Healthcare-associated infections, including surgical site infections, ventilator-associated pneumonia, urinary tract infections, and catheter-associated bloodstream infections, prolong length of hospital stay and increase cost, morbidity and mortality [1–3]. Additionally, antibiotic resistance is emerging worldwide as a serious health threat [4].

Transmission of potential pathogens between patients occurs primarily via healthcare worker (HCW) hands when hand hygiene (HH) is omitted at critical moments [5, 6]. Such hand-to-surface exposures (HSE) occur frequently [7], resulting each time in a bi-directional exchange of microorganisms between the hand and the touched surface [6]. In consequence, hands transport microorganisms sequentially between surfaces [6]. Depending on the nature of the microorganisms and of the receiving surface, this can result in patient harm. If microorganisms feature antibiotic resistance, their transmission to a patient can result in prolonged carriage. If the microorganisms are virulent and the receiver surface is a skin lesion or an invasive device such as a central venous line, the transmission may result in healthcare-associated infection.

Several studies show that infectious microorganisms can survive on human skin long enough to be cross-transmitted and that hand hygiene using alcohol-based handrub is an effective way to decrease this transmission [8, 9]. With the WHO “My five moments for hand hygiene”, a user-centered concept based on education, training, monitoring and reporting of hand hygiene has been introduced with the goal to bridge the gap between scientific evidence and daily healthcare practice [10]. Yet, HCWs still fail to consistently apply hand hygiene. The lack of awareness regarding what people touch during their routine work may play an important role in this failure to adhere to established rules [11]. Today’s gold standard to monitor HH performance consists of direct observation of healthcare workers by trained observers during patient care [5, 12–14]. This method may not capture every HSE during fast-paced care and thus, underestimates the true risk of pathogen transmission [7, 15]. On the other hand, automated electronic hand hygiene monitoring systems still fall short of detecting all hand hygiene opportunities [16].

To better understand the nature of microbial hand-transmission in a real-life intensive healthcare setting, we built and pilot-tested a new observation and coding system that would consistently capture every HSE, and thus allow to study true transmission risks via HCWs’ hands.

The 10 active care video sequences totaled 296.5 min and featured eight nurses of whom seven were female and two physicians of whom one was female, all right handed. Overall, 4222 HSE occurred, translating in an overall density of 14.2 HSE per minute or one HSE every 4.2 s. Exemplarily, Fig. 3 demonstrates the coding timeline of all HSE and hand hygiene actions in the first 3 min of video #7. Details on the frequency and nature of HSE and hand hygiene actions overall and per each video sequence appear in Table 2.

