Bacterial transfer to fingertips during sequential surface contacts with and without gloves

Bacterial transmission from contaminated surfaces via hand contact plays a critical role in disease spread. However, the fomite-to-finger transfer efficiency of microorganisms during multiple sequential surface contacts with and without gloves has not been formerly investigated. We measured the quantity of Escherichia coli on fingertips of participants after 1-8 sequential contacts with inoculated plastic coupons with and without nitrile gloves. A Bayesian approach was used to develop a mecha nistic model of pathogen accretion to examine finger loading as a function of the difference between E coli on surfaces and fingers. We used the model to determine the coefficient of transfer efficiency ( λ ), and influence of swabbing efficiency and finger area. Results showed that λ for bare skin was higher (49%, 95% CI = 32%-72%) than for gloved hands (30%, CI = 17%-49%). Microbial load tended toward a dynamic equilibrium after four and six contacts for gloved hands and bare skin, respectively. Individual differences between volunteers’ hands had a negligible effect compared with use of gloves ( P < .01). Gloves reduced loading by 4.7% (CI = −12%-21%) over bare skin contacts, while 20% of participants accrued more microorganisms on gloved hands. This was due to poor fitting, which created a larger finger surface area than bare hands.

cases. 4,5 With an average cost of £5239 ($6600 USD) in hospital care per case, 6 the prevention of hospital-acquired infections (HAIs) is a major priority for both the UK National Health Service and health organizations worldwide. While the transmission routes for these pathogens are still poorly understood, 7 it is thought that 20% of HAIs are spread through direct or indirect hand-to-mucosa contact. 8 Hands of healthcare workers provide a dynamic vector for the transfer of microbes from contaminated hospital surfaces to susceptible patients. 9,10 Although there is strong emphasis on hand hygiene, the role of indirect contact, surface, or fomite transmission through contaminated surfaces may be underestimated. 11 Since there is also a relationship between contact and airborne transmission, pathogens are known to deposit onto surfaces from the air and may then survive for many hours. [12][13][14][15] The transfer efficiency (λ) between microorganisms on surfaces and hands is an important parameter for understanding infection transmission risk. 16,17 It is used in contact transmission models to predict contamination of hands and hence model the exposure of patients and healthcare workers to infectious pathogens. Values for λ are therefore derived from experimental studies and are known to vary with microorganism, surface material, use of gloves, and type of contact. In terms of relevant hospital-acquired pathogens, the transfer efficiency of Gram-negative bacteria such as E coli and Acinetobacter baumannii from dry non-porous fomites to fingertips has been reported to lie between 0.1% and 76% 18 under dry air conditions: 20%-40% relative humidity. Since experimental variability is so large, it is hard to say whether higher environmental humidity increases this transfer rate. However, for 40%-65% relative humidity, experiments have reported values between 0.3% and 100% for surfaces such as glass, stainless steel, ceramic, and granite with an average λ of 52% (σ = 19%) (see Table 1). 18,19 The use of nitrile gloves has been shown to reduce Staphylococcus aureus transfer to 38% (σ = 18%), 20

while no significant difference in transfer between
Gram-positive and Gram-negative bacteria has been found without the use of gloves. 19 When adding an intermediary plastic or metal surface, bacteria are transferred between gloved hands with efficiencies ranging from 0.1% to 75% (mean = 37%, σ = 32%) 20 ; applying a twisting shear stress seems also to increase transfer between

