Extent and Predictors of Microbial Hand Contamination in a Tertiary Care Ophthalmic Outpatient Practice

Abstract

purpose. To measure the extent of microbial hand contamination among ophthalmologists during routine clinic practice and examine its association with hand cleansing practices and beliefs, glove use, and patient load.

methods. This was a single-masked analysis of resident and transient flora of ophthalmologists before and after patient examination and after handwashing by agar imprints of the dominant hand. Standardized questionnaires were used to collect information concerning subjects’ hand cleansing practices and patient load.

results. Of the 108 cultures, 107 (99.1%) were culture positive, yielding 15 separate organisms. Gram-negative bacilli were the most common transient flora, followed by Gram-positive cocci and fungi. Thirty-five (97.2%) ophthalmologists were culture positive for at least one resident and 8 (22.2%) ophthalmologists were culture positive for at least one transient organism, before patient contact. Regression models showed alcohol-based hand rub use, transient and resident floral load before patient contact, and patient load collectively accounted for 58.7% of the variance in resident floral load after patient contact. Use of alcohol-based hand rubs was associated with a mean resident floral reduction of 324.4 CFUs (95% confidence interval [CI] = 185.4 to 463.5; P < 0.01) and 31.6 CFUs (95% CI = 1.2 to 62.0; P < 0.05) after patient contact and handwashing, respectively. Handwashing with chlorhexidine was a significant predictor for transient floral load after handwashing (unstandardized β = –17.2; 95% CI = –10.2 to –24.2; P < 0.01).

conclusions. The extent of contamination with pathogenic organisms after contact with eye outpatients, who have traditionally been perceived as relatively “clean,” was of concern. Previously identified risk factors for hand contamination in inpatient settings, such as patient load, only explained a small proportion of variance in microbial load in the ophthalmic outpatient setting.

Ocular examination frequently requires eyelid eversion, putting the ophthalmologist’s hands in direct contact with potentially infectious body fluids, such as tears and other ocular secretions. Such secretions have been shown to contain many highly infectious viruses, among them adenoviruses ¹ and the newly discovered SARS-associated coronavirus. ² Not infrequently, the attending ophthalmologist is also infected with, ^{3 4} and has been implicated in the subsequent spread of, these viruses, posing an occupational hazard to self in addition to acting as a transmission vector in the course of iatrogenic, nosocomial infection. ⁵ Many reports of nosocomial conjunctivitis outbreaks, ⁶ keratitis with ^{7 8 9} or without ⁹ perforation, and even endophthalmitis ¹⁰ have been published. 

Consequently, there have been renewed efforts to identify preventive strategies against such infectious outbreaks and risks. One of the research foci concerns the nature of hand flora and the potential value of hand cleansing. Most studies thus far in this area, however, have been carried out in traditional “high-risk” inpatient environments, such as intensive care units and pediatric wards. ^{11 12 13 14 15 16 17} Relatively little attention has been paid to the outpatient setting, where most eye care services are based. As a result, the present study was conducted to quantify the extent of microbial hand contamination among ophthalmologists during routine practice and to examine its relationship with potential risk behaviors and environmental factors. 

Materials and Methods

A cross-sectional study was conducted in November 2004 at the outpatient clinic of Hong Kong Eye Hospital, which is a university ophthalmic teaching hospital with a population of approximately 1.6 million out of 6.8 million in Hong Kong. More than 600 outpatients are seen at the hospital daily. Handwashing facilities are available throughout the entire facility, with at least one washbasin in every consultation room, together with plain (non-antimicrobial) soaps, chlorhexidine gluconate solution, and alcohol-based hand rubs. The medical team consists of 3 consultant, 13 fellow, and 20 resident ophthalmologists. Consultation session hours are 9:00 am to 1:00 pm every working day. 

The following definitions will be used throughout the study. Handwashing refers to the use of water, with or without plain soap or a chlorhexidine gluconate solution (4%, Hibiclens; ZeneCA Pharmaceuticals, Wilmington, DE) ¹⁸ ; hand anti-sepsis refers to the application of alcohol-based (60%) hand rubs, the use of which does not require water. Hand cleansing refers to either action. ^{19 20} One consultation refers to a single patient encounter, and consultation session refers to the entire morning’s patient encounters. 

Microbiologic Procedures

To assess the nature of hand flora before and after patient contact and the effectiveness of handwashing, blood agar imprints of the 5 fingertips of the dominant hand ^{17 20 21} were obtained from each participating ophthalmologist at the following times:

1.  
   Immediately before patient contact in the consultation session (culture A);
2.  
   Immediately before the first handwashing episode after patient contact, which might or might not have been after contacting the first patient (culture B);
3.  
   Immediately after the first handwashing episode after patient contact (culture C).