Video	#1	#2	#3	#4	#5	#6	#7	#8	#9	#10	Overall
ICU specialty	Trauma	Trauma	Trauma	Trauma	Cardio-surgery	Cardio-surgery	Cardio-surgery	General surgery	Cardio-surgery	General surgery
Length of coded care sequence; min:sec	34:50	34:50	34:50	36:20	31:17	32:38	33:31	16:39	32:19	16:32	296:30
Gender	M	F	F	F	F	F	M	M	F	F
Profession	N	N	N	N	N	N	N	P	N	P
HSE; n	494	472	314	495	474	671	553	176	526	47	4222
HSE density; n/min	14,2	13,6	9,0	13,6	15,2	20,6	16,5	10,6	16,3	2,8	14,2
Mean HSE duration (SD); sec	8.9 (16.0)	6.8 (10.0)	11.7 (17.0)	8.1 (18.5)	7.0 (14.9)	5.3 (10.5)	5.6 (11.8)	11.7 (15.1)	6.8 (10.7)	11.8 (26.5)	7.44 (14.1)
Hand
Right hand (%)	273 (55.3)	254 (53.8)	173 (55.1)	277 (56.0)	250 (52.7)	358 (53.4)	304 (55.0)	93 (52.8)	292 (55.5)	25 (53.2)	2299 (54.5)
Left hand (%)	221 (44.7)	218 (46.2)	141 (44.9)	218 (44.0)	224 (47.3)	313 (46.7)	249 (45.0)	83 (47.2)	234 (44.5)	22 (46.8)	1923 (45.6)
Gloves worn during HSE
No (%)	355 (71.9)	420 (89.0)	221 (70.4)	214 (43.2)	406 (85.7)	671 (100)	334 (60.4)	176 (100)	474 (90.1)	47 (100)	3318 (78.6)
Yes (%)	139 (28.1)	52 (11.0)	93 (29.6)	281 (56.78)	68 (14.4)	0	291 (39.6)	0	52 (9.9)	0	904 (21.4)
Any surface inside patient zone (% of all HSE)	289 (58.5)	222 (47.0)	196 (62.4)	133 (26.9)	131 (27.6)	134 (20.0)	225 (40.7)	122 (69.3)	278 (52.85)	45 (95.7)	1775 (42.0)
Patient intact skin (% of HSE inside patient zone)	12 (4.2)	30 (13.5)	12 (6.1)	13 (9.8)	0	15 (11.2)	45 (20.0)	8 (6.6)	29 (10.4)	14 (31.1)	178 (10.0)
Mobile object inside patient zone (% of HSE inside patient zone)	241 (83.4)	163 (73.4)	130 (66.3)	112 (84.2)	117 (89.3)	80 (59.7)	168 (74.7)	109 (89.3)	234 (84.2)	29 (64.4)	1383 (77.9)
Immobile surface inside patient zone (% of HSE inside patient zone)	36 (12.5)	29 (13.1)	54 (27.6)	8 (60.2)	14 (10.7)	39,829.1)	12 (5.3)	5 (4.1)	15 (5.4)	2 (4.4)	214 (12.1)
Any surface outside patient zone (% of all HSE)	148 (30.0)	220 (46.6)	89 (28.3)	350 (70.7)	322 (67.9)	506 (75.4)	115 (20.8)	53 (30.1)	148 (28.1)	2 (4.3)	1953 (46.3)
HCW own body (outside patient zone) (% of HSE outside patient zone)	7 (4.7)	36 (16.4)	25 (28.1)	47 (13.4)	60 (18.6)	107 (21.2)	79 (68.7)	50 (94.3)	28 (18.9)	0	439 (22.5)
Mobile object outside patient zone (% of HSE outside patient zone)	114 (77.0)	160 (72.7)	49 (55.1)	235 (67.1)	175 (54.4)	346 (68.4)	18 (15.7)	3 (5.7)	92 (62.2)	2 (100)	1194 (61.1)
Immobile surface outside patient zone (% of HSE outside patient zone)	27 (18.2)	24 (10.9)	15 (16.9)	68 (19.4)	87 (27.0)	53 (10.5)	18 (15.7)	0	28 (18.9)	0	320 (16.4)
Any critical site (inside patient zone) (% of all HSE)	57 (11.5)	30 (6.4)	29 (9.2)	12 (2.4)	21 (4.4)	31 (4.6)	213 (38.5)	1 (0.6)	100 (19.0)	0	494 (11.7)
Sterile equipment (% of HSE at critical site)	1 (1.8)	0	0	0	0	0	123 (57.8)	1 (100)	82 (82.0)	0	207 (41.9)
Invasive device access (% of HSE at critical site)	55 (96.5)	30 (100)	29 (100)	9 (75.0)	21 (100)	21 (67.7)	65 (30.5)	0	8 (8.0)	0	238 (48.2)
Mucous membrane (% of HSE at critical site)	1 (1.8)	0	0	3 (25.0)	0	0	0	0	0	0	4 (0.8)
Wound (% of HSE at critical site)	0	0	0	0	0	10 (32.3)	25 (11.7)	0	10 (10.0)	0	45 (9.1)
Infectious risk events	41	42	43	26	44	80	117	31	72	2	508
Patient colonization events	13	26	23	16	25	54	65	30	37	2	291
Patient infection events	38	16	20	10	19	26	52	1	35	0	217
Hand hygiene actions; n	8	13	14	7	11	9	15	4	11	5	97
Hand hygiene action at colonization event; n (% of patient colonization events, i.e. ‘adherence’)	0	3 (11.5)	2 (8.7)	1 (6.2)	3 (12.0)	2 (3.7)	0	2 (6.7)	1 (2.7)	0	14 (4.8)
Hand hygiene action at infection event; n (% of patient infection events, i.e. ‘adherence’)	0	0	0	0	0	0	2 (3.9)	0	1 (2.9)	NA	3 (1.4)
Average density of hand hygiene actions; n/hour	13.8	22.4	24.1	11.6	21.1	16.5	26.9	14.4	20.4	18.1	19.6
Mean duration of hand hygiene actions (SD); sec	8.6 (4.7)	14.9 (6.6)	22.2 (11.0)	18.8 (7.4)	11.7 (4.5)	9.2 (4.8)	10.5 (9.6)	16.3 (12.0)	7.9 (3.7)	11.6 (6.2)	13.2 (8.6)

