If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Healthcare-associated infections are associated with increased patient mortality. Hand hygiene is the most effective method to reduce these infections. Despite simplification of this easy intervention, compliance with hand disinfection remains low. Current assessment of hand hygiene is mainly based on observation by hygiene specialists. The aim of this study was to investigate additional benefits of eye-tracking during the analysis of hand hygiene compliance of healthcare professionals in the intensive care unit.
In a simulated, randomized crossover study conducted at the interdisciplinary intensive care unit at University Hospital Zurich, Switzerland, doctors and nurses underwent eye-tracking and completed two everyday tasks (injection of 10 μg norepinephrine via a central venous line, blood removal from the central line) in two scenarios where the locations of alcoholic dispensers differed (‘in-sight’ and ‘out-of-sight’). The primary outcomes were dwell time, revisits, first fixation duration and average fixation duration on three areas of interest (central venous line, alcohol dispenser, protective glove box) for both scenarios. Compliance with hand hygiene guidelines was analysed.
Forty-nine participants (35 nurses, 14 doctors) were included in this study. Eye-tracking provided additional useful information compared with conventional observations. Dwell time, revisits, first fixation duration and average fixation duration did not differ between the two scenarios for all areas of interest. Overall compliance with recommended hand hygiene measures was low in both doctors (mean 20%) and nurses (mean 42.9%).
Compared with conventional observations, eye-tracking offered additional helpful insights and provided an in-depth analysis of gaze patterns during the recording of hand hygiene compliance in the intensive care unit.
]. Despite the simplification of hand hygiene with the introduction of alcohol-based disinfectants in the 1990s and the publication of hand hygiene guidelines by the World Health Organization (WHO) in 2009 [
]. Multiple possible reasons for non-compliance have been investigated. One possible explanation is an inverse relationship between staff workload and compliance. Accordingly, low compliance rates were detected in intensive care units (ICUs) [
]. Further possible reasons include distraction, oblivion, missing or empty dispensers, or inconvenient positioning of hygiene equipment.
The current standard to assess hygiene compliance is direct external observation and detection of possible mismatches between the number of indications (opportunities) and the number of effective hand disinfections by specialized hygiene staff [
], where individuals change their behaviour in response to the awareness of being observed.
A more objective, non-biased methodology to assess adherence to hand hygiene is desirable. Eye-tracking is one potential technique that may meet these criteria, is safe and has been used in healthcare settings in a variety of scenarios, including ICUs [
]. Using infra-red light reflected by a participant's cornea, precise characterizations of visual behaviour, situational awareness and decision-making have become possible. The aim of the present study was to investigate advantages and potential additional benefits of eye-tracking during the analysis of compliance with hand hygiene rules in the ICU.
All participants gave written informed consent. The local ethics committee (Kantonale Ethikkommission Zurich, BASEC ID Req-2017-00798) approved the study protocol.
Study design and population
This prospective, simulated, randomized crossover study was conducted at the interdisciplinary ICU of University Hospital Zurich, Switzerland. All doctors and nurses (females and males) working in the ICU were eligible to participate, provided that informed consent was given. Participation in the study was voluntary. Subjects with visual disturbances (lack of stereoscopic vision, monocular vision or achromatopsia) were excluded from this study.
Based on participant numbers of previous eye-tracking studies in comparable settings [
], a minimal participant number of 25–30 participants was targeted. After signing the informed consent form, demographic data and data that could potentially influence eye-tracking, such as visual disturbances, were collected. The participant was included in the study if three-point calibration of the eye-tracker was possible and after habituation (15–30 min) to the device.
In order to obtain non-biased results, the participants were not aware of the aim of the study, and were not trained in hygiene measures for the purpose of the study. The participants were instructed to behave as they would in everyday professional life. The experiment started in front of a standard ICU patient treatment room, where the eye-tracking glasses were set up and the recordings were initiated. After recording commenced, the participant had to enter the ICU patient treatment room and was asked to complete two tasks in two scenarios; these were the same for all participants. For the tasks to be performed, a plastic simulator mannequin was placed in an ICU patient bed in the real ICU environment. The mannequin had a right subclavian central venous catheter (CVC) inserted. The first task was to inject 10 μg norepinephrine (standard dilution used regularly in the ICU) via the central venous line using a 10-mL Braun syringe (B. Braun, Melsungen, Germany). In the second task, blood had to be taken from the CVC with the in-house syringe used for blood gas analysis (safePICO, Radiometer, Thalwil, Switzerland). The drawing of blood was simulated with the aid of a bag filled with red coloured saline implanted in the mannequin.
