ABSTRACT
Hand hygiene interventions are critical for reducing farmworker hand contamination and preventing the spread of produce-associated illness. Hand hygiene effectiveness may be produce-commodity specific, which could influence implementation strategies. This study’s goal was to determine if produce commodity influences the ability of handwashing with soap and water or two-step alcohol-based hand sanitizer (ABHS) interventions to reduce soil and bacteria on farmworker hands. Farmworkers (n = 326) harvested produce (cantaloupe, jalapeño, and tomato) for 30 to 90 minutes before engaging in handwashing, two-step ABHS (jalapeño and cantaloupe), or no hand hygiene. Hands were rinsed to measure amounts of soil (absorbance at 600 nm) and indicator bacteria (coliforms, Enterococcus sp., generic Escherichia coli, and Bacteroidales universal [AllBac] and human-specific [BFD] 16S rRNA gene markers). Without hand hygiene, bacterial concentrations (0.88 to 5.1 log10 CFU/hand) on hands significantly differed by the produce commodity harvested. Moderate significant correlations (ρ = −0.41 to 0.56) between soil load and bacterial concentrations were observed. There were significant produce-commodity-specific differences in the ability of handwashing and two-step ABHS interventions to reduce soil (P < 0.0001), coliforms (P = 0.002), and Enterococcus sp. (P = 0.003), but not the Bacteroidales markers AllBac (P = 0.4) or BFD (P = 0.3). Contamination on hands of farmworkers who harvested cantaloupe was more difficult to remove. Overall, we found that a two-step ABHS intervention was similar to handwashing with soap and water at reducing bacteria on farmworker hands. In summary, produce commodity type should be considered when developing hand hygiene interventions on farms.
IMPORTANCE This study demonstrated that the type of produce commodity handled influences the ability of handwashing with soap and water or a two-step alcohol-based hand sanitizer (ABHS) intervention to reduce soil and bacterial hand contamination. Handwashing with soap and water, as recommended by the FDA’s Produce Safety Rule, when tested in three agricultural environments, does not always reduce bacterial loads. Consistent with past results, we found that the two-step ABHS method performed similarly to handwashing with soap and water but also does not always reduce bacterial loads in these contexts. Given the ease of use of the two-step ABHS method, which may increase compliance, the two-step ABHS method should be further evaluated and possibly considered for implementation in the agricultural environment. Taken together, these results provide important information on hand hygiene effectiveness in three agricultural contexts.
INTRODUCTION
Between 2004 and 2013 in the United States, contaminated fresh produce caused at least 36% of all reported foodborne illnesses (629 outbreaks, 19,932 cases) (1). Produce may become contaminated at multiple points from farm to fork (2). Farmworker hands are an important vehicle for pathogen contamination in the harvest and postharvest environments (3–5). Thus, the U.S. Food and Drug Administration (FDA)’s Food Safety Modernization Act, Final Rule on Produce Safety requires that personnel wash their hands with soap and water (6). Compliance with the FDA’s Produce Safety Rule can be difficult; farmworkers may have limited access to running water and convenient handwashing facilities in the field, and incentives for hand hygiene may conflict with those for productivity (7). Alcohol-based hand sanitizers (ABHSs) have been proposed as an alternative in the health care industry (8, 9), but the FDA’s Produce Safety Rule clearly states that “you may not use antiseptic hand rubs as a substitute for soap (or other effective surfactant) and water” because they are ineffective at removing bacteria when dirt, grease, and oil are present on hands (6). An ABHS-based method that may address this limitation is SaniTwice, a two-step technique where an excess of ABHS is applied to the hand and removed by a paper towel, followed by a second ABHS application (10). Previous studies showed that the SaniTwice (two-step) method was more effective at removing bacterial contamination than traditional (one-step) ABHS methods and that SaniTwice was either equivalent to or better than handwashing with soap and water at removing bacterial contamination from farmworker hands (10–12). The handwashing facilities on the farms from these studies were a few kilometers, an estimated 10- to 30-minute walk, from the harvesting fields (3), which may be a barrier to handwashing compliance. The SaniTwice method may improve hand hygiene compliance because this method takes a similar amount of time (∼35 to 45 seconds) as handwashing and can be completed on the spot (e.g., in the field while harvesting produce).
Unfortunately, there is limited evidence on the effectiveness of hand hygiene interventions, including both traditional handwashing with soap and water and two-step ABHS, in the agricultural environment. Most of these studies have been conducted in health care or community settings (8, 9, 13, 14), which may not accurately reflect agricultural settings. Agricultural workers’ hands likely differ from health care provider hands (15, 16), and even in the health care setting, worker hand characteristics are likely to vary substantially depending on subspecialty and tasks performed. Therefore, it is plausible that the degree of soil and microbial contamination on farmworker hands and the effectiveness of hand hygiene interventions will differ depending on the produce commodity harvested. We previously showed that the hands of farmworkers harvesting cantaloupe, compared with those of workers harvesting jalapeño and tomato, harbored higher concentrations of bacterial indicators and somatic coliphages (17).