Legend: HSE hand-to-surface exposure; ICU intensive care unit; F female; M male; N nurse; P physician; NA not applicable; SD standard deviation. Definition for patient colonization event and patient infection event s. main text

The mean and median duration of the 97 observed hand hygiene actions were 12.9 (SD, 8.7) and 11 (range, 2–48) seconds, respectively. Patient colonization events occurred overall 291 times, 139 for the left and 152 for the right hand. Patient infection events were observed overall 217 times, 103 for the left and 114 for the right hand. Importantly, 117 (61%) of colonization events and seven (2.3%) infection events occurred after HCWs touching their own body. HCWs touched themselves 439 times (10% of all HSE), including their clothes 165 (38%), personal protective equipment 21 (5%), their face 24 (6%), and remaining bare skin or hair 229 (52%) times; 13 (3%) times with gloved hands.

Hand hygiene occurred prior to 14 of the 191 colonization events and three of the 217 infection events, resulting in a hand hygiene ‘adherence’ of 5% and 1%, respectively.

This unique video observation and coding approach, that considers each single HSE by both HCW hands, revealed a surprising reality of transmission opportunities during real-world intensive care. The overall density of 14.2 HSE per minute with which HCWs’ hands touched surfaces during active patient care is high, suggesting that hands acquire and deposit – and thus likely transmit – potentially harmful microorganisms every 4 s onto patients and surfaces in the care environment. We identified sequences of particular interest for infection prevention, such as patient zone entries and transitions to critical sites, which each occurred roughly every 2 min of active patient care in an ICU. Hand hygiene was performed on average 19.6 times per hour, which equals one hand hygiene action every 3 min. It is not surprising that participants only sustained hand rubbing for a median of 11 s against the recommended 20–30 s [17]. In fact, if meeting the recommended duration for hand rubbing, almost one fifth of active patient care time would have been spent on this activity. Recent data indicating that 15 s might suffice are comforting in this respect [18].

The approach used in this study is in line with a human factors task analysis, whose underlying principle is to break down a task to study its individual elements [19]. In doing so, we aim to understand the factors that influence the way work is being done and, ultimately, what can be done to improve it [20, 21]. In doing so, the moments we report here are more frequent than those usually reported in direct hand hygiene observation studies. For example, tasks such as a dressing change are typically seen as constituting one single hand hygiene opportunity with an indication ‘Before clean/aseptic procedure’ before the task and ‘After body fluid exposure risk’ at the end of the task [10]. In the current approach, each care task is split into multiple HSEs, taking into account both mobile objects [22] and the HCWs own body, each scrutinized for potential hand contamination and transmission. Furthermore, traditional hand hygiene models are based on the assumption that surfaces within the patient zone are colonized primarily with the patient’s own flora. Our results [11], however, demonstrate that frequent transitions of hands into the patient zone without hand hygiene may lead to contamination of the patient zone with foreign microorganisms. Such lapses lead to an unsafe system state, which creates ambiguity [23] and may result in unintentional patient harm.