All participating ophthalmologists were allowed to use alcohol-based hand rubs and gloves at his or her own discretion during the consultation session. If gloves were worn, then all 3 cultures were taken with gloves on. All gloves available at the clinic are of nonsterile latex type. The change in hand flora between cultures A and B represented the microbial load acquired after patient contact, whereas the change between cultures B and C was used as a proxy measure for the effectiveness of handwashing. 

All blood agar plates were incubated aerobically at 37°C and supplemented with 5% CO_2. The agar plates were examined daily up to 48 hours. All growths were identified, counted, and tested for antibiotic sensitivities. The maximum colony count was fixed at 600 colony-forming units (CFUs). Beyond this, colonies formed a confluent surface. All growths were classified into resident or transient flora. ¹⁹ Resident flora were defined as microorganisms that rarely cause infection unless introduced into the body by trauma or medical devices. ²² For the purpose of this study, coagulase-negative Staphylococcal species (CNSS), micrococci, Bacillus species, diphtheroids, and Moraxella species were regarded as resident flora. Transient flora were microorganisms that have the potential to cause infection regardless. ¹⁹ Laboratory staff responsible for organism identification and counting were masked to the identity and timing of individual culture plates. Reliability of culture results was evaluated in a pilot study involving 3 of the study team members (RFL, DYLL, and BNML). 

To minimize the Hawthorne phenomenon (the potential bias associated with subjects being aware that they are being observed), an information sheet containing the instructions and the culture plates were given to each ophthalmologist at the time of informed consent. Each ophthalmologist was asked to make the impressions at the appropriate time, without direct observation, and to return the plates in an anonymous manner to a study team member (RFL) after the last culture was taken. All cultures plates were coded by a member of the clinic not involved in the study. The purpose of coding was not to enable individual identification but to link the culture results with the information on the questionnaires. All ophthalmologists were assured of this arrangement and were guaranteed that their individual culture results would be untraceable with this coding method. 

We took two additional measures to ensure that the cultures were taken appropriately. First, immediately before the consultation session, a member of the study team (RFL) personally explained to, and on some occasions showed, each ophthalmologist how the fingertip impressions should be made. Unfortunately, the study team member was not physically present in the room when the cultures were taken because it could induce the Hawthorne phenomenon. Second, when the cultures were returned, each plate was checked macroscopically to ensure that all 5 fingertip impressions were present and of appropriate size before they were sent to the laboratory for culture. All cultures were taken satisfactorily. 

Survey Instrument

All participants were asked to complete a standardized questionnaire immediately after the consultation session. The same coding method was used for the questionnaires. Information collected included the frequency, method, and duration of hand cleansing practices, patient load, glove use during patient care, and self-reported factors, if any, for poor adherence with hand cleansing. ²³  

Statistical Analysis

All analyses were performed using special software (StatLab, SPSS for windows, ver. 12.0; SPSS, Inc., Chicago, IL). Continuous variables were expressed as mean (± SD), and categorical variables were expressed as individual counts and proportions. Univariate analyses to determine the association between baseline demographics, handwashing practices and beliefs, use of alcohol-based hand rubs and gloves, patient load, and microbial load were performed using the Mann-Whitney U test and the Spearman rank correlation test as appropriate. To minimize the potential for type 1 error caused by multiple comparisons, we used P < 0.01 as the critical value to determine statistical significance. We also constructed six stepwise regression models to determine the most robust predictors for resident and transient floral load in cultures A, B, and C. Independent variables were chosen based on empirical ^{14 17 20 23} and statistical (from univariate results) association with the outcome variable. The F probability at entry and removal in the regression models were set at 0.05 and 0.10, respectively. Colinearity diagnostics, using an averaged variance inflation factor >5 as cutoff, were included. There was no evidence of multicollinearity for any of the independent variables included in all six models. The critical value of significance in the regression models was set at P < 0.05. 

The study protocol followed the principles in the Declaration of Helsinki, and informed consent was obtained from all participating ophthalmologists. The Ethics Committee of the Chinese University of Hong Kong approved the study protocol. 

Results

All ophthalmologists (n = 36) working at the outpatient clinic agreed to participate. Baseline demographics, hand cleansing practices and beliefs, and patient load are summarized in Table 1 . No significant gender differences were present except for the qualification of the participating ophthalmologist, likely due to a historical cohort effect. 

Of the 108 cultures (36 ophthalmologists × 3 culture samples each), 107 (99.1%) were culture positive, yielding 15 separate organisms. Five were residual flora and the rest were transient flora. Transient flora were further classified according to morphology (Table 2) . 