Only two studies, both conducted by our group on jalapeño and tomato farms in Northern Mexico, have evaluated the effectiveness of hand hygiene interventions (handwashing with soap and water versus two-step ABHS) in the agricultural setting (11, 12). Both studies suggested that handwashing (∼1.4 log10 reduction) and two-step ABHS, (∼ 0.4 log10 reduction), compared with the control groups, were statistically associated with reductions of soil on farmworker hands (P < 0.05). Only the tomato study (11), but not the jalapeño study (12), showed that handwashing, compared with the control group, was effective at removing some bacterial indicators (Enterococcus sp. [1.3 log10 reduction] and Escherichia coli [0.2 log10 reduction]). The two-step ABHS method resulted in lower concentrations of some, but not all, bacterial indicators in both studies as compared to controls. Descriptively, there seems to be produce-commodity-specific differences in the ability of handwashing and two-step ABHS interventions to reduce bacterial contamination on farmworker hands, but there is a need to expand to other produce commodities and to statistically compare results.
Bacterial indicators, such fecal coliforms, generic E. coli, and Enterococcus sp., are typically used to measure fecal contamination in the agricultural environment, which, when present, indicate a higher risk of fecally transmitted pathogens (18). More recently, Bacteroidales 16S rRNA genetic markers have been used in microbial source tracking because they are specifically found in the gut of warm-blooded animals and do not propagate in the environment (19, 20). When multiple bacterial indicators are assessed, which is typical for many research studies (18), multivariable modeling techniques can be helpful in identifying trends and relationships among the data (21). The goal of this study was to determine if the effectiveness of handwashing with soap and water and two-step ABHS interventions differs depending on the produce commodity harvested using both uni- and multivariable statistical approaches.
RESULTS
Characteristics of study population and perceptions on hand hygiene.Three hand hygiene studies were conducted on cantaloupe (n = 129), jalapeño (n = 120), and tomato (n = 77) farms. As described in the Materials and Methods, all three studies contributed control and handwashing data, but only jalapeño and cantaloupe studies contributed two-step ABHS data. Farmworker characteristics (age and gender) and harvest time (number of minutes workers spent conducting harvesting activities) for the three studies are presented in Table 1. Harvest time was significantly different between studies in each group (control, handwashing [P < 0.0001] and two-step ABHS [P < 0.01]), while harvester age was only significantly different between cantaloupe and jalapeño studies in the handwashing group (P < 0.01). None of these variables were strongly correlated with the concentration of soil or bacterial indicators nor did they confound the relationship between produce commodity and hand hygiene (data not shown); thus, they were not controlled for in subsequent analysis.
Demographic characteristics of farmworkers in control and intervention groups harvesting produce in Nuevo León, Mexico
Thirty-nine workers in cantaloupe farms completed an optional survey assessing their perceptions on hand hygiene interventions. Most respondents reported that they believed that hand hygiene improves fresh produce quality (82%) and worker health (67%). When asked about which method they preferred using, responses varied; 33% preferred handwashing with soap and water, 13% preferred ABHS, 33% said both, and 21% did not provide a response. Most respondents (85%) expressed a desire to continue performing hand hygiene interventions at work. Responses for how many times participants thought hand hygiene should be performed during an 8-hour work shift varied between once and as often as every half hour or hour, with the most common answer being 3 times per shift (49%).
Comparison of hand contamination in the absence of hand hygiene interventions between cantaloupe, jalapeño, and tomato farmworkers. (i) Concentrations of soil and bacterial indicators.Farmworker hand contamination represented by soil amounts (optical density at 600 nm [OD600]), bacterial indicator concentrations (coliforms, Enterococcus sp., and generic E. coli), and concentrations of the Bacteroidales 16S rRNA markers AllBac and BFD (tomato data not collected) were compared for the three produce commodities in the absence of hand hygiene interventions (i.e., control groups). OD600 values were used as a proxy for the amount of soil in hand rinsate samples and are proportional to the amount of particulate matter in the sample. The concentrations of soil and some bacterial indicators on hands significantly differed across produce commodities (Fig. 1). The hands of tomato harvesters had the highest soil loads (median OD600, 0.48) compared with those of jalapeño (median OD600, 0.29; P < 0.001) and cantaloupe (median OD600, 0.11; P < 0.0001) workers (Fig. 1). There were not, however, any differences in coliform concentrations between cantaloupe, jalapeño, or tomato harvester hands. In contrast, cantaloupe harvester hands had 1.2 log10 higher CFU/hand concentrations of Enterococcus sp. than jalapeño harvester hands (P = 0.004) but had levels equivalent to those found on tomato harvester hands. Cantaloupe harvester hands harbored the highest E. coli concentrations of the three produce commodities (1.4 log10 higher CFU/hand concentrations than jalapeño and 1.0 log10 higher CFU/hand concentrations than tomato farmworker hands) and had higher concentrations of the universal Bacteroidales AllBac (2.4 log10 higher genomic equivalent copies [GEC]/hand concentrations) and the human-specific Bacteroidales BFD markers (4.1 log10 higher GEC/hand concentrations) than the jalapeño harvester hands. Tomato farmworker hands were not assessed for Bacteroidales markers.
Soil and bacterial indicator levels on farmworker control group hands collected from farms producing cantaloupe, jalapeños, and tomatoes. The Steele-Dwass test for multiple comparisons was employed to compare soil levels (log10 OD600) (A) and the concentrations of coliforms (B), Enterococcus sp. (C), E. coli (log10 CFU/hand) (D), and AllBac (E) and BFD Bacteroidales markers (F) (log10 GEC/hand) in farmworker control group hand rinsate samples between different produce commodities harvested. Bar graphs display the quartiles (25th, 50th, and 75th), and whiskers extend to 1.5 times the interquartile range. Any data points outside the whiskers are displayed individually as dots. Statistically significant differences (P < 0.05) between produce are indicated with an asterisk.