Each participant had to complete both tasks in two scenarios. The scenarios were the same for all participants. In Scenario 1, the alcohol dispenser was placed ‘in-sight’ of the participants (i.e. next to the CVC and close to the head of the mannequin). As such, the dispenser was visible for participants without changing their spatial position while performing the tasks (Figure 1a). In Scenario 2, the dispenser was placed ‘out-of-sight’, with the need to change spatial position to perform hand disinfection (Figure 1b). The walking distance to the dispenser was not significantly longer in Scenario 2. A computer randomization program (www.random.org) was used to determine if each participant started with Scenario 1 or Scenario 2.
After the two scenarios, all participants completed a post-experiment questionnaire assessing disturbance caused by the eye-tracking device, subjective stress during the experiment, previous training in hand hygiene, familiarity with the WHO guidelines for hand hygiene, the extent to which they subjectively adhered to the WHO guidelines for hand hygiene during the experiment, and visibility and accessibility of the alcohol dispensers in the two scenarios.
Eye-tracking technology, data recording and analysis
The study was conducted using the SMI Eye-tracking Glasses 2 Wireless system (SensoMotoric Instruments, Teltow, Germany) (Figure S1, see online supplementary material). Eye-tracking describes the recording and study of movements of the eyes in following a moving object, lines of printed text or other visual stimuli, used as a diagnostic procedure or a means of evaluating and improving the visual presentation of information. It involves in-depth characterization of participants' eye movements, as well as the behaviours of the participants' pupils using emitted infra-red lights reflected by the cornea and detected by cameras integrated in the eye-tracking glasses. Eye-tracking glasses are generally not relevantly larger than regular corrective glasses. Specific gaze metrics provide indices to assess distinct visual patterns of participants [
If the participants were myopic or hyperopic, corrective glasses provided by the manufacturer were available. Gaze tracking was executed at a sampling rate of 60 Hz. The angle of view was measured with an accuracy of 0.5° over all distances. The scene video was recorded with a resolution of 960 × 720 pixels at 30 fps. Raw data were processed using SMI BeGaze Version 3.6 (SensoMotoric Instruments). The SMI algorithm for fixation determination was used. All recorded eye-tracking videos were stored in the study computer. ‘BeGaze’ Version 3.6 (SensoMotoric Instruments) was used for data analysis. Every ocular fixation of participants during the simulated events was manually assigned to the areas of interest (AOIs), which were mapped on to a reference snapshot image.
Areas of interest and outcome parameters
Gaze patterns of participants related to the simulated interventions were recorded, irrespective of participants' subsequent actions. For each scenario, a total of three AOIs (spatial areas of importance for the study purpose) were defined prior to the experiments: the CVC, the alcohol dispenser (‘in-sight’ and ‘out-of-sight’), and the protective glove box. Outcome measures were dwell time (cumulated time spent on an AOI, including fixations, blinks and saccades), revisits (frequency of revisiting a particular AOI after gazing at other areas), first fixation duration, and average fixation duration for all AOIs.
Observational video analysis
In an analysis of the video recordings conducted by the study team, objective adherence to hand hygiene rules compared with the opportunities that would mandate hand hygiene was investigated, independent of participant gaze patterns. For this analysis, mandatory hand hygiene moments in accordance with the published WHO guidelines [
] were: before initial patient contact; before aseptic procedure; after body fluid exposure risk; after touching a patient; and after touching a patient's surroundings. In order to provide a sufficient amount of disinfectant, the button of the dispensers used at the study institution needed to be used twice in series.
All members of the study were familiarized with the current WHO hand hygiene rules in a training session conducted by the Department of Infectious Diseases and Hospital Epidemiology before the recordings started. The analysis of compliance was conducted by two members of the study team. Both members assessed the video recordings independently. In the case of possible discrepancy, a third member of the study team was involved in order to resolve the issue.
Data were modelled by means of linear mixed-effects models, considering eye-tracking measurements as dependent variables, and observed object as well as in-sight or out-of-sight position of the dispenser as fixed effects while accounting for subject random effect. Additional multi-variable adjustment for age and experience of the individual subjects was considered, but this was dropped as it did not modify the effects inferred by the simpler model. P-values for individual fixed effects were obtained by Satterthwaite's degrees of freedom method.
Statistical analysis was performed via a fully scripted data management pathway using the R Environment for Statistical Computing Version 4.0.2. A two-sided P<0.05 was considered to indicate significance.
After assessment for eligibility, 49 participants were included in the study. No adverse events occurred during the recordings.
Table I provides an overview of the baseline characteristics of the participants. Three-quarters (73.5%) of the participants were female, the median age was 34 years, and the median job experience was 10 years.