CNSS were the most common resident flora isolated, followed by micrococci, Bacillus species, diphtheroids, and Moraxella species (Fig. 1) . CNSS were present on 88.9% of hands before patient contact (culture A), 97.2% of hands before first handwashing episode after patient contact (culture B), and 63.9% of hands after handwashing (culture C). Gram-negative bacilli were the most common transient flora in all 3 cultures, followed by Gram-positive cocci and fungi. Thirty-five (97.2%) ophthalmologists were culture positive for at least one resident and 8 (22.2%) ophthalmologists were culture positive for at least one transient flora, before patient contact. As expected, most organisms showed an increase in prevalence after patient contact (culture B versus culture A) and a decrease in prevalence after handwashing (culture C versus culture B). 

Individual organism counts are summarized in Table 3 . Hands were most heavily colonized by CNSS, with a mean ± SD count of 116.8 ± 200.7, 223.0 ± 241.9, and 20.6 ± 38.2 CFUs in cultures A, B, and C, respectively. Contamination by transient flora was minimal before patient contact. Transient floral load as a proportion of total floral load increased from 1.4% to 9.5% after patient contact and decreased to 2.1% after handwashing. 

Univariate Analyses

Significant associations (P < 0.01) were found between the resident floral load in culture B and both alcohol-based hand rub use and transient floral load in cultures A and B. The mean resident floral load was less (mean difference = 354.2 CFUs; P < 0.01) for alcohol-based hand rub users than for nonusers after patient contact. The resident floral load in culture B was also strongly associated with the resident floral load in culture C (ρ = 0.46; P < 0.01). Self-perception of adequacy of handwashing and patient load only showed a weak positive association with resident floral load in culture B (both P < 0.05). No significant associations were present between the baseline demographic predictors and the microbial load in any of the cultures. 

Multivariate Analyses

Stepwise multiple linear regression models were constructed to determine the most robust predictors for the resident and transient floral load in cultures A, B, and C. None of the demographic variables, hand cleansing practices, or beliefs significantly explained the variances in resident and transient floral load in cultures A. On the other hand, alcohol-based hand rub use, transient and resident floral load in culture A, and patient load (between cultures A and B) combined, accounted for 58.7% of variance in resident floral load in culture B. Of high significance (P < 0.01) was the impact of use of alcohol-based hand rubs, explaining 30.7% of the variance and a mean resident floral reduction of 324.4 CFUs (95% CI = 185.4 to 463.5) in culture B after adjusting for other predictors. 

Resident floral load in culture B also significantly predicted the resident floral load in culture C (unstandardized β = 0.1; 95% CI = 0.0 to 0.1; P < 0.01). It alone explained 18.3% of the variance in resident floral load after handwashing and another 7.6% with alcohol-based hand rub use (unstandardized β = –31.6; 95% CI = –1.2 to –62.0; P < 0.05). None of the handwashing predictors (with water ± soap or chlorhexidine gluconate) significantly explained the variance in the resident floral load in culture C. 

Total variances of transient floral load explained by the predictors were low compared with those of resident floral load. A weakly significant positive correlation (unstandardized β = 0.1; 95% CI = 0.0 to 0.3; P < 0.05) was present between resident and transient floral load in culture B, whereas a more significant negative correlation (unstandardized β = –17.2; 95% CI = –10.2 to –24.2; P < 0.01) was present between handwashing with chlorhexidine gluconate and transient floral load in culture C. 

Discussion

Our results showed that although resident flora dominated all 3 cultures, some clearly pathogenic organisms, such as methicillin-resistant Staphylococcus aureus (MRSA) and Enterobacter species, were isolated after contacts with eye patients, who have been traditionally regarded as relatively clean of infection. 

Microbial contamination of hands has been described as a dynamic process that results from multiple factors. ²⁰ Our stepwise multiple linear regressions ranked the relative importance of some predictors of microbial load after patient contact. Not unexpectedly, patient load was a significant predictor of resident floral load after patient contact. ^{17 20 24} It was ranked fourth, however, after alcohol-based hand rub, transient floral load, and resident floral load before patient contact. Of high significance (P < 0.01) was the impact of use of alcohol-based hand rubs, explaining 30.7% of variance and a mean resident floral reduction of 324.4 CFUs (95% CI = 185.4 to 463.5) after adjusting for other predictors. This finding is also supported by previous studies, ^{19 20 21} with one reporting a mean reduction of 52 CFUs in bacterial load in healthcare workers who routinely used alcohol-based hand rubs. ²⁰  

We were unable to identify any significant predictors for transient and resident floral load before patient contact. In univariate and multivariate analyses, none of the demographic variables, hand cleansing practices, or beliefs significantly explained the floral load before patient contact. Their high correlations with resident floral load after patient contact suggest that hands heavily contaminated with microbes before patient contact tended to remain heavily contaminated after patient contact, whereas hands that are relatively free of contamination tended to remain relatively contamination free, even adjusting for other predictors such as patient load. Further studies are needed to explore the determinants of colonization by these organisms. 