(ii) Correlations between soil and microbes.There were significant positive correlations between soil and bacterial indicators on farmworker hands in the absence of a hand hygiene intervention (i.e., control groups) (Fig. 2). There was no clear pattern across produce commodities, and the correlations observed were only moderate, not strong associations (see Materials and Methods and figure legends for definition)(Fig. 2). Interestingly, the only significant correlation between Bacteroidales markers and other indicator bacteria was a negative correlation of BFD with Enterococcus sp. (ρ = −0.41) on the hands of jalapeño farmworkers, but not those of cantaloupe workers (tomato farmworker hands not assayed).
From control groups, heatmap graphs from the Spearman’s rank correlation results between soil and bacterial indicator concentrations on workers’ hands on farms producing cantaloupes (A), jalapeños (B), and tomatoes (C). Colors represent Spearman’s rank ρ values, as indicated in the legend. Correlations with P values of <0.05 are indicated with an asterisk. Correlations were classified by the strength of the association (Spearman’s ρ value), namely, weak (0.00 to ±0.30), moderate (0.31 to 0.7 or −0.31 to −0.7), or strong (>0.71, or ≤0.71) (45).
(iii) Hand contamination profiles.We sought to define distinct hand contamination profiles using a principal-component analysis (PCA) that incorporated the following log10-transformed variables: soil (OD600), coliforms, E. coli, and Enterococcus sp. Both principal component 1 (PC 1) and PC 2 were significantly associated with the produce commodity harvested (Fig. 3A and B). Hands of tomato, jalapeño, and cantaloupe workers exhibited distinct PC 1 and PC 2 clustering, with hands of cantaloupe harvesters exhibiting the greatest separation from hands of tomato and jalapeño workers (Fig. 3C). Because the hands of cantaloupe farmworkers had the lowest PC 2 values, this indicated that cantaloupe farmworker hands harbored the lowest amount of soil but still had moderate levels of indicator bacteria, as evidenced by their PC 1 vales. Taken together, these results highlight that farmworker hand contamination profiles in the absence of hand hygiene interventions differ by the produce commodity harvested.
Principal-component analysis (PCA) based on measures of contamination on worker hand rinsate samples from control groups that were evaluated in all three hand hygiene intervention studies (OD600, coliforms, Enterococcus sp., and E. coli [n = 124]). (A, B) Bar graphs display the quartiles (25th, 50th, and 75th), and whiskers extend to 1.5 times the interquartile range. Any data points outside the whiskers are displayed individually as dots. Statistically significant differences in PC 1 and PC 2 scores between produce commodities as assessed using the Steele-Dwass test for multiple comparisons are indicated with an asterisk (P < 0.05). (C) Principal component 1 (PC 1) and PC 2 scores are depicted in a two-dimensional scatterplot for each farmworker hand rinsate sample across cantaloupe (blue, n = 42), jalapeño (green, n = 40), and tomato (red, n = 42) control groups. PC 1 (x axis) describes the greatest variation in the data set (39%), while PC 2 (y axis) describes the second greatest variation in the data (26%).
Impact of produce commodity on the effect of hand hygiene interventions.We found that the effectiveness of hand hygiene interventions on reducing soil and bacterial indicators was dependent on the produce commodity being harvested, as indicated by the F statistic and P values shown below the graphs in Fig. 4. Handwashing was more effective at removing soil on the hands of jalapeño (1.4 log10 reduction, P = 0.002) and tomato (1.4 log10 reduction, P = 0.0004) farmworkers than those of cantaloupe workers (Fig. 4A) (1.0 log10 reduction). Handwashing was more effective at reducing Enterococcus concentrations on tomato (1.6 log10 reduction) than that on jalapeño (0.10 log10 reduction, P = 0.002) and cantaloupe (0.11 log10 reduction, P = 0.002) farmworker hands (Fig. 4C). The two-step ABHS intervention was more effective at reducing coliform concentrations on jalapeño farmworker hands (1.8 log10 reduction) than reducing that on cantaloupe farmworker hands (0.26 log10 reduction, P = 0.004) (Fig. 4B). There were no produce-commodity-specific differences in the effectiveness of either the handwashing or the two-step ABHS intervention on Bacteroidales marker concentrations (Fig. 4D and E).
The effect of hand hygiene interventions on farmworker hand contamination by produce commodity harvested. Two-way fixed-effects models were used to assess statistically significant interactions between study groups (hand hygiene interventions [hand washing and two-step ABHS] and control groups) and produce commodities (cantaloupe, jalapeño, and tomato). Plots depict the log10 difference between interventions and control groups for soil OD600 (A), coliforms (B), Enterococcus sp. (C), Bacteroidales AllBac (D), and Bacteroidales BFD (E) and the differences for PC 1 scores (F) and PC 2 scores (G). Bar graphs delineate the least-squares (LS) means difference and the standard error derived from fixed-effect model estimates. The fixed-effect F statistic and P values (α = 0.05) for the interaction between produce commodity and log10 differences between intervention and control groups are shown below each plot. Statistically significant differences (P < 0.05) in LS means differences (control versus intervention group) between produce commodities harvested are indicated with an asterisk.