Our approach revealed further noteworthy realities. We considered the HCW’s own body as an ‘Outside patient zone’ surface. More than half of all HSE sequences (61%) from the “outside” to the “inside” patient zone were due to ‘self-contact’. Current hand hygiene guidelines often fail to address HCW self-contact as an indication for hand hygiene [17]. Hence, such HSE are usually ignored by observers. Second, much variation exists in whether HCWs perceive their professional apparel as a potential source of bacteria, leading to variations in hand hygiene [24]. Additionally, as described by Sax & Clack, relying on automatic, unconscious behaviors fuelled by “mental models” for routine tasks is inherent to the nature of human beings, allowing mental resources to be spared for more complex tasks [11]. This suggests that people often are unaware of what exactly their hands do while they are focused on the main task goal [11]. The average of 1.48 exposures per minute to a HCW’s own body is consistent with previous findings [25, 26]. However, with only 4.87 exposures per hour to “HCW Face”, our results differed from studies who found that face contact occurred up to 15–23 times per hour among students during 2-h lectures [26] or during office-type work [25]. Finally, glove use was frequent, representing one fifth of all HSE. Gloves represent mobile surfaces that transport microorganisms like bare hands. Further research could explore the nature of HSE during glove use to inform best practice for glove indications.

The “first-person view” of a head-mounted action camera provides the advantage of an unobstructed view of both hands and the surfaces they touch following the healthcare worker [27] even when leaving the immediate care area, neither of which can be guaranteed with a fixed-position camera. From anecdotal reports by the participants, their awareness of wearing a camera and their activity being registered waned quickly, suggesting a minor Hawthorne effect, yet this remains to be studied systematically. Contrary to concerns about video recording in acute care settings, we found that once healthcare workers, patients, and their relatives were informed of the study goals, objections to filming were rare. Video observation of hand hygiene behavior has been used before [28–31] but never from a first-person view and never to record HSEs.

Our approach has limitations. The analysis is limited to a small sample of healthcare workers in three ICUs and in consequence not representative for care in general. We do not expect, however, the main findings of frequent HSE to be categorically different. Furthermore, while the sequential analysis we report here considers only pairs of two consecutive HSE leading up to “colonization” or “infection” events, it is important to recognize that HSE occur in long sequential chains. The exact benefit of hand hygiene at any of these moments has not been considered in our current calculation, nor in the WHO ‘Five moments’ concept. In this line of thought, our approach might serve as basis for more advanced future transmission risk modelling. Our definition of a colonization event deviated from ‘Moment 1’ of the WHO hand hygiene concept by including any object within the patient zone, not only the patient. We did this intentionally to identify the transmission trajectories most likely leading to contamination of high-touch surfaces near the patient and ultimately, the patient. On a technical note, the specific software is expensive and its use requires expertise. Video coding is more time-consuming than live observation. Hence, before introducing this instrument into day-to-day practice beyond research, simplification and automation is a desirable next development step. Finally, the videos were coded by a single coder (MS) and supervised by a second person (LC) due to feasibility. The possibility to pause and rewind the video likely minimized the risk of miscoding.

In conclusion, our approach produced a valid video and coding instrument for analysis of detailed HSE trajectories. Using a head-mounted action camera and a comprehensive coding system, we could show for the first time in a fast-paced, real clinical setting how frequently healthcare workers’ hands touch surfaces, corroborating the fast spread of microorganisms in healthcare settings. Further development and use of this method may contribute to the design of more efficient preventive strategies.

Using a new head-mounted action camera and a systematic coding tool, we could show for the first time how healthcare workers’ hands touch surfaces in a real-world clinical setting. This human factors approach to task analysis demonstrated the hand trajectories via which microorganisms can spread in healthcare and revealed that hand hygiene adherence is lower than usually reported by traditional on-site observations. This new instrument may assist in designing more efficient preventive strategies on an individual and systems level.