Our results showed that handwashing can reduce the mean resident and transient load by at least an order of magnitude (Table 3) . Linear regression models suggested that resident flora before handwashing alone explained 18.3% of the variance of resident floral load after handwashing and another 7.6% with alcohol-based hand rub use. Not surprisingly, none of the handwashing predictors (with water ± soap or chlorhexidine gluconate) significantly explained the variance in the resident floral load after handwashing. Resident flora are usually attached to deeper layers of skin and consequently are more resistant to elimination. ²⁵ Therefore, it appears that alcohol-based hand rubs may be more effective in removing resident flora than any of these methods of handwashing, ^{20 26} and our results suggests that their effectiveness may persist even after handwashing (unstandardized β = –31.6; 95% CI = 1.2 to 62.0; P < 0.05). 

Handwashing with chlorhexidine gluconate was significantly associated with the transient floral load after handwashing (unstandardized β = –17.2; 95% CI = –10.2 to –24.2; P < 0.01). Chlorhexidine gluconate is frequently recommended because of its broad antimicrobial spectrum and low toxicity. ¹⁴ Although the bactericidal activity of chlorhexidine gluconate is not as rapid as that of alcohol, previous studies reported significant flora reductions after approximately 15 seconds of hand washing ²⁷ and residual activity of >6 hours. ²⁸ Unfortunately, we do not have chlorhexidine-based hand rubs at our clinic. It would be highly desirable to compare the effectiveness of these time-efficient hand rubs ²⁹ in busy ophthalmic outpatient settings. In our practice, hand antisepsis with alcohol-based hand rubs and handwashing with chlorhexidine gluconate appear to be most effective at reducing resident and transient flora. 

Our study had certain limitations. First, there was no randomization for different hand-cleansing practices. Given that the aim of the study was to describe the dynamics of microbial hand contamination during routine clinic practice, we tried to mimic the actual clinical setting as much as possible. The ethical acceptability of control groups in situations that may be threatening to the ophthalmologist and to other patients (e.g., after contacting ocular secretions of patients with conjunctivitis) poses another obstacle. Second, the use of self-administered agar gel and questionnaires with the aim to minimize the Hawthorn phenomenon might have led to performance or reporting bias. The magnitude of these biases would depend on how much each ophthalmologist was concerned with the stigma of being labeled as dirty in case his or her cultures grew spurious or high loads of organisms. We specifically designed an anonymous coding system to alleviate such concerns. Third, it is difficult to translate the present findings into clinical impact and infectious risk. We were unable to estimate nosocomial infection rates and their relationships with microbial hand contamination because such a task would require a much larger sample size with prospective follow-up, processes that were beyond the scope of the study. In the paradigm of evidence-based medicine, our results would only constitute disease-orientated rather than patient-orientated evidence. ³⁰ Nonetheless, 8 of the 13 bacteria, including all 5 resident flora, isolated in the present study have been implicated as the sole causative agent of bacterial keratitis in our country ³¹ and of conjunctivitis outbreaks in other parts of the world. ^{32 33 34} Findings from this study should prompt further study in this often neglected area, ultimately leading to implementation of enhanced infection control measures and, it is hoped, to the successful prevention of nosocomial infections in an ocular care setting. 

In conclusion, the nature and load of pathogenic organisms isolated after contact with relatively clean eye patients were of concern. Well-recognized factors for hand contamination in inpatient settings, such as patient load, only explained a minority of variance in microbial load in our clinic sample. Instead, the microbial load already present on the hands before patient contact appears to have significant influence on the microbial load present after patient contact. Further studies are warranted to explore the determinants of these microbes. 

 Submitted for publication February 17, 2005; revised April 26 and June 10, 2005; accepted August 10, 2005.

 Disclosure: R.F. Lam, None; M. Hui, None; D.Y.L. Leung, None; V.C.Y. Chow, None; B.N.M. Lam, None; G.M. Leung, None; D.S.C. Lam, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Table 1.

