Which fatigue measurement tools best predict workplace errors and accidents?

Self-report scales have been tied to real accident records

The most direct evidence comes from a study of 160 tire-industry workers whose shifts all ran longer than 10 hours. They completed the Karolinska Sleepiness Scale, which asks people to rate how sleepy they feel, and the Chalder Fatigue Scale, which asks about physical and mental tiredness. Researchers compared those scores with the workers' accident records from the previous year and with follow-up interviews. Higher fatigue scores were strongly associated with more errors, and the authors concluded that fatigue was a major contributor to human error and workplace accidents in this high-intensity setting [1].

That result is useful because it links a simple questionnaire to something that actually happened — recorded accidents — rather than to a laboratory mistake. But the study is limited to one industry and one country's tire plants, and the abstract does not report how well the scales separated workers who had accidents from those who did not. So the finding supports a correlation, not a proven predictive threshold.

Pairing a fatigue questionnaire with a daily readiness test adds a behavioral signal

A larger study at a Brazilian mining company gave the Chalder Fatigue Scale to 552 employees and split 334 of them into a group showing signs of fatigue and a group without. Both groups then took a computerized work-readiness assessment called the FOCOS/Prontos System, which measures daily performance and cognitive function. The two groups differed significantly in daily readiness, cognitive impacts, and risky behaviors, and those differences showed up before and after workers reported fatigue [2].

The practical value here is the combination: a questionnaire captures how tired a worker feels, while the computerized test captures how they actually perform that day. The authors describe the system as having predictive value for fatigue-related risks and accident prevention, and they argue for continuous daily monitoring rather than one-off surveys [2]. The study was cross-sectional and case-control, meaning it compared groups at a point in time; it did not follow workers forward to see who later had accidents. That makes the predictive claim plausible but not yet confirmed by accident outcomes.

Wearable sensors can quantify fatigue in real time, but accident prediction is still untested

A newer approach uses a smart full-face mask with electronic modules that track 26 physiological, motion, and environmental parameters — things like heart rate, skin temperature, air quality, and carbon dioxide levels. In a small study of 10 healthy participants doing daily activities and controlled mental and physical fatigue tasks, machine learning models reduced those 26 signals to 6 useful ones and combined them with the 20-item Multidimensional Fatigue Inventory (MFI-20) to produce a single Cumulative Fatigue Index [5].

The system correlated better with the MFI-20 than a commercial smartwatch did, and it is designed to support analysis of how individual fatigue relates to work-related incidents and injuries [5]. But the testing involved 10 people over short laboratory-style sessions, not workers on real job sites with recorded accidents. The authors present the wearable as offering added value over survey-only methods, yet the evidence so far establishes measurement capability, not predictive power for workplace errors.

Fatigue models and management reviews point to the same gap

Reviews of fatigue modeling describe two broad approaches: one-step models that estimate fatigue directly from work schedules and other inputs, and two-step models that first derive a fatigue measure and then link it to error or accident risk. These models are used in work settings to manage fatigue-related errors and accidents, but the review literature treats them as tools for risk management rather than as validated accident predictors [3]. A separate review of workplace fatigue management likewise discusses the association between fatigue and performance, errors, and accidents alongside various measurement methods, without identifying a single best instrument [4].

Taken together, the evidence supports a layered approach rather than a single winning tool. Self-report scales are cheap and have been linked to accident records in at least one industry [1]. Adding a daily computerized readiness test appears to sharpen the picture by capturing performance and risky behavior [2]. Wearable multimodal systems promise continuous, objective measurement but have not yet been tested against real accident outcomes [5]. The models and management reviews frame all of these as parts of a risk-management system, not as stand-alone predictors [3][4].

Sources used in this answer

Related research questions