This study was designed as a crossover trial, which took place between October 2021 and March 2022. The experiments took place in climate chamber (5 × 3 × 2.2 m (L/W/H)) with temperature and humidity control. The interior included a table with a chair, a treadmill, a 170 patient bed and patient dummy (CLA1®, 21 kg, Coburger Lehrmittelanstalt, Coburg, Germany). A photo of the climate chamber setting can be found in the supplemental information (Fig. S1).

Four independent experiments were performed by each the participants: (1) at 22.0 °C without PPE (NN), at 22 °C with PPE (NP), (3) at 27 °C without PPE (WN) and (4) at 27 °C with PPE (WP). The selection of the higher temperature was based on the guidelines of the German Federal Ministry of Labour and Social Affairs, recommending that the ambient temperature in workspaces should not exceed 26 °C (BMAS 2022). However, we chose 27 °C in order to achieve a sufficient temperature difference to the reference temperature of 22 °C. The relative humidity was set to 40% for all four scenarios to limit the number of experimental variables. Participant wore a standard hospital gown and additional PPE (disposable plastic gown, FFP2 face mask, face shield and gloves) if applicable. To alleviate a habituation or training effect, the setting´s sequence was randomized. During the experiment participants performed HCW-related activities within a given time. These activities include, amongst others, walking up to 5.8 km/h on a treadmill, mobilization and washing of a dummy and the simulation of administrative tasks. The full protocol can be found in the supplementary information. Except for temperature and PPE, all other conditions were identical. One experiment lasted 3.5 h. Each participant conducted all experiments at the same time of the day (either mornings or afternoons) and the interval between two individual experiments was at least seven but maximum ten days. These prerequisites aimed to minimize interfering effects of circadian rhythms and differences in the participants´ physiological condition (Goel et al. 2013).

Inclusion of eligible participants was based on the medical history, by considering the following criteria: (1) age 18–60 years; (2) medical background or experience as HCW, since they had to perform several HCW-related activities; (3) no sensitivity against heat (e.g., dizziness, redness on the skin); (4) no obesity (BMI < 30); and (5) no severe chronic diseases. Prior to each experiment, participants were asked about their general state of health. Furthermore, heart rate, blood pressure and body temperature (forehead) were measured to exclude illnesses that may interfere with the study.

A QUESTemp 34 Heat Stress Monitor® (Quest Technologies, Wisconsin, USA) was utilized for assessing the heat stress by the environmental conditions in the climate chamber. The instrument was placed on a table at a height of approximately one meter. It was also ensured that this device was placed away from any barriers that might block radiant heat or flow. The participants were also requested not to move too close to this instrument in order to minimize variations in temperature and radiant heat. The sampling interval was one minute.

A cosinuss° Two® (Cosinuss GmbH, Munich, Germany) in-ear sensor was utilized to monitor participants´ in-ear temperature (IET) and heart rate (HR) during the experiments (Fig. 1). For each participant, the appropriate sensor size (small or medium) was selected during the anamnesis. The sampling interval was one second. After 14 min of recording, the data were sent to the cosinuss° Health cloud server via a Gateway within one minute. This cycle was repeated during the whole experiment.

Wearable sensors used during the trials. Cosinuss° Two in-ear sensor® (A) to monitor body temperature and heart rate and Thermochron iButton® (B) to record peripheral temperatures

Moreover, skin temperatures were monitored using Thermochron iButton® temperature loggers (CK electronic GmbH, Cologne, Germany) (Fig. 1). The sensors were placed at five different central and peripheral locations (left/right infraclavicular, belly and left/right midthigh). The sampling interval was one minute. Finally, to evaluate the participants´ clinical state, their weight, forehead temperature, blood pressure, and heart rate were measured before and after each experiment using a digital body scale, an infrared thermometer and a medical blood pressure cuff, respectively.

Prior to the analysis, all data were pre-processed. In detail, sections outside the trials, including those with apparent sensor malfunctions, were removed and the remaining measured data were used for the analysis. Furthermore, only physiological data measured under the dry bulb indoor temperature between 20 and 24 °C for normal conditions and between 25 and 29 °C for warm conditions were included in the analysis. For in-ear temperature and heart rate, one-minute medians were calculated to match the interval of the skin sensors. Additionally, only heart rate results with corresponding signal quality index above 50 were used. The signal quality index is an algorithm quantitatively assesses the functional near-infrared spectroscopy signal quality on a numerical scale from 0 (very low quality) to 100 (very high quality). Moreover, longitudinal data from the five skin temperature measurements were corrected using an external calibration. For the calculation of the mean skin temperature (MST) for one experiment, the results of all five sensors were averaged.

Each physiological parameter was tested for normality using the Kolmogorov–Smirnov test. The analysis of variance (ANOVA) test was utilized to compare the measured data. All parameters were found to be non-normally distributed. Therefore, the Wilcoxon rank-sum test was considered. Since this method is not robust against systematic interindividual variability, a linear mixed-effects model analysis was used to properly consider the interindividual differences. The mixed model approach was broadly used in previous accelerometer studies (Van Dongen et al. 2004; Haapalainen et al. 2008; Pfeiffer et al. 2009; Bolton et al. 2021). In particular, we were interested in making conclusions about how the trial settings over time (fixed effects) impact the measured physiological parameters by controlling the individual differences (random effects). Alpha (α) level at 0.05 was set for all statistical tests. All p-values were two-tailed. The data cleaning process and statistical analysis were performed using R statistical software (version 4.1.2.®).