 

Baseline Demographics, Hand Cleansing Practices and Beliefs, and Patient Load of Participating Ophthalmologists

	Sex					Overall
		Male (n = 23) (63.9%)	Female (n = 13) (36.1%)	P (2-tailed)	n = 36	Range (%)
Year(s) in ophthalmology*	9.0 ± 7.9	5.6 ± 6.9	0.09, †	7.8 ± 7.7	0–29
Qualification of participating ophthalmologist
Resident	10	11	0.03, ‡	21	(58.3)
Consultant and fellow	13	2		15	(41.7)
Duration of hand-washing episode between cultures Band C (sec)*	20.2 ± 15.3	12.7 ± 6.7	0.17, †	17.5 ± 13.3	5–60
Self-perception of adequacy of hand washing
Insufficient	17	12	0.38, ‡	29	(80.6)
Sufficient	6	1		7	(19.4)
No. of hand-washing episodes per consultation session*	4.7 ± 4.3	4.0 ± 3.1	0.75, †	4.4 ± 3.9	1–20
No. of hand-cleansing episodes per consultation session*	8.5 ± 9.7	7.7 ± 9.2	0.96, †	8.2 ± 9.4	1–40
Alcohol-based hand rub use
No	15	10	0.11, ‡	25	(69.4)
Yes	8	3		11	(30.6)
Glove use
No	19	7	0.12, ‡	26	(72.2)
Yes	4	6		10	(27.8)
Patient load between cultures A and B (no. of patients)*	13.1 ± 12.7	20.1 ± 15.9	0.25, †	15.6 ± 14.1	1–50
Patient load for consultation session (no. of patients)*	30.1 ± 15.9	39.1 ± 16.3	0.13, †	33.3 ± 16.4	5–55
Patient load for consultation session/total no. of hand-washing episodes*	12.5 ± 12.6	17.2 ± 16.6	0.36, †	14.2 ± 14.1	1–52
Patient load for consultation session/total no. of hand-cleansing episodes*	10.5 ± 13.0	12.5 ± 14.1	0.45, †	11.2 ± 13.3	1–52

*  Mean ± SD.

†  Mann-Whitney U test.

‡  Fisher exact test.

Table 2.

 

Morphologic Classification of Cultured Organisms

Classification	Morphology	Cultured Organisms
Resident flora		Coagulase-negative Staphylococcal species
		Bacillus species
		Micrococcus species
		Diphtheroids
		Moraxella species
Transient flora	Gram-negative bacilli	Enterobacter aerogenes
		Sphingomonas paucimobilis
		Brevundimonas vesicularis
		Weeksella virosa
		Acinetobacter species
	Gram-positive cocci	Methicillin-resistant Staphylococcus aureus
		Methicillin-sensitive Staphylococcus aureus
		Viridans strepococci
	Fungi	Aspergillus fumigatus
		Penicillium species

Figure 1.

 

Prevalence of positive cultures.

Table 3.

 

Microbial Load for Cultures A, B, and C

Organism	Microbial Load (CFUs)
		Culture A			Culture B			Culture C
		Mean ± SD	Range	Mean ± SD	Range	Mean ± SD	Range
Resident flora
CNSS	116.8 ± 200.7	0–600	223.0 ± 241.9	0–600	20.6 ± 38.2	0–160
Micrococcus species	7.4 ± 13.6	0–75	26.0 ± 99.1	0–600	1.1 ± 3.6	0–15
Bacillus species	17.3 ± 99.9	0–600	0.6 ± 1.7	0–10	0.5 ± 1.2	0–5
Diphtheroids	3.6 ± 14.9	0–88	14.3 ± 48.4	0–279	1.0 ± 5.8	0–35
Moraxella species	0.1 ± 0.5	0–3	0.0 ± 0.0	0	0.0 ± 0.0	0
Transient flora
Gram-negative bacilli	0.7 ± 3.2	0–19	2.7 ± 8.6	0–40	0.4 ± 1.6	0–9
Gram-positive cocci	1.3 ± 5.7	0–34	22.8 ± 104.4	0–600	0.2 ± 0.3	0–2
Fungi	0.0 ± 0.0	0	0.0 ± 0.2	0–1	0.1 ± 0.2	0–1
Total flora
Resident	145.1 ± 259.7 (98.6%)*	0–1209	263.9 ± 289.4 (90.5%)*	0–1200	23.3 ± 39.4 (97.9%)*	0–160
Transient	2.1 ± 6.6 (1.4%)*	0–34	27.6 ± 104.5 (9.5%)*	0–600	0.5 ± 1.6 (2.1%)*	0–9

 Ranges are truncated at 600 CFUs for individual counts.

*  Proportion of total floral load.

The authors thank Alison Kan for her assistance in organism identification and quantification and Keith Tin and Irene Wong for their statistical support.