The interaction between produce commodity and intervention group was statistically significant for PC 2 scores—the composite PC score most associated with differentiating soil versus microbes (Fig. 4G) (F statistic, 11; P < 0.0001)—but not for PC 1 scores (Fig. 4F). There was a significantly higher PC 2 score difference, with the control, on the hands of jalapeño (1.7 log10, P = 0.006) and tomato (1.6 log10, P = 0.02) farmworkers than cantaloupe workers (1.2 log10) performing the handwashing intervention. For the two-step ABHS intervention, PC 2 score differences, with the control, were greater for the cantaloupe (0.53 log10) than the jalapeño (0.04 log10, P = 0.005) group (Fig. 4G).
We then used the fixed-effects model to compare the levels and proportions of soil and bacterial indicators across intervention and control groups within produce commodities. Handwashing resulted in significantly lower soil levels than two-step ABHS, and both interventions yielded lower soil levels than controls on farmworker hands for the produce commodities analyzed (Table 2). However, handwashing was not significantly different than controls for most bacterial indicators, except for lower Enterococcus sp. on tomato farmworker hands (3.7 log10 [handwashing] versus 5.4 log10 [control] CFU/hand, P < 0.0001). For both PC 1 and PC 2 scores, handwashing resulted in significantly lower scores than the control group for all three produce commodities.
Concentration and proportions of positive hand rinsate samples for soil and indicator bacteria from control and intervention groups harvesting cantaloupe, jalapeño, and tomato on farms in Nuevo León, Mexico
The two-step ABHS intervention resulted in significantly lower coliform concentrations than controls on jalapeño farmworker hands (1.5 log10 [two-step ABHS] versus 3.3 log10 [control] CFU/hand, P < 0.0001). Furthermore, the two-step ABHS intervention resulted in significantly lower coliform concentrations than the handwashing intervention on jalapeño farmworker hands (1.5 log10 [two-step ABHS] versus 2.8 log10 [handwashing] CFU/hand, P = 0.02). The two-step ABHS group resulted in significantly lower PC 1 scores on jalapeño farmworker hands than the control group. For PC 2 scores, the two-step ABHS group had significantly lower scores than the control group on cantaloupe farmworker hands.
There were no differences in the proportion of positive samples between control, handwashing, and two-step ABHS for bacterial indicators, except for a lower significant proportion of coliform-positive samples in the two-step ABHS group than both the control and handwashing groups on jalapeño farmworker hands. In summary, the handwashing intervention was more effective at reducing soil than the two-step ABHS intervention. Both the two-step ABHS intervention and handwashing intervention performed similarly at reducing bacteria on hands. The degree to which both handwashing and two-step ABHS reduced bacterial indicators compared with the control group varied across produce commodities. Importantly, neither the handwashing nor the two-step ABHS intervention was effective at removing bacterial indicators on cantaloupe farmworker hands.
DISCUSSION
This study demonstrated that soil and bacterial contamination on farmworker hands, in the absence of a hand hygiene intervention, differed by the produce commodity harvested. Moreover, results showed that the produce commodity harvested influenced the ability of handwashing with soap and water or two-step ABHS interventions to reduce this hand contamination. Last, the two-step ABHS intervention and handwashing intervention performed similarly at reducing bacteria on hands.
We found that soil and bacterial contamination on farmworker hands differed by the produce commodity harvested (tomato, jalapeño, and cantaloupe). The composition of soil and microbes that farmworker hands are exposed to may be influenced by the type of soil, harvest practices, and contact with irrigation water, which may differ depending on what type of produce commodity is being harvested (5, 22–24). Our results, consistent with a previous study (17), found that farmworkers who harvested cantaloupes harbored higher levels of bacterial indicators on their hands than those who harvested jalapeño and cantaloupe.
The degree of soil contamination or the concentration of each bacterial indicator type on farmworker hands, in the absence of intervention, were not proxies of one another. Similar to our previous findings (23), soil level was positively correlated with some bacterial indicators, but not all. Similar to this study, previous research on the correlations between bacterial indicators has been contradictory (25). Furthermore, Ravaliya et al. found no statistically significant relationships between Bacteroidales and generic E. coli in produce, irrigation water, and farmworker hand rinsates obtained from produce farms in Northern Mexico (4), which is in line with the findings here. The lack of clear relationships between bacterial indicators suggests that the measurement of hand contamination is complex and that one metric is not sufficient for all settings. By using PCA to combine the soil and bacterial indicators into a single latent variable, we demonstrated that hand contamination profiles were different among the three produce commodities harvested.
Taken together, our results showed that the soil and bacterial loads, correlation patterns between indicators, and contamination profiles on farmworker hands differed by the produce commodity harvested. It is plausible that different farm environments might favor various compositions of microbial species, which could explain the inconsistencies in bacterial indicator relationships between studies and produce commodity. Previous studies have shown that contamination of fresh produce can be influenced by geographical region, farm management practices, postharvest processing, seasonality, climate, and the source of irrigation water, which can all be produce specific (5, 22–24). For example, cantaloupes are grown close to the ground, are harvested by hand, are propagated during warmer seasons, and have surface characteristics that can aid in microbial attachment and survival (22, 26–28). These factors can influence the transfer of microbes onto farmworker hands and may explain why cantaloupe farmworker hands harbored increased concentrations of Enterococcus sp., E. coli, and Bacteroidales compared with jalapeño and tomato farmworker hands. Even though tomatoes in this study were grown under protective cover in a semienclosed system with a shade cloth, as opposed to being grown in the open air as were cantaloupes and jalapeños, tomato farmworker hands harbored the highest levels of soil.