Practical Implications
• Transfer efficiency is an important parameter in contact infection transmission models.
• Approximate Bayesian computation is a flexible method for fitting a model to data.
• Swabbing efficiency was found to be highly significant in this experimental study, and needs to quantified in future experiments.
• Finger surface area is increased by gloves that do not fit well, and this encourages the acquisition of microorganisms.
• Gloves should be the correct size and fingers should fit snugly.
• Healthcare staff should be reminded about the need to change gloves between specific duties in accordance with local protocol. surfaces by a factor of three. 21 After a third contact, transfer efficiency substantially reduces to 1.25% (σ = 0.9%) 21 and transfers to and from porous surfaces, including fabrics to hands, are often slightly lower (mean = 0.5%). 19,22 Although these percentages may seem small, putting them into a clinical context such differences could significantly alter patient outcome.
A simple glance at Escherichia coli, which has traditionally been used as a fecal indicator, has also been used as a model Gram-negative pathogen in previous transfer studies since it is safer for direct hand-to-surface contact in transfer studies than pathogens of interest and is still considered representative of several multidrug-resistant organisms such as Klebsiella spp., Pseudomonas aeruginosa, and Shigella spp. 19,23 Specific transmission parameters, such as transfer efficiencies from fomite to finger, are needed to measure the risk of pathogen transmission; this will enable the use of quantitative microbial risk assessment in diverse indoor environments, for example, hospitals, offices, and transportation.
The aim of this study is to quantify the transfer of E coli between plastic surfaces and fingers through sequential contacts and to determine how the wearing of nitrile gloves affects this transfer.
An experimental study is carried out to measure contamination of fingers of volunteers following sequential contact with between 1 and 8 surfaces. Significant novelty is through fitting the data to a model using approximate Bayesian computation to assess experimental variability and to estimate values for the transfer efficiency for gloved and un-gloved hands.

| ME THODOLOGY
In summary, participants were asked to touch a sequence of up to eight E coli inoculated plastic coupons using each of their fingers in turn (thumbs were controls) (see Figure 1). All participants took part in the investigation with and without nitrile gloves. Fingertips were sampled and swabs plated onto selective growth media for quantitative evaluation. A detailed method is laid out in what follows.

| Preparation of volunteer hands (n = 35)
Thirty-five volunteers participated in the study over a 7-week period:

| Preparation of inoculum
A laboratory strain of E coli was prepared at the beginning of the experiment phase using a loopful of bacteria transferred to 100 mL of nutrient broth (Oxoid Ltd). This was incubated at 37°C for 18 hours with a 10-mL sample centrifuged for 30 minutes. The pellet was resuspended in 10 mL of 98% Ringer's solution (Oxoid Ltd) and 2% Tween-80 (Oxoid Ltd). 18 Serial dilution followed by culture on F I G U R E 1 Schematic diagram showing the sequence for touching inoculated surfaces with fingertips tryptone bile X-glucuronide (TBX) (Oxoid Ltd) was used to approximate the concentration (~1.12 × 10 9 CFU/mL).

| Preparation of surface coupons
High-pressure laminate plastic (ELS Panels) was marked into 3 cm × 3 cm squares using an indelible marker. Since autoclaving damaged the surface properties, this was sterilized by submerging in 70% isopropyl alcohol, allowed to dry overnight in a fume cupboard, and screened for live organic bioburden using an adenosine triphosphate swab (Hygiena Int.) as a qualitative assurance. A 100-μL aliquot of culture solution was pipetted onto the marked coupons, spread out using a sterile spreader, and allowed to air-dry at 21°C ± 1°C and 48% ± 2% relative humidity for 60 minutes.

| Fingertip contacts and sampling
Coupons were flat on a laboratory bench and could not be moved by participants. Participants touched the inoculated coupons with one fingertip at a time at an interval of 1 second following the pattern in it might have been used as finger 5 in another volunteer's experiment set. Participants were trained to apply 50 g (±5 g) pressure for 1 second during each surface contact using a top-balance (sensitivity 0.002 g), which relates to a "light-touch". 28 Contact with the surface was standardized to a time of 1 second.
Immediately after a fingertip was used to touch the required number of inoculated coupons, the fingertip was sampled with a sweeping and rotating motion using a sterile cotton swab moistened with sterile Ringer's solution + Tween-80 solution. This was done to remove E coli transferred to the fingertip during surface contacts. 29 All samples were transferred to 10 mL Ringer's solution + Tween-80, shaken for 30 minutes at 36.6°C before being serially diluted (1 mL in 10 mL), and then, 0.1 mL was spread onto Petri dishes containing TBX agar. Plates were incubated for 24 hours at 37°C, and visible colonies were counted. The experiment was crossed as all participants performed un-gloved conditions followed by gloved conditions to avoid cross-contamination (so not counterbalanced); however, whichever finger they used to touch a surface was randomized à priori but maintained for both conditions.

