Recruitment

We identified fruit and vegetable growers via Maryland’s Best website (https://marylandsbest.maryland.gov) and emailed them invitations to participate in a questionnaire about their farming activities. The Maryland’s Best website is a public database of farm operations maintained by the Maryland Department of Agriculture’s Marketing and Agribusiness Development section to connect producers and consumers. We also contacted growers who previously participated in a qualitative study of soil contact [9] or in the Safe Urban Harvests Study [4, 31]. We contacted 240 growers via email; 5 who responded were ineligible (i.e., not currently growing fruits or vegetables or not located in Maryland); 10 who responded indicated they were not interested; 187 did not respond; and 38 participated in at least one season of data collection. Growers were eligible if they were currently a farm owner/manager, farm employee, or community gardener in Maryland; ≥18 years of age; had completed farm activities related to edible plant production (e.g., planting, harvesting, weeding, mulching) within the past 12 months; and expected to be engaged in farm activities in the upcoming 12 months. We excluded growers who worked on farms that engaged solely in food animal production.

Data collection

We developed and administered a soil contact activity questionnaire (Supplementary Material) to 38 growers beginning in the spring of 2020. Growers were invited to respond to the same questionnaire each subsequent season (i.e., summer, fall, and winter). Questionnaires were administered between April 2020 and March 2021. Because the questionnaire queries growers about meso-activities conducted in the previous 30 days, administration began 30 days after the start of each season (i.e., administration of the fall questionnaires began on October 22, 2020, 30 days after the first day of fall). Due to restrictions on in-person research due to the COVID-19 pandemic, all questionnaires were administered via telephone. The questionnaire was semi-structured and contained a combination of open- and closed-ended questions. Some growers provided additional information above and beyond their answers; this material was recorded to enhance context. All data were collected and managed using Research Electronic Data Capture (REDCap) tools hosted at the Johns Hopkins Bloomberg School of Public Health [32, 33]. REDCap is a secure, web-based software platform that supports data capture for research studies.

Informed consent was obtained from all growers prior to the administration of the questionnaire. Questionnaire administration times ranged from 5 to 88 min, with a mean of 28 min. Mean questionnaire administration times decreased with each subsequent season (i.e., spring = 40; summer = 27; fall = 28; winter = 18 min). Growers were offered a gift card on an increasing scale (i.e., $5, $10, $15, $20) for each questionnaire completed. All study tools and protocols were reviewed and approved by the Johns Hopkins Institutional Review Board (IRB00009866).

Soil contact activity questionnaire

The questionnaire was informed by a framework of environmental, activity, timing, and receptor (EAT-R) factors for estimating soil ingestion [30]. Specifically, in this investigation, we aim to generate quantitative and semi-quantitative estimates to characterize further the six meso-activities or tasks (i.e., bed preparation, planting seeds, transplanting, irrigation, weeding, and harvesting) identified by that previous qualitative characterization. These quantitative exposure factors can then be used to inform occupational exposure assessments for agricultural workers. We asked growers to estimate the frequency (days per week) and duration (hours per day) spent engaged in agricultural tasks on-site over a typical month during a given season. We also asked growers to estimate the frequency (days per month) and duration (hours per day) of six specific meso-activities (i.e., bed preparation, planting seeds, transplanting, irrigation, weeding, and harvesting). We also asked growers to estimate what fraction of time their hands were in contact with soil while conducting each task. We then followed up with specific questions about how the task was conducted, i.e., what tools were used, what clothing items were worn, use of PPE, and in what ergonomic position(s) they were in (e.g., kneeling, standing).