Table IBaseline characteristics of participants
Male sex (N, %)
Nurses (N, %)
Doctors (N, %)
Volume of work (%)
Median job experience (years)
Wearing corrective glasses
No (N, %)
Yes (N, %)
Data are provided as absolute numbers (%) or median [interquartile range], as appropriate.
Figure 2 shows dwell time, revisits, first fixation duration and average fixation duration on all AOIs for the two scenarios. Dwell time on the glove box and dispenser did not differ between the two scenarios (glove box 81612.63 ms vs 78868.01 ms, P=0.825; dispenser 44343.82 ms vs 41793.36 ms, P=0.77). Dwell time on the central venous line was slightly higher in the ‘in-sight’ scenario (1444783.91 ms vs 1396935.73 ms, P=0.705), but the difference was not significant (Figure 2a).
Revisits on the CVC were higher in the ‘out-of-sight’ scenario (20.01 vs 18.20, P=0.157), and revisits on the glove box (2.76 vs 2.03, P=0.113) and dispenser (2.76 vs 2.02, P=0.11) were higher in the ‘in-sight’ scenario; however, the differences were not significant (Figure 2b).
First fixation duration was higher in the ‘in-sight’ scenario for the dispenser (312.07 ms vs 310.50 ms, P=0.98) and the CVC (307.49 ms vs 284.49 ms, P=0.505). First fixation duration on the glove box was lower in the ‘in-sight’ scenario (232.16 ms vs 262.72 ms, P=0.516) (Figure 2c). Figure 2d shows the average fixation duration on the glove box, dispenser and CVC, with no differences between the two scenarios (glove box 312.07 ms vs 325.22 ms, P=0.71; dispenser 340.17 ms vs 349.33 ms, P=0.812; CVC 446.96 ms vs 449.77 ms, P=0.835)
The results of the post-experiment questionnaire are shown in Table II. Overall, only 22% of the participants felt disturbed by the eye-tracking device, and only 18% of the participants felt stressed during the experiment. Two-thirds of the participants declared that they perform CVC-related tasks multiple times daily, and almost 90% stated that they wear protective gloves regularly (Table II). Interestingly, 63% of the participants declared that they normally disinfect their hands both before and after wearing protective gloves. All participants were familiar with dispenser locations on the ICU. Easy accessibility of dispensers was afforded by 65% of participants, and participants felt that there were enough dispensers during the experiment (Table II). In relation to theoretical aspects of hand hygiene, >70% of participants felt subjectively familiar with the five moments of hand hygiene. Almost all nurses and doctors had been trained in hand hygiene before the study (Table II).
Table IIData derived from the post-experiment questionnaire
Disturbed by eye-tracker
Range of motion impaired
Stressed during experiment
Frequency of performance of CVC-related tasks
Subjective safety during handling of CVC (scale 0–5)
Overall compliance with recommended hand hygiene measures was low in both doctors (mean 20%) and nurses (mean 42.9%).
While participants declared subjective compliance with hand hygiene in 70% of all possibilities (Table II), objective adherence as analysed by the video recordings was profoundly low for both doctors and nurses, and was minimally higher in nurses (Table III).
The aim of this study was to investigate additional benefits of eye-tracking in the assessment of compliance with hygiene rules of ICU healthcare professionals. Important additional information could be collected using eye-tracking. Overall compliance with hand hygiene was low in both doctors and nurses.
Conventional approaches for the assessment of hand hygiene compliance are often based on direct observations by specialized hygiene staff, audits or hygiene campaigns [
]. Unfortunately, observation biases are associated with such measures, leading to overestimation of the true hygiene compliance of healthcare workers. As a consequence, hygiene compliance may be impeded, and inadequate conclusions may be drawn with respect to appropriate target interventions. In contrast, the present eye-tracking study revealed mismatches between participants' subjective assessments in the questionnaires and objective eye-tracking recordings, similar to a lie detector. While a substantial proportion of participants declared high subjective compliance with hand hygiene measures, the corresponding objective compliance as assessed by eye-tracking analysis was profoundly low for doctors and nurses. The fact that compliance with hand hygiene was markedly poorer compared with previous literature suggests that eye-tracking is a valid and novel methodology to minimize the occurrence of the Hawthorne effect, which is a strength of the study design and eye-tracking [
]. Further refinements of the technology in the future may be less intrusive and lead to further avoidance of the Hawthorne effect. Despite the observed low compliance, participants declared familiarity with dispenser locations in the ICU in the questionnaire, and had been trained in hand hygiene previously; this discrepancy suggests that further and novel educational measures in ICU staff training are required, beyond pure observational analyses and assessments by questionnaires.