| Statistical analysis
Statistical analysis of the experimental data was carried out using R (R project version 3.3.2) to investigate the effect of gloves, number of sequential contacts, and the participants' implicit differences in pressure and contact surface area on CFU loading. À priori sample size was estimated conservatively at 35 based on a medium effect size of 0.71. 18,23 Welch's two-tailed t test was used to assess statistical differences between groups as a preliminary measure.
A linear mixed-effects model 30 with log 10 (CFU) as a dependent variable was applied using the lme4 R package (version 1.1-18-1) 31 to investigate the joint effect of gloves, contact number (fixed effect), and individual variability between participants (random slopes). Mixed-effects models account for shared variance within participants while modeling between-participant differences (see Appendix S1 for mathematical details).

| Modeling of bacterial counts on fingers
Current microbial loading on fingers (C n ) after contact n can be described using a recurrent relation that is dependent on previous loading (C n−1 ), surface loading (C (s) ), and a transfer efficiency parameter specific to that surface type and microorganism (λ). 19,23 Participant finger area was measured with a ruler: mean of 1.71 cm 2 and standard deviation 0.34 cm 2 . But the exact contact area (A f ) is not known and hence must be estimated. Exact inoculate numbers on the surface (C (s) ) after die-off % (d) are not known (C (s) ·d) either and so are estimated (see Table 2). A modified version of Julian  1 to account for finger area and die-off. A subscript g was introduced to denominate whether the hands were gloved or un-gloved, creating two forms of the equation In this equation, transfer efficiency (λ) represents the ratio of the number of bacteria recovered from the finger after contact with the inoculated surface divided by the number of bacteria present on the surface during the contact. Transfer efficiency is notoriously complex to measure, however, and often thought to be underestimated due to variability in initial inoculum, bacterial inactivation rates, and swabbing efficiencies during the experiment. 23 (1) to the experimental data while calibrating transfer efficiency that optimally represents the mechanism of sequential contacts. It also included the effect of variability in initial surface inoculum concentration (C (s) ), inactivation of bacteria during drying (d), and sampling efficiency using cotton swabs (S eff ). 18 Since swabs are known to underestimate the CFU count because bacteria are retained in the fibers, 35 we took this into consideration when comparing with the experimental data by artificially multiplying by a swabbing efficiency chosen from Table 2 for each case: C g,n ⋅ S eff . We assumed that all other parameters were the same between cases, regardless of whether the participant wore gloves (eg, the surface inoculum did not vary just because the participant wore gloves, although we will later see that finger area might).
The CFU concentration in the initial inoculum on the surface (C (s) ) was represented by a uniform distribution between 5 × 10 7 CFU and 5 × 10 9 CFU per mL. 18 Survival of bacteria over the 60-minutes  Table 2). An initial guess or prior guess of transfer efficiency is taken from the work by Lopez et al. 19 The parameters used in the ABC method are summarized in Table 2, and the ABC algorithm is described in Appendix S2. Figure 2 shows the comparison of log 10

| Effect of gloves
A comparison plot of E coli loading on fingertips for both gloved and un-gloved hands after sequential contacts is shown in Figure 3.
Pairwise comparison of gloved and un-gloved loadings reveals a statistically significantly difference for all contact counts (P < .005).

| Effect of individual participant variability
Examining data from individual participants, correlation between gloved and un-gloved burdens is significant (P < .01), meaning that if a participant acquired a large microbial burden when gloved, then they were also likely to exhibit a high CFU burden when un-gloved.

| Effect of sequential contacts
Oldham's method 37 was used to calculate the correlation between fingertip CFU counts from the first contact (baseline) vs CFU counts after n contacts. This suggests that the amount of E coli accrued during the first contact has a statistically significant effect on loadings up to, and including, contact number 4 (P < .001). Figure 4 shows boxplots of percentage differences of CFU between subsequent contacts ((c g,n − c g,n−1 )∕c g,n−1 × 100), where n represents the contact number and g is either gloved or un-gloved.  Significant levels of stochasticity or loading variance increase after the first contact as can be seen by the whiskers in Figure 4, which on average are larger for un-gloved hands by 16%.

