of Korea), and the developed EMA application as well as a smartphone (Galaxy S7 or later released; Samsung Electronics Co., Ltd., Republic of Korea). Our interdisciplinary research team developed the integrated fatigue measure using both wearable devices and the EMA methodology. The interdisciplinary research team comprised construction management, nursing, data science, and IT experts for sensor data collection. We conducted a preliminary survey on 108 construction workers and selected a smartwatch as the wearable device. We also consulted user interface designers who confirmed that the user interface design of the EMA application was better supplemented by inserting emoticons, such as by applying a color scale. The collected EMA data were stored in local storage. In addition, the smartwatch sent the collected sensor data to the smartphone via Bluetooth in real time, and the smartphone sent the data to the cloud server via Wi-Fi every hour (Fig. 1).
Smartwatch and smartphone
The main hub of our system is a smartwatch (Galaxy Watch Active 2; Samsung Electronics Co., Ltd., Republic of Korea) and smartphone (Galaxy S7 or later released; Samsung Electronics Co., Ltd., Republic of Korea). The product dimensions of the smartwatch are 1.6 × 1.6 × 0.41 in., and its weight is 26 g. It has an internal storage of 1.4 GB and a battery life of up to 95 h per charge. This smartwatch was selected due to its several advantages. Being based on the Korean language, it has an easy-to-understand interface compared with other products. In addition, it is also more user friendly because it provides various applications and functions that are well-integrated in the Android OS that is mostly used by Korean construction workers. It also uses Wi-Fi and is efficient at sending data.
Application for sensor data collection, LASoR
This smartwatch consists of several passive sensors including an accelerometer, gyro sensor, heart rate sensor, light sensor, and global positioning system. It can collect various biometric signals, including physical activity, sleep patterns, and psychological distress. The biometric data collected through sensors embedded in the smartwatch include heart rate, three-axis accelerometer, and three-axis gyroscope data. Heart rate is the most widely used form of physiological information for personal health status [2, 33]. Photoplethysmography (PPG) sensors are used to measure heart rate. Accelerometer and gyroscope data are collected to evaluate the amount of activity of workers. In previous studies, the feasibility of recognizing the activity of workers was verified using smartwatch acceleration data, without interfering with their ongoing work [34, 35]. The three-axis acceleration is the acceleration force data along each axis (x, y, and z axes) collected from the accelerometer, and the three-axis gyroscope is the rotational speed around each axis. These data can be used to detect motion and measure the amount of activity. The three-axis accelerometer, which measures inertial body motions, provides information-rich data regarding the workers’ activities, without considerable additional computational expenses [36].
EMA application
We developed the EMA application to load on the smartwatch. Our interdisciplinary research team consulted construction site managers and occupational nurses to detect symptoms of fatigue in a timely manner without interfering with the daily lives of the construction workers. We have developed applications for smartwatches that can respond to questions, considering that it is difficult for construction workers to use their smartphones to respond while working. Because construction workers wear protective gloves during work, the application was developed such that they can respond even when wearing gloves by pressing a button on the smartwatch.
The EMA app collected two types of self-reported data regarding the worker’s fatigue. First, the worker reports the overall fatigue levels after receiving an hourly alarm prompt (hereinafter referred to as EMA (type 1)). The worker answers the second EMA question regarding the participant’s fatigue symptoms four times a day when (a) starting the work, (b) taking regular breaks, and (c) finishing the work (hereinafter referred to as EMA (type 2)). Five specific questions were extracted from the Korean version of the Swedish occupational fatigue inventory (K-SOFI) [16] to evaluate momentary occupational fatigue among construction workers. In this study, the fatigue level was measured on a 6-point Likert scale (0 = “not at all”, 5 = “severe fatigue”). Originally, the SOFI was on a 7-point Likert scale (0 = “not at all”, 6 = “very high level”) [9]; however, we revised it to a 6-point Likert scale based on the findings from interviews with construction site managers and several construction workers that high levels of fatigue were not frequent [16].