Our multivariable model results suggested that the effect of hand hygiene interventions was influenced by the produce commodity harvested. One of the major differences observed was that hand rinsates from cantaloupe harvesters, compared with those from tomato and jalapeño harvesters, had the lowest log10 differences in soil (handwashing versus control), Enterococcus sp. concentrations (handwashing versus control), and coliform concentrations (two-step ABHS versus control). In fact, bacterial indicator concentrations on cantaloupe farmworker hands were no different for the handwashing or the two-step ABHS intervention groups compared with the control. We hypothesize that the amount and types of soil and microbes present on farmworker hands prior to intervention might influence the efficacy of different hand hygiene methods at reducing this contamination. This hypothesis should be evaluated in future studies.
There are several mechanisms by which produce-specific differences in hand rinsate profiles could influence hand hygiene interventions. One mechanism may be that specific produce commodities, when harvested, result in higher levels of soil and organic matter, which has been shown to inhibit the effectiveness of hand hygiene interventions (10). Our results, however, argue that a high degree of soil contamination on farmworker hands is likely not the only factor that impedes hand hygiene intervention effectiveness. Even when soil levels were low, as was the case for cantaloupe farmworker hands, there was, relative to other commodities, low hand hygiene effectiveness. A second mechanism could be differences in harvesting practices, such as glove or knife use. For example, glove use may result in different amounts of soil and bacterial concentrations transferred between hands and produce harvested (29). The Produce Safety Rule does not require the use of gloves but does require that if gloves are used they must be in an intact and sanitary condition (6). Although all produce was harvested by hand in this study, tomato and jalapeño, but not cantaloupe, farmworkers used knives to cut stalks, and this may have further contributed to differences in hand contamination profiles and hand hygiene effectiveness. A third mechanism explaining the differences in hand hygiene effectiveness may be due to the differing ability of nonantimicrobial soaps and ABHS to inactivate microbes (30). A fourth mechanism may be the degree of skin roughness or calluses (15, 16) and other physical hand characteristics (e.g., temperature and moisture) that may affect the attachment of different bacterial species to hands and, in turn, the efficacy of microbial removal by hand hygiene products. Last, another possibility may be that harvesting different produce commodities results in a different number of touch events within the same harvest period, which could influence hand contamination levels and in turn hand hygiene effectiveness.
Across all produce commodities, both the handwashing and two-step ABHS interventions were found to be effective at removing soil, with handwashing being more effective than two-step ABHS. This result is consistent with that of previous studies and is likely due to the fact that soap acts in part as an emulsifier, absorbing and suspending dirt particles on contaminated hands (11, 12, 31, 32). The finding that the two-step ABHS intervention also removed soil from hands irrespective of produce commodity is contrary to what has been observed for standard (i.e., one-step) ABHS methods (33, 34), which are not intended to remove solids. The physical removal of the excess ABHS application with a paper towel during the first step facilitates the removal of soil particles compared with standard ABHS protocols, which do not require this initial step.
For some, but not all, bacterial indicators, the handwashing and two-step ABHS interventions resulted in lower bacterial indicator concentrations in hand rinsates than that in controls. The two-step ABHS intervention demonstrated lower bacterial indicator loads than controls on hands (cantaloupe for PC 2, jalapeño for coliforms and PC 1). In the case of jalapeño, the two-step ABHS intervention had significantly lower coliform concentrations than handwashing, consistent with previous findings (10–12, 34–37). Bacteria were likely inactivated by the 70% ethyl alcohol, but the exposure to two large doses of the alcohol-based formulation along with a paper towel wipe in between, which creates friction, may have also led to bacterial removal and inactivation (38). Neither intervention was effective at reducing the concentrations of Bacteroidales irrespective of the produce commodity harvested. Bacteroidales species are unique from coliforms, E. coli, and Enterococcus sp. in that they are limited to warm-blooded animals, are found in higher quantities in the gut, and are unable to propagate in the environment and thus may more accurately represent human fecal contamination. In summary, the two-step ABHS intervention and handwashing intervention performed similarly at reducing bacteria on hands.
One strength of this study was that it compared two hand hygiene interventions across diverse produce commodities using multivariable statistical methodologies. Another strength was that the breadth of measurements (bacterial, soil, and composite scoring) allowed for more comprehensive comparisons and analysis. Moreover, all studies were conducted in an agricultural setting; therefore, results are more likely to reflect the real-world impacts of hand hygiene interventions.
Despite these strengths, this study did not assess hand hygiene technique or compliance, and compared only three different produce commodities, and may not be generalizable to other high-risk produce commodities. In addition, only a limited number of farms for each produce commodity, all from a similar geographic region in Mexico, were evaluated in the study. Nevertheless, this study represents an important first step in understanding whether hand hygiene effectiveness is impacted by the produce commodity harvested. Future studies are warranted to determine whether these results translate to other produce commodities, geographic regions, and harvesting seasons. A second limitation to this study was that the brand of soap, and thus formulation ingredients and levels, used for the tomato study was different than the product used for the cantaloupe and jalapeño studies. Although in vitro testing of both soap brands showed that they were comparable in their ability to kill six different pathogenic bacterial species, slight differences in the brands may have impacted hand hygiene effectiveness (39, 40). Despite this limitation, differences in hand hygiene effectiveness were still observed between cantaloupe and jalapeño farmworkers in which the same products and methodology were utilized.
