A qualitative characterization of meso-activity factors to estimate soil exposure for agricultural workers

Soil can act as a reservoir for pesticides, metals, PFAS and other contaminants that pose risks to human health. Agricultural workers have frequent, direct contact with soil, but current methodologies inadequately characterize their soil exposure. Traditional activity pattern approaches focus on macro-activities (e.g., farming) or micro-activities (e.g., hand-to-mouth), which do not fully capture task-specific variability affecting exposure via ingestion, dermal contact, and inhalation. The study proposes an intermediate activity level—meso-activity (specific farming tasks)—to better link macro- and micro-activity data and to capture the frequency, duration and intensity of soil contact. The authors aim to identify key agricultural meso-activities, characterize contextual factors influencing soil contact, and present an Environmental, Activity, Timing, and Receptor (EAT-R) framework to improve soil exposure estimation and demonstrate its application within US EPA exposure models.

Existing exposure factor data for soil contact are limited, often derived for the general population, and rarely tailored to occupational settings such as agriculture. Prior work emphasizes macro-activity data for air pollutants and micro-activity data mainly for children’s hand- and object-to-mouth behaviors; few micro-activity data exist for adults or agricultural workers. The EPA’s Pesticide Handler Exposure Database (PHED) incorporates task scenarios for pesticides but does not address broader agricultural soil exposures. Job titles have been used as surrogates for exposure assessment, yet they can be imprecise due to task variability within occupations. Recent modeling suggests higher soil ingestion rates for adults in high-contact occupations than earlier tracer studies. These gaps highlight the need for task-specific (meso-activity) characterization and inclusion of broader behavioral and environmental modifiers in soil exposure assessments for agricultural workers.

Design: Qualitative study using in-depth interviews (IDIs) and a framework analysis approach, followed by a demonstration translating qualitative insights into quantitative exposure modeling.

Participants and recruitment: Purposive sampling identified 16 fruit and/or vegetable growers in Maryland (farm owners/managers, employees, or community gardeners, ≥18 years old) who had conducted food production tasks within the previous year and anticipated future engagement. IDIs were conducted on-site between January and February 2020. All participants provided informed consent; IRB approval: Johns Hopkins IRB00009866.

Data collection: A semi-structured guide elicited descriptions of typical workdays, farm operations, detailed conduct of tasks (e.g., planting, irrigation, weeding, harvesting), soil contact (including incidental ingestion), PPE and attire, hand hygiene facilities, and health and safety concerns. Interviews (21–92 minutes; median 55, mean 50) were audio-recorded, transcribed verbatim via NVivo, and verified for accuracy.

Qualitative analysis: Using a framework method (transcription, familiarization, coding, developing and applying an analytical framework), transcripts were coded deductively (e.g., farm description, background, tasks) and inductively for emergent themes. Six meso-activities were identified: bed preparation, planting, irrigation, harvesting, pest management, produce handling. Factors affecting soil contact were coded and mapped to exposure science constructs and grouped into the EAT-R framework: Environmental (natural; social/built), Activity (crop type; growing practices; ergonomic positioning), Timing (season; day of week; time of day), and Receptor (biological; behavioral).

Framework demonstration and quantitative translation: The EAT-R qualitative factors were mapped to quantitative exposure model inputs (e.g., intake rate, exposure frequency/duration, soil-to-skin adherence, skin surface area, body weight, averaging time). Using US EPA Risk Assessment Guidance for Superfund (RAGS) and Exposure Factors Handbook defaults where applicable, the authors conducted sensitivity analyses to estimate average daily dose (ADD) via ingestion and dermal contact. Iterative models incorporated: (1) defaults; (2) meso-activity-specific exposure factors; (3) timing (seasonal/task frequencies and durations); (4) environmental modifications to intake/adherence by season; and (5) receptor-specific biological and behavioral differences (e.g., body weight, attire/PPE, tool use).