We asked all growers, “In the current month, do you recall getting soil in your mouth while working at your farm/garden?” For growers who answered “yes” and reported ingesting soil in the past 30 days, we followed up with two additional questions: 1) “How many days this month do you recall getting soil in your mouth?” and 2) “On a typical day, when you got soil in your mouth, how much soil do you think entered your mouth?” showing them images of varying amounts of soil on spoons (Fig. S1) and asking them to estimate the amount of soil they ingested each day by selecting the image that most closely matched the amount of soil they ingested. For growers who reported ingesting soil in the past 30 days, we showed them images of varying amounts of soil on spoons (Fig. S1) and asked them to estimate the amount of soil/dust they ingested each day by selecting the image that most closely matched the amount of soil/dust they ingested. The amounts of soil on each spoon were based upon EPA’s recommended default exposure factors for daily soil ingestion [10] (10 mg/day = general adult population central tendency; 50 mg/day = general adult population upper percentile; 1000 mg/day = child with soil pica) or dust ingestion (20 mg/day = general adult population central tendency) or soil and dust ingestion (100 mg/day = general adult population upper percentile; 200 mg/day = general child population upper percentile). Two additional options for less than the smallest (<10 mg/day) and greater than the largest amounts (>1000 mg/day) were included. We also queried growers about two events (i.e., incidentally ingesting soil, getting soil on the face) and two behaviors (i.e., sampling produce, and eating a meal or snacks onsite while working) that may increase the likelihood or amount of soil contact. We also asked growers to provide demographic information on themselves (e.g., age, means of compensation) and their farms (e.g., farm size, organic certification status).

Data analysis

We used Kruskal-Wallis tests to assess differences in growers’ questionnaire responses to four exposure factors: days per month spent at a growing site, hours per month at a site, hours per month engaged in all tasks, across seasons and hours per month engaged in all tasks, across meso-activities. For each exposure factor, we conducted Shapiro-Wilk tests of the distributions to determine whether the distributions were normally or lognormally distributed. We used Pearson correlations to assess correlations between farm size and hours worked per month (both on-site and across all six activities). We used non-parametric Kendall’s tau tests to assess correlations between employment status and hours worked per month (both on-site and across all six activities).

We reviewed all open-ended responses to identify emergent themes, and then through an iterative process SL coded each response according to those themes. Analysis of soil contact activity questionnaire responses was conducted in Excel. Empirical exposure estimation and Monte Carlo simulation, data displays, and hypothesis testing were conducted in Python v. 3.6.

Exposure assessment

We estimated exposure by calculating ADDs for each of the 38 growers queried in our study using grower-reported data about the frequency of farming-related meso-activities (Fig. 1), and interactions with soil (Fig. 2). One of the challenges of working with the default soil ingestion rates (e.g., 100 mg/day for the general population and 330 mg/day for highly exposed workers) is the fact that they are presented as daily rates. While daily rates are easier to implement in a regulatory context (i.e., for deriving soil screening levels), they are problematic when used in an occupational context to estimate exposure since they are not easily adaptable for the variable tasks (and their associated soil contact intensities, durations, and frequencies) that constitute agricultural work. For example, the 330 mg/day ingestion rate default is used whether a construction laborer works 8 h or 12 h per day. No adjustments can be made to account for the reasonable assumption that a laborer working more hours would logically incur more exposure than one working fewer than 8 h per day or that a worker may engage in various tasks with different intensities of soil contact. To demonstrate the relevance of incorporating grower-reported frequencies and durations of specific agricultural tasks, we estimated ADDs for each grower who completed questionnaires (empirical) and used Monte Carlo simulation over a 1 month, i.e., (30-day) period using three methods: daily, hourly, and hourly, task-specific soil ingestion rates. The three methods are summarized on GitHub [23].

Fig. 1: Grower-reported frequency and duration of six meso-activities.

a Frequency (days per month) of each meso-activity performed, by season. b Duration (hours per day) of each meso-activity performed, by season. c Cumulative duration (hours per month) of each meso- activity performed, by season. Boxes represent interquartile ranges (Q1-Q3) and black lines represent medians. Results represent only growers who engaged in each activity, i.e., zero values were not factored into box plots. Each dot indicates one grower’s response for a given season.

Fig. 2: Grower-reported soil contact and ergonomic positioning for six meso-activities.

a Percentage of time hands were in soil during each activity. b Ergonomic positioning during each meso-activity. Responses are not mutually exclusive, i.e., a grower might kneel and sit during the same activity. Examples of “other” positions include: riding on a tractor, squatting, etc. Boxes represent interquartile ranges (Q1-Q3) and black lines represent medians. Each dot represents one grower’s response for a given season. Results represent only growers who engaged in a given activity during a given season.