The FDA’s Produce Safety Rule provides general guidelines for when personnel are required to wash their hands, including “before starting work, before putting on gloves, after using the toilet, upon return to the work station after any break or other absence from the work station, as soon as practical after touching animals… and any other time when the hands have become contaminated” (6). The last frequency requirement may be difficult for compliance, given our past results showing that farmworker hands rapidly become recontaminated after performing hand hygiene interventions (within one cycle of jalapeño harvest of ∼30 minutes) (12). Furthermore, based on these and past results, the amount of soil on hands should not be used as the sole criteria for when hands have become contaminated because hands can harbor microbial contamination in the absence of large amounts of soil. Last, our results suggest that handwashing with soap and water may not be sufficient to remove or inactivate microbial hand contamination in diverse agricultural contexts. Evidence suggests that newly developed antimicrobial soap formulations could be more effective at removing or inactivating microbes (41, 42) and should be considered for further evaluation in the agricultural environment. Thus, it is unclear what criteria agricultural workers should use to determine when their hands may have become contaminated and what frequency and method of hand hygiene intervention should be implemented. Although the FDA’s Produce Safety Rule does not allow ABHS products to be used in lieu of soap and water for hand hygiene (6), this method warrants further investigation because our results suggest that the two-step ABHS method can remove dirt and some bacterial indicators from farmworkers’ hands. The two-step ABHS method may be an effective alternative to handwashing with soap and water when soap and water are not readily available, which could improve hand hygiene frequency and compliance.
While it is important to define the optimal hand hygiene agents, protocols, and programs for use in specific agricultural settings and across produce commodities, doing so should not delay the implementation of currently recommended hand hygiene protocols as required by the FDA’s Produce Safety Rule. Comparison testing of multiple product formulations in a variety of agricultural environments is resource intensive, which might delay the adoption of more effective hand hygiene on farms. Therefore, it may be more practical and better for overall risk reduction to define a minimum performance standard that all hand hygiene products must achieve such that they would be effective in the highest risk agricultural settings. Finally, the methods utilized (diversity of bacterial indicators and multivariable analyses) by and results obtained (variable hand contamination profiles and hygiene product effectiveness) from this study may have application beyond the agricultural setting to other settings (e.g., health care, food preparation) to ensure adequate risk reduction.
MATERIALS AND METHODS
Study location and participants.Three separate hand hygiene intervention studies were conducted on two farms that produced cantaloupes (Cucumis melo var. cantalupensis), three farms that produced jalapeño peppers (Capsicum annuum), and one farm that produced tomatoes (Solanum lycopersicum); all farms were located in Nuevo León, Mexico. The study location, participant characteristics, and results of the tomato and jalapeño studies have been previously described (11, 12), but the cantaloupe study is described here. All farms used drip irrigation with additives (fungicides, insecticides, and synthetic fertilizers), with water obtained from deep wells. In the field, the tomato farmworkers were normally required to wear gloves during pre- and postharvesting activities but removed them for participation in this study. All produce commodities were harvested by hand with no glove use, but jalapeño and tomato farmworkers reported using knives to cut stalks. Jalapeños and cantaloupes were grown in open fields, while tomatoes were staked and grown under protective cover in a semienclosed system with shade cloth. All three produce commodities were grown on plastic mulch. A total of 326 participants that included 129 from the cantaloupe farms, 120 from the jalapeño farms, and 77 from the tomato farm were enrolled in the study. For each of the three studies, data collection occurred over several nonconsecutive days during the harvest season in order to capture variability. The tomato study took place over a 4-week period during the 2014 harvest season (August to September), with data collection occurring on 5 nonconsecutive days. The jalapeño study took place during May 2013, with data collection occurring on four nonconsecutive days. The cantaloupe study took place during the 2014 harvest season (May and July), with data collection split between two nonconsecutive days—one at the beginning of harvest season and one at the end. Before enrollment, staff explained the study to potential participants and obtained their consent. Inclusion criteria for the participants were as follows: 10 years of age or older; an employee of the farm assigned to harvesting cantaloupes, jalapeños, or tomatoes; and provision of oral informed consent to participate in the study. There were no exclusion criteria. Study design and protocols for research on human subjects were reviewed and approved by both the institutional review boards (IRBs) of Universidad Autónoma de Nuevo León in Monterrey (UANL), Mexico, and Emory University in Atlanta, GA (IRB number 00035460).