• Six routine meso-activities essential to fruit and vegetable production were identified: bed preparation, planting, irrigation, harvesting, pest management, and produce handling. Growers also described relevant micro-activities (e.g., hand-to-mouth, object-to-face) occurring within these tasks.
• Ten modifying factors affecting frequency, duration, and intensity of soil contact were grouped into the EAT-R framework: Environmental (natural: climate/weather; social/built: farm size/type, technology/facilities, workforce), Activity (crop type, growing practices, ergonomic positioning), Timing (season, day of week, time of day), and Receptor (biological, behavioral including PPE use and habits like produce sampling or smoking).
• The framework clarifies how contextual factors influence soil exposure within and across tasks and suggests interactions among factors (e.g., season and day-of-week structuring tasks; mechanization and workforce size; positioning and tool use).
• Quantitative demonstrations showed that incorporating EAT-R factors can materially change exposure estimates compared to default models: • Ingestion (ADD, mg/kgBW/day):
  • Traditional model: 4.38×10^-5.
  • With meso-activity-informed exposure factor: 9.13×10^-5.
  • With meso-activity + timing: seasonal task-specific ADDs produced a total around 5.08×10^-5; other seasonal parameterizations yielded totals up to 8.73×10^-5 and as high as 1.37×10^-4 when environmental adjustments to intake rate were included.
  • Receptor-specific example: Grower A total 1.74×10^-4 vs Grower B 7.08×10^-5. • Dermal (ADD, mg/kgBW/day):
  • Traditional model: 1.26×10^-7.
  • With meso-activity: 1.58×10^-7.
  • With meso-activity + timing: total 1.45×10^-7 (task-specific ADDs on the order of 3.03×10^-8 to 6.06×10^-8).
  • With environmental (seasonal soil-to-skin adherence): totals reported around 1.94×10^-7 and up to 4.08×10^-7 depending on parameterization.
  • Receptor-specific example: Grower totals ranged from 8.02×10^-9 to 3.82×10^-7, illustrating substantial inter-individual variability driven by biological and behavioral factors.
• Overall, the meso-activity-centered EAT-R framework produces more nuanced estimates that better reflect true variability in agricultural work and can identify higher-impact tasks (e.g., transplanting vs watering for ingestion) for targeted risk management.

The study addresses a critical gap in agricultural exposure assessment by introducing meso-activities as a bridge between macro-activities and micro-activities. By organizing ten contextual factors into the EAT-R framework, the authors demonstrate how environmental conditions, task-specific practices, timing, and worker characteristics jointly shape soil exposure. Applying this framework within established EPA dose models shows that default assumptions can under- or over-estimate exposure, while task-, season-, and receptor-specific parameterization captures real-world variability. The approach can guide more precise exposure estimation across ingestion, dermal, and inhalation pathways and help prioritize interventions (e.g., modifying planting techniques, ergonomics, PPE use) at the task level. The framework is compatible with indirect tools (surveys/diaries) and can inform direct observation protocols to quantify parameters needed for robust modeling. Broader data collection to parameterize EAT-R factors would enable improved risk assessment, development of occupational health guidelines, and potential application to other populations with soil contact (e.g., construction workers), with appropriate adaptation.

This work proposes and demonstrates a meso-activity-centered EAT-R framework to refine soil exposure estimation for agricultural workers. By identifying six core tasks and ten modifying factors across environmental, activity, timing, and receptor domains, the framework enables translation of qualitative context into quantitative model inputs, improving the rigor and specificity of dose estimates over default approaches. The study illustrates substantial variability in exposure by task, season, and individual factors, highlighting opportunities for targeted risk mitigation. Future research should: (1) develop a comprehensive database of meso-activity exposure factors; (2) empirically quantify the magnitude and direction of EAT-R factors and their interactions; (3) validate models via direct observation and measurement; (4) expand to inhalation pathway integration; and (5) adapt the framework to other populations and settings while considering population-specific behaviors and tasks.

• Qualitative study with a small sample (n=16) of Maryland fruit and vegetable growers; findings may not capture all agricultural contexts or practices.
• The identified factors may not be exhaustive; interactions among factors were not quantified.
• Quantitative demonstrations relied on professional judgment for certain parameters due to limited empirical data (e.g., seasonal intake and soil-to-skin adherence adjustments), and were not intended to produce definitive exposure estimates.
• Indoor and less soil-intensive tasks (e.g., office, retail, some animal care) were discussed but not included in the framework due to limited relevance or data.
• The study was not designed to derive numerical distributions for micro-activities or to link exposure with specific health outcomes.