(Traditional) Method 1. Daily soil ingestion

As a baseline, we estimated monthly/seasonal ADDs of a hypothetical chemical using the following equation:

$$=\frac_}}}* }_}}}* }_}}}* }_}}}_}* }_}}$$

where C is the concentration of a hypothetical soil contaminant held constant at 400 mg/kg; IR is the daily default soil + dust ingestion rate held constant at 378 mg/day, (adjusted from mg/day to kg/day); and BW is the bodyweight of the grower (kgBW). The 378 mg/day soil ingestion rate represents a high-end (95th percentile) estimate derived from a recent study that specifically modeled a high-contact soil scenario [34]. We derived an exposure factor representing the fraction of days worked in a month by multiplying the exposure frequency (EF) or grower-reported days worked per week by 4.35 (i.e., the average number of weeks in a month) to obtain the typical number of days worked per month [23]. The exposure factor was derived using the EF of grower-reported days worked in a typical 30-day period (for that season); an exposure duration (ED) of 1 month, and an averaging time (AT) of 1 month (equivalent to 30.5 days). Bodyweight (BW) was based on the median body weight (kgBW) from the Exposure Factors Handbook [35] associated with the grower’s age bracket and sex [23].

Method 2. Hourly soil ingestion

To more precisely account for the variability in the number of hours growers worked per day, we estimated monthly/seasonal ADDs of a hypothetical chemical from outdoor (i.e., working exposures) and indoor (i.e., non-working exposure) using hourly ingestion rates. We derived hourly ingestion rates separately for time spent working outdoors and time spent indoors (not engaged in agricultural work) modeled using ingestion rates in ref. [34]. We derived an hourly ingestion rate for outdoor agricultural work of 45.25 mg/h by dividing the 362 mg/day soil ingestion rate default for growers in a high-contact soil scenario while working outdoors by 8 (assuming the worker works outside for 8 h). Hubbard et al. provide a revised modeled estimate of dust ingestion while indoors of 22 mg/day. To derive an hourly ingestion rate of 1.38 mg/h for dust ingested while indoors (and not engaged in agricultural work), we divided the daily dust ingestion rate (22 mg/day) by 16, assuming that the grower spends all non-working hours indoors.

We estimated seasonal/monthly ADDs for indoor and outdoor exposure to a hypothetical contaminant using the following equation:

$$}_\,\,}=\frac_}}}* }_}}}* }_}}}* }_}}}_}* }_}}$$

where C is the concentration of a hypothetical soil contaminant (mg/kg), held constant at 400 mg/kg and IR is the hourly default soil ingestion rate held constant at 45.25 mg/hour (adjusted to kg/hour) for hours worked outdoors and 1.38 mg/h (adjusted to kg/hour) for hours not worked and spent indoors. The exposure factor (for outdoor hours) was derived using the EF grower-reported (i.e., number of grower-reported hours worked per day by the number of grower-reported days worked per week, multiplied by 4.35 weeks per month), an ED of 1 month, and an AT of 1 month (equivalent to 730.8 h). The EF (for indoor hours) was assumed to be the complement of outdoor hours and derived by subtracting the grower-reported EF for outdoor hours from 730.8 h. We assumed an ED of 1 month, and an AT of 730.8 h (equivalent to 1 month) for indoor exposures. To convert the average hourly daily dose to an average daily dose we multiplied by 24. Bodyweight was based on the median body weight (kgBW) from the Exposure Factors Handbook [35] associated with the grower’s age bracket and sex. ADDs for both indoor and outdoor exposures were summed.