Phase 2: validation study to evaluate feasibility
In the second phase, we collected the data from 100 construction workers in the field and evaluated data acquisition, compliance with EMA, feasibility, acceptability, and usability. Feasibility of this developed system was assessed in line with recommendations of previous studies [18, 37, 38] using the following metrics: (a) sensor data acquisition rates as an objective fatigue measurement, (b) rates of EMA compliance as subjective fatigue measurement, and (c) self-reported acceptability and usability of smartwatch-based EMA. All study participants provided informed consent, and the study design was approved by the Institutional Review Board of the affiliated university (IRB No. XXX-2019-11-001 for anonymous review). Researchers explained the purpose, protocol, and strategies to the construction site manager working on site to protect personal information.
Study participants
Because this feasibility study measured fatigue by using EMA and collecting physiological data of construction workers, the sample size was determined by referring to previous studies. First, according to studies related to physiological data collection of construction workers, the advantage of sensor data measurement is that it can be analyzed with relatively few subjects (10–25 people) compared to self-report studies such as surveys [2, 26, 27, 29]. Second, in previous studies on the EMA of fatigue, approximately 40–80 participants were analyzed [23, 39, 40]. The number of subjects to be used for analysis was chosen to be 80, referring to previous studies, and the sample size was calculated as 100 corresponding to the expected dropout rate of 20%.
A sample of 100 Korean construction workers was enrolled from five construction sites. Participants were recruited via an announcement posted on the bulletin boards or via word-of-mouth, and the construction site managers assisted recruiting the participants. The inclusion criteria were as follows: (1) age ≥ 19 years, (2) the ability to use a smartwatch and smartphone, and (3) the ability to understand the EMA instructions. The exclusion criteria were as follows: (1) non-Korean workers and (2) workers using a smartphone other than an Android smartphone. After checking the data completeness, seven participants were excluded from the analyses due to data for classifying fatigue groups. Of the 93 study participants, 80 completed the entire protocol: (a) five dropped out during data collection due to device connection problems, (b) six dropped out due to work schedule changes, and (c) two dropped out by accidentally uninstalling the EMA app during the study. The final data of 80 participants were included for the data analyses (Fig. 2).
Measures
Study participants completed the baseline questionnaires on the socio-demographic, work-related, and health-related characteristics. Socio-demographic characteristics included age, sex, marital status, education, and living status. Work-related characteristics include work experience, working hours of a day, employment form, and working intensity. Health-related characteristics include height, weight, smoking, drinking, and exercise.
K-SOFI was used to assess fatigue levels at the baseline for classification into the high and low fatigue groups [16]. The SOFI is internationally used to measure self-reported fatigue [9]. It comprises measures for lack of energy, physical exertion, physical discomfort, lack of motivation, and sleepiness (range: 0–120) [13]. The Korean version of the SOFI has demonstrated suitable reliability (Cronbach alpha ranging from .70 to .90) and has been tested on construction workers [16]. Based on the mean value of the K-SOFI of the participants, 1.56 (SD = 1.28), the high fatigue (higher than 1.56) and the low fatigue groups (lower than 1.56) were classified through a group comparison, based on the data acquisition rate or compliance with EMA.
Data collection procedure
The data were collected using standardized self-report questionnaires, K-SOFI, and the smartwatches with the EMA application were installed between July and November 2020. Our strategy was to evaluate construction workers’ fatigue over three working days using the developed system. The research team installed an application for data transfer on their smartphones. Trained research assistants explained the purpose of the self-reported EMA and smartwatches. In order to facilitate accurate self-reports by the participants and maintain strong inter-rater and intra-rater reliability, we provided video supplements to explain the approach to EMAs and the specific fatigue symptoms depending on levels. Whenever needed, trained research assistants additionally taught the participants to operate the smartwatches. Participants were expected to wear the smartwatch at all times, except during charging and bathing. To manage battery limitations, researchers instructed the participants to charge their smartwatches daily. To incentivize involvement, researchers encouraged the completion of all measures and indicated the remote monitoring of compliance; additionally, participants received a reward worth US $100 during the study.