Study design and hand hygiene intervention description.All three studies had the same design and protocols except that farmworker age, gender, and Bacteroidales data were not collected for the tomato study. The tomato study used a different brand of soap (Pearl Lotion Hand Soap; Noble Chemical, Inc., Lancaster, PA) and ABHS (Purell Advanced Instant Hand Sanitizer; GOJO Industries, Akron, OH) than the jalapeño and cantaloupe studies, which both used the same brand of soap (Green Certified Foam Hand Cleanser; GOJO Industries) and ABHS (disinfectant gel; Desinfectantes y Aromatizantes, S.A., Monterrey, Mexico). Soap data from the tomato study were included in this study because in vitro efficacy studies (43) showed no statistical difference between the killing of six pathogenic bacterial species (Salmonella enterica serotype Cholerasius, Listeria monocytogenes, Escherichia coli, Staphylococcus aureus, Klebsiella pneumoniae, and Staphylococcus epidermidis) between the soap brand used in the tomato compared with that in the jalapeño and cantaloupe (same brand) studies (data not shown). ABHS data from the tomato study were excluded because significant in vitro (43) differences in antibacterial efficacy were found between the tomato ABHS used and the jalapeño and cantaloupe ABHS (both used the same locally sourced ABHS).
After giving consent, the participants for each study were randomly assigned to either a no hand hygiene (control) group or one of two intervention groups: (ii) handwashing with soap and water (handwashing) or (ii) two-step ABHS, as described below and trained (11, 12). To standardize the bacterial load on farmworker hands, all farmworkers washed their hands with soap and potable tap water following the same protocol for the handwashing intervention. All potable tap water utilized throughout the studies was provided by the laboratory at the Universidad Autónoma de Nuevo León in Monterrey (UANL), Mexico. Potable tap water in Nuevo León, Mexico, is chlorinated at water treatment plants prior to distribution, but residual chlorine levels in the water were not measured prior to use in this study. All potable tap water used in this study tested negative for coliforms, E. coli, and Enterococcus sp. (100-ml aliquot). The farmworkers then harvested their respective produce items for one cycle (∼30 to 90 minutes), completed the intervention activities described, and provided a hand rinsate sample. For all three studies, exact harvest times were recorded for each participant. At the cantaloupe and jalapeño farms, but not tomato farms, study staff also recorded participant demographic information (farmworker age and gender). After sample collection, cantaloupe participants were invited to complete an optional survey on their perceptions regarding the hand hygiene interventions. All participants were compensated for their time (e.g., with a cold beverage, bandana, t-shirt, or baseball cap).
Control group.Participants in the control group did not perform any hand hygiene intervention after harvesting produce.
Handwashing group.Handwashing was performed as previously described (11, 12). Briefly, participants in this group rinsed their hands under potable water, rubbed 2 ml of nonantimicrobial hand soap onto their hands for ∼15 to 20 seconds, rinsed their hands again with potable water, and then dried them off with a single-use paper towel. For the jalapeño and cantaloupe studies, the nonantimicrobial soap used was the same (with ingredients in the following order on the label: water, sodium laureth sulfate, cocamidopropyl betaine, citric acid, disodium cocoamphodiacetate, glycerin, PEG-80 sorbitan laurate, polyquaternium-39, methylchloroisothiazolinone, methylisothiazolinone, and sodium benzoate; Green Certified Foam Hand Cleanser; GOJO Industries, Akron, OH). The tomato study used a different brand of soap (with ingredients in the following order on the label: sodium laureth sulfate, dodecanamide, N, N-dimethyl, fatty acid ester, and fragrance; Water Pearl Lotion Hand Soap; Noble Chemical, Inc., Lancaster, PA). As described above, there was no significant difference in in vitro antimicrobial activity between the two brands.
Two-step ABHS group.Two-step ABHS was performed as previously described (11, 12). Briefly, 3 to 4.5 ml (two to three dispenser pumps) of ABHS were dispensed onto participants’ hands. After rubbing their hands for ∼15 to 20 seconds, participants removed excess ABHS with a single-use paper towel. Thereafter, an additional pump of ABHS was dispensed, and participants again rubbed their hands together, this time until dry. For the jalapeño and cantaloupe studies, the ABHS brand used was the same (containing an active ingredient of 70% ethanol [vol/vol] with inactive ingredients in the following order on the label: water, glycerin, TAE [Tris-acetate-EDTA] buffer, carbomer, propylene glycol, and fragrance; disinfectant gel; Desinfectantes y Aromatizantes, S.A., Monterrey, Mexico).
Hand rinsate sample collection and testing.Hand rinsate samples were collected for all groups as previously described (11, 12) immediately following the above described interventions or after produce harvesting for the control group. Briefly, participants placed one hand in a Whirl-Pak bag containing 750 ml of sterile 0.1% peptone water. The participant agitated the hand while a study staff member massaged their hand within the bag for 30 seconds. The procedure was repeated for the other hand using the same Whirl-Pak bag (11, 12). After collection, samples were placed on ice and transported to the Laboratory of Microbial Biochemistry and Genetics at the Universidad Autonóma de Nuevo León (UANL), where they were stored at 4°C until analysis. Analysis was performed within 24 to 48 hours of field collection. If the microbial analysis results were outside the quantifiable range and a repeat analysis was necessary, the repeat analysis was conducted within 72 hours of field collection.