Method 3. Hourly-task-specific soil ingestion

There is variation across agricultural tasks in the rate or intensity of soil contact, but scant research exists to inform precise modifications to hourly ingestion rates by meso-activity. To account for this variation in intensity, we used responses from the grower soil contact activity questionnaires to create scaling factors to adjust the baseline hourly outdoor ingestion rate (45.25 mg/h) to each task, as follows: For each of the six tasks we queried growers about, we asked them to report the percentage of time that involved direct soil contact and we either maintained or adjusted the hourly intake rate up or down accordingly. For example, the mean rate of grower-reported time in contact with soil while engaging in transplanting and weeding was 87% and 72% of the time, respectively [23] so we doubled the baseline hourly soil ingestion rate (45.25 mg/h) for these two tasks to 90.5 mg/h. The mean rate of grower-reported time in contact with soil while engaging in seeding and preparing beds was 49% and 41% of the time, respectively, so we maintained the baseline hourly soil ingestion rate of 45.25 mg/h for these tasks. The mean rate of grower-reported time in contact with soil while engaging in harvesting and watering was 35% and 8% of the time, respectively, so we halved the baseline hourly soil 45.25 mg/h) ingestion rate for these two tasks to (22.63 mg/h). These subjective adjustments to the hourly ingestion rate represent our professional judgment, as directly informed by the grower-reported experiences of soil contact obtained via the soil contact activity questionnaire.

We estimated ADDs of a hypothetical chemical (at 400 mg/kg concentration) over a month using the task-specific hourly ingestion rates and the following equation:

$$\sum }_\,n}\frac_}}}* \right)}_}}}* }_}}}* }_}}}_}* }_}}+\frac_}}}* \right)}_}}}* }_}}}* }_}}}_}* }_}}$$

We derived an exposure factor representing the total number of hours worked in a typical month (on each task) by multiplying the number of grower-reported hours worked per day by the number of grower-reported days worked per month (for that task) to obtain the EF. As in method 2, we assumed an ED of 1 month, and an AT of 730.8 h (equivalent to 1 month). To convert the hourly daily dose to an ADD we multiplied by 24. We added the ADDs for each of the 6 tasks to obtain an overall ADD for outdoor exposure for that grower. All hours not accounted for by the grower reported time spent working on each task was assumed indoors and multiplied by the indoor ingestion rate (1.38 mg/h). ADDs for both indoor and outdoor exposures were summed to yield the total daily exposure. As in methods 1 and 2, we used the median body weight (kgBW) from the Exposure Factors Handbook [35] associated with the grower’s age bracket and sex [23].

Complementing the empirical exposure assessments, we simulated exposures among a population of hypothetical growers (n = 5000) using Monte Carlo simulation in Python v. 3.6. Simulated exposure assessments follow the three methods described above, adapted as follows:

For all three methods, the bodyweight of each simulated grower was randomly sampled from the distribution provided in the 2017–March 2020 NHANES data [36] filtered to include adults aged 21 and over. To reflect the general United States (US) population, the probability of selecting a given body weight value was weighted by the number of US citizens represented by that datum, as described in Centers for Disease Control and Prevention (CDC) documentation [37].

For the simulated variant of method 1 (traditional, daily soil ingestion), the number of days per month spent at a site, by season, was randomly sampled from the associated values reported in questionnaires. For example, the number of days per month each simulated grower spent at a site during the spring was sampled from the pool of empirical responses specific to the spring. Daily soil ingestion rates were the same as those used in the empirical exposure assessment.

Similarly, for the simulated variant of method 2 (hourly soil ingestion), the number of hours per month spent at a site, by season, was randomly sampled from the associated values reported in questionnaires. Hourly soil ingestion rates were the same as those used in the empirical exposure assessment.

For the simulated variant of method 3 (hourly-task-specific soil ingestion), the number of hours per month engaging in each task, by season, was randomly sampled from the associated values reported in questionnaires. For each task, soil ingestion rates were randomly sampled from a uniform distribution centered around the ingestion rates used in the empirical exposure assessment for method 3 [23]. The ingestion rate used for weeding in the empirical assessment, for example, was twice that of the baseline soil ingestion rate; for the simulated assessment, the ingestion rate for weeding was sampled from a uniform distribution between 1.8 and 2.2 times that of the baseline soil ingestion rate. All task-specific distributions for the simulated assessment similarly ranged from 0.2 times lower to 0.2 times higher than the soil ingestion rates used for the empirical assessment.