| Predicting transfer efficiency (%) using the ABC method
The ABC algorithm was used to compute transfer efficiency using Equation 1 instead of the linear mixed-effects model, which does not account for vagaries of pathogen transfer mechanisms.
Density histograms of predicted transfer efficiencies (λ G and λ U ), which most closely represent the experimental CFU for all contacts, are plotted in Figure 5. These are used to estimate mean, standard deviation, and 95% CI for gloved and un-gloved contacts and are presented in Table 3. The prior experimental distribution of transfer efficiencies collated from Lopez et al's 19 experimental data used as an initial guess or prior distribution is laid under the posterior predictions. 16 We note that the gloved transfer efficiency tends to be lower than that reported in Table 2 and that the un-gloved version is higher. Table 3 shows the mean, standard deviation, and 95% confi- the 95% confidence interval range. Quantitatively, we note that this represents an average of 11% relative error (min 4%, max 23%) between prediction for the un-gloved cases and a 28% error for the gloved case (min 14%, max 50%).

| D ISCUSS I ON
Between participants, variation appears to be modest in comparison with the effect of a glove barrier, meaning that contact surfaces area, roughness, skin temperature, pH (in un-gloved tests), and con- isms. 39 We also note from our own unpublished pressure data from a separate experiment that participants tend to apply less pressure when not wearing gloves, making these findings potentially "bestcase" scenarios.
Nevertheless, gloves generally afford a consistency in loadings, with a dynamic equilibrium appearing at four contacts. A similar trend was seen for un-gloved hands, although equilibrium was seen after six contacts. This value may shift depending on surface type or porosity and the difference in CFU values between finger and surface. 27 The highest stochasticity is observed from the first contact, as shown in Figure 4, leading to the conclusion that contact with a single contaminated object such as a door handle may have greater variability than repeated contacts, say with a soiled computer keyboard. This reinforces that gloves should be removed after patient care and hands thoroughly washed. While gloves often became less contaminated, they facilitate transfer from fingers to surfaces more readily than un-gloved hands. Although an average of 5% difference between CFU glove loadings and un-gloved fingers may appear small, this could be clinically significant when dealing with pathogens with tiny infectious doses.
It is important to recognize that this laboratory study measured transfer under idealized conditions that possibly differ from real surface contacts in a hospital or other indoor environments including pressure, sheer force, or other ways people manipulate objects.
The current analysis does not include quantification of asymmetric (bidirectional) transfer from fomite to fingertip and vice versa as sampling is destructive; however, it has been found previously that transfer from skin to fomite is often substantially smaller in magnitude, although not negligible. 19,23,33 Transfer efficiency is likely to also be related to inoculum quantity or more specifically to the difference in CFU counts between fomite and finger. 25 Percentage transfer of microorganisms appeared to decrease linearly as inoculum concentration increased, while the concentration gradient between surface and finger dictates the transfer efficiency. 40 Although Oldham's method shows that the effect disappears after multiple contacts, this hypothesis is still visible in Figure 4. The study utilizes an inoculum concentration range of ~5 × 10 7 to 5 × 10 9 /mL, which could be representative of a sample of human fluid waste, 40  were completely dry, that swabbing efficiency was overestimated, that bacterial inactivation was underestimated, or that initial inoculum was actually lower than anticipated. As a result, the ABC method allows all of these variables to be considered simultaneously and a posterior distribution to be given for transfer efficiency. Surfaces, in our experiment, were also deliberately inoculated to have a relatively uniform concentration on each surface, but some coagulation of fluid was still noticeable in the center of the coupon after an hour.
Since similar conditions were used by Lopez et al, 19 it stands to reason that this would have happened in that study, also explaining their similar transfer efficiency rates.
The data enabled evaluation of the statistical properties of CFU loading on fingertips and transfer efficiency for a multiple sequential contacts, and statistical 18,19,23 or risk analysis models 32  3. The ABC method provides a novel, accessible, and flexible method for parameter estimation. Transfer efficiency that best represents this experimental data set, estimated using the ABC model, was higher with bare skin (49%, 95% confidence interval CI = 32%-72%) than gloved hands (30%, CI = 17%-49%), highlighting high variability arising from latent variables. This upholds UK national recommendations for glove use during patient care, especially when fluids are involved 45 ; however, the fit of the gloves was shown to have an important influence on the potential for bacterial contamination especially when in 23% of the cases, a higher loading was found. Glove fit is therefore of crucial importance.

CO N FLI C T O F I NTE R E S T
None to declare.