Rinsate samples were inverted several times to resuspend any particulate matter, and then an aliquot was taken for an absorbance measurement of an optical density at 600 nm (OD600) using a spectrophotometer (Sequoia Turner, Mountain View, CA) at UANL. A second aliquot was shipped overnight on ice to the Department of Food, Bioprocessing, and Nutrition Sciences, North Carolina State University for microbial source tracking analysis. Detection and quantification of universal and human-specific Bacteroidales 16S rRNA genetic markers were performed by quantitative real-time PCR (qPCR) using the AllBac and BFD primers as previously described (4). Serial dilutions of the remaining hand rinsate sample were processed by membrane filtration in duplicate, and indicator bacteria (total coliforms, total Enterococcus sp., and generic E. coli) were enumerated (CFU/hand) on selective media at UNAL as previously described (11, 12). Following filtration through duplicate membranes for each serial volume of rinsate, each membrane was placed on a separate petri dish containing solidified agar for bacterial enumeration. Membranes were placed onto chromogenic Rapid′E. coli 2 agar (Bio-Rad, Hercules, CA) and incubated at 44°C for 24 hours to enumerate E. coli (pink/purple colonies) and coliforms (both blue/green and pink/purple colonies). To quantify Enterococcus bacteria, membranes were placed on Kenner fecal Streptococcus agar plates (BD, Franklin Lake, NJ) and incubated at 37°C for 48 hours, and red-centered colonies were counted. Bacterial concentrations (CFU/hand) were calculated as described from bacterial colony counts from the serial sample filtrations. Assay limits of detection (LODs) were calculated as 1 CFU for the largest effective volume tested. Sample concentrations below the LOD were assigned a value half the LOD.
Data entry and statistical analysis.All data from the tomato and jalapeño studies have been previously described (11, 12). Statistical analysis was performed using Statistical Analysis Software version 9.4 (SAS Institute Inc., Cary, NC), except for the principal-component analysis (PCA), which was performed using JMP Pro 13. Figures were created using Prism version 7.04 for Windows (GraphPad Software, La Jolla, CA USA). Because all bacterial indicator concentrations and OD600 values were nonnormally distributed (data not shown), either nonparametric tests or log10-transformed variables were used. Pearson’s χ2 test (44) was used to compare the proportion of males, and the Kruskal-Wallis test was used to compare farmworker age and duration of harvest across produce commodities and study groups. Correlations between soil, bacterial indicators, and Bacteroidales markers were evaluated using Spearman’s rank correlation tests. Correlations were classified by the strength of association (Spearman’s ρ value), namely, weak (0.00 to ±0.30), moderate (0.31 to 0.7 or −0.31 to −0.70), or strong (>0.71 or ≤0.71) (45).
Two separate PCA models were constructed with the following variables: log10-transformed values of OD600, total coliforms, total Enterococcus sp., and E. coli counts (first PCA model only). In the first PCA model, only control group data were included to assess differences in hand contamination profiles in the absence of the intervention, while for the second model, all control and intervention group data were included to assess whether intervention efficacy differed by produce commodity. PCs were calculated based on correlations using centered and scaled log-transformed bacterial indicator variables. Statistical comparisons of PC scores between produce commodities were made using the Steele-Dwass test for multiple comparisons (46).
To assess the impact of produce commodity on the ability of hand hygiene interventions to reduce soil and microbes on farmworker hands, multivariable two-way fixed analysis of covariance (ANCOVA) was constructed (47). This method was chosen because it allowed consideration of the effect of intervention group and produce commodity across multiple studies with a simultaneous Tukey’s correction for multiple comparisons (48). PCA was performed using the soil and bacterial indicator concentrations from all combined control and intervention groups to assess the impact of hand hygiene interventions at reducing hand contamination profiles; the loading matrices are shown in Table 3. To ensure that bacterial indicator concentrations that were below the limit of detection (LOD) were not significantly affecting the two-way fixed ANCOVA model results, a Tobit analysis (49) that censored bacterial indicator values at one-half the value of the LOD was performed. Because Tobit models cannot test for interactions between variables, we compared an ANCOVA model without the hand hygiene intervention and produce variable interaction term to an analogous Tobit model.
Loading matrices for principal-component analysis of hand contamination across cantaloupe, jalapeño, and tomato farmworkers
We excluded E. coli from the fixed-effects and PCA model because its low percentage of positives (0% to 24%) led to low-confidence model estimates, as assessed by the Tobit model comparison. To compare differences in indicator bacteria and Bacteroidales proportions across cantaloupe, jalapeño, and tomato study groups, Pearson’s χ2 test (44) with post hoc Bonferroni correction (50) was used. All statistical tests were conducted with an alpha level of 0.05.
ACKNOWLEDGMENTS
This work was supported by NIFA grants (2010-85212-20608, 2011-67012-30762, 2015-67017-23080, and 2018-07410). The tomato study was also partially financially sponsored by GOJO Industries, Inc., through an unrestricted research grant to cover partial salary for the effort of the study team, supplies, and communication of results.
We thank the participating farmers for their collaboration; the scientific and community advisory board members Elizabeth Bihn, James Gorny, John (Jack) Guzewich, Robert Mandrell, José Luis Rodríguez Cavazos, José Elías Treviño Ramírez, and Lorenzo J. Maldonado Aguirre; Cindy Caballero, Carmen Cárdenas, Rafael García, Karina Molina, Luisa Solís-Soto, and Fabiola Venegas for sample collection and microbial analyses; and Sanemba Aya Fanny, Valerie Morril, Carol Ochoa, Vidisha Singh, Amelia Van Pelt, Dominique Watson, Ryan Hall, and Adam Lipus for assistance with laboratory assays, data management, and statistical analysis.
FOOTNOTES
- Received 8 April 2020.
- Accepted 8 July 2020.
- Accepted manuscript posted online 17 July 2020.
- Copyright © 2020 American Society for Microbiology.