The methodology of eye-tracking, which has not – to the authors' knowledge – been tested previously in similar study settings, offered several further benefits. Owing to its precise technology, objective eye-tracking parameters related to hygiene interventions could be recorded and analysed. In contrast to observations made by health professionals, such as hygiene or hospital-epidemiological staff, subtle differentiated spatial and temporal measurements of gaze patterns and visual behaviour could be collected in the two study scenarios. Eye-tracking revealed accurate gaze metrics, and characterized parameters such as dwell time, revisits, first fixation duration and average fixation duration on different AOIs in high detail. Conventional observations do not allow such analyses. Even if no significant differences in visual behaviour were detected between the ‘in-sight’ and ‘out-of-sight’ scenarios, comparable observations or conclusions could not be drawn by pure observation of workflows. In this context, relatively high dwell times, revisit numbers and average fixation durations on the CVC (compared with other AOIs) may be associated with frequent rechecking gazes and constitute surrogate parameters of the complexity of this particular AOI. Eye-tracking thus allowed in-depth analysis of the participants' consequences of their actions and the characterization of AOIs which are visually difficult to grasp by ICU professionals or which have importance for a specific task. Moreover, with the technology of eye-tracking, accurate analyses of gaze patterns could be conducted depending on different spatial locations of AOIs, such as the alcohol dispensers. The positioning of the dispensers (‘in-sight’ vs ‘out-of-sight’) did not appear to play a crucial role related to the participants' gaze movements; this finding has implications for the design of future investigations. Eye-tracking is thus a useful tool to analyse visual behaviour of ICU professionals in real-life scenarios, and proves feasible as a future innovative approach to hygienic control where spatial positions of AOIs or the correct, ergonomic positioning of dispensers in ICUs are key. A further advantage of eye-tracking is the fact that participants can perform their routine, healthcare-related tasks in an undisturbed, unobserved way, which may minimize further biases compared with other study designs. In line with this, most participants did not feel disturbed by the eye-tracker device, and reported free range of motion during the measurements in the post-experiment questionnaire.
Due to its easy methodology, and uncomplicated and unbiased recordings, eye-tracking may be helpful in the future to evaluate staff teaching or to design further hygiene studies. In future hygiene research projects, eye-tracking could be of particular benefit when the complexity of mandatory actions, spatial locations of AOIs or high precision measurements are key to address specific research questions. Future research scenarios and clinical practice should again include settings where participants are not familiar with the aim of the study (i.e. analysis of hygiene compliance). In such settings, compliance with hygiene could be analysed as a ‘bystander analysis’, while participants principally have to focus on other tasks not related to hygiene. With such approaches, true hygiene compliance may not be overestimated, which may provide the most accurate estimates of hygiene adherence. This will hopefully translate into improvements in hand hygiene compliance and reduced transmission of pathogenic bacteria from patient to patient, which remains unacceptably high [
The study has some limitations. First, it was a single-centric study, probably limiting its external validity. Second, it was performed in a simulated setting. It was therefore not possible to exclude residual confounders completely. Nevertheless, participants in comparable simulated studies still afforded high reality [
]. Third, eye-tracking primarily assesses gaze patterns, which do not themselves reflect cognition of participants to a full extent, nor are they necessarily associated with subsequent participant actions. Fourth, overall compliance with hand hygiene was profoundly low, possibly obfuscating the detection of potential differences in eye-tracking-related gaze patterns between the two scenarios.
In conclusion, eye-tracking provided additional useful information compared with conventional observation methods during the recording of hand hygiene measures in the ICU. Eye-tracking allowed in-depth analysis of the participants' consequences of their actions and detailed characterization of AOIs with high relevance.
The authors wish to thank all participants for their contribution to this study.
Conceptualization: RV, PKB, DAH.
Data curation: PDWG, PKB.
Formal analysis: RV, PDWG.
Investigation: RV, PDWG, RAS, PKB, DAH.
Project administration: RV, PKB, DAH.
Writing – original draft: RV.
Writing – review and editing: RV, PDWG, RAS, PKB, DAH.
Conflict of interest statement
Appendix A. Supplementary data
The following is the Supplementary data to this article:
Example of an eye-tracker device with a small recording camera built in (SMI eye-tracking glasses 2 wireless system, SensoMotoric Instruments, Teltow, Germany). A small wire transmits measurements to a processor for analysis.
Selected aspects of the socioeconomic impact of nosocomial infections: morbidity, mortality, cost, and prevention.