Online Exclusive: Workplace Exposure Risk Assessments: Worst-Case vs. Random Sampling

Bernard L. Fontaine, Jr., CIH, CSP, FAIHA

Introduction

An occupational hygiene risk assessment is the process used to evaluate a worker’s exposure to a chemical, physical, biological, ergonomic or radiological agent. The exposure risk assessment strategy used depends on the purpose and goal of the monitoring and what the number and type of sample(s) should be representative of the occupational exposure.

There are two types of sampling strategies to consider when planning an exposure assessment study – “worst-case” sampling and random sampling. These terms represent two distinct methodologies, each with its unique advantages and shortcomings. The broad difference is that “worst-case” sampling is more subjective than a random sampling approach.¹ When comparing a worst-case sampling strategy to random (statistical) sampling strategy in occupational hygiene exposure risk assessments, it’s all about the goal—to rule out risks quickly or build a defensible, data-driven understanding of exposure. Let’s break it down:

Understanding Risk

In practice, most occupational exposure risk assessments do not fully address the concepts of risk quantitatively. Comparison of measured or estimated exposure airborne concentrations to the reference occupational exposure limit (OEL) provides insight as to relative exposure acceptability, but does not provide full information of the likelihood or severity of adverse effects as exposure exceeds or is below the OEL.

Interpretation of the comparison of workplace exposure concentrations with an OEL (i.e., risk characterization) is important when making decisions about effective exposure control strategies (i.e., risk management measures). Such information will be the driving input for decisions such as identification of processes where engineering controls are necessary, new work practices may be introduced or respiratory protection may be needed. Risk management decisions often depend on the anticipated amount of risk reduction among exposure control alternatives.²

Worst-Case Sampling

Tracks the highest plausible exposures by targeting workers or time periods where exposure is believed to peak—subjectively selected without chance or randomness. Based on experience, some of the worst-case occupational health exposures occur to maintenance and repair workers. These individuals perform both routine and nonroutine work tasks that expose them to chemical, biological, ergonomic and physical agents, either alone or in combination. Often times, these worst case exposures only occur regularly each month or randomly depending on facility operating conditions.

Maintenance workers may climb into confined spaces such as tanks, pits, or vessels with limited ventilation to clean out or repair. They are often subjected to loud noise, hand-arm or whole-body vibration injuries, and adverse thermal conditions in hot or cold environments. In older building, they may be exposed to asbestos fibers, respirable crystalline silica dust, and heavy metals like inorganic lead, cadmium, and hexavalent chromium. Maintenance tasks involving water systems, HVAC, or confined spaces can expose workers to Legionella, mold, bacteria, and parasites. Airborne exposure and inhalation risks are common. Some maintenance work is performed under time pressure or during off-hours, leading to stress, long shifts, and shift work challenges. These factors increase the likelihood of mistakes by taking shortcuts to expedite the work and risk-taking by not wearing respirators or personal protective equipment.

Other mitigating factors include odd work practices, obvious signs of smoke, dust and fume or the smell of solvent vapors in the air.  Heavy lifting, awkward postures, repetitive tasks, and vibrations contribute to musculoskeletal injuries—common among mechanics and facility maintenance workers.

Many of the formal workers can help identify which routine or nonroutine work tasks are the worst to do. Sometimes, additional information may be available such as trending reports of illness or disease on the OSHA summary of recordable incidents, information on the Safety Data Sheets (SDSs), workers’ compensation loss reports or interviews with workers. Be advised that all SDSs are not prepared similarly. There are differences in reporting rules in the United States, Canada, and Europe.³ Some important information may be intentionally left out since hazardous materials less than 1% of the product can be left out along with cancer-causing agents containing less than 0.1% by weight or volume.

For example, worst case air sampling was done in a small machine shop after it was discovered that benzene was present below reportable quantities in the petroleum coolant. Smoke billowed from several CNC grinders, which triggered an observation of a potential exposure. The coolant supplier confirmed a trace amount of benzene in the product. Upon conducting worst-case sampling, the CNC grinder operators were well overexposed to benzene vapor even though the amount of the human carcinogen was below 0.1% in the oil.

In another business case, the organization was considering new business which included lathe cutting of beryllium stock. Several short-term samples were collected to evaluate the work practice and ventilation inside the machine. Airborne beryllium dust was found the outside of the machine when the CNC operator continued to reopen the door to evaluate the work. The ventilation system installed inside helped reduce the airborne exposure but was circumvented each time the door to the machine was opened.

Strengths

• Faster and resource-light—just need to check worst expected scenarios.
• Useful for initial screening or urgent risk decision-making when quick answers matter.
• If even worst-case samples are below the OEL, the investigator can subjectively conclude exposure is acceptable.

Limitations

• Too subjective: it hinges on judgment—risking bias or missed high-exposure events.
• Doesn’t yield statistical confidence or variability insights.
• Can’t robustly characterize “typical” exposure or measure distribution.

Random (Statistical) Sampling

Occupational hygienists begin their work by anticipating potential hazards—this often includes reviewing industry processes, new materials, or historical incidents. They recognize actual hazards through walk-through surveys, which include on-site observations of work tasks, use of older equipment, workspace layout, and environmental conditions.

Sample locations are chosen randomly across workers and time periods within Similar Exposure Groups (SEGs), often aided by spreadsheet tools or statistical tables. Using available data and professional judgment, occupational hygienists assess exposure levels by using calibrated survey instruments, health impact based on records, and uncertainty for each SEG from random worker interviews and observations.

Exposures can occur when handling cleaning agents, oils, acids, solvents, or encountering residues from processes like paint stripping or brake repair. Based on the prioritization, occupational hygienists decide which workers in which SEGs should undergo further evaluation. It’s important to target the highest-risk SEGs first to ensure efficient time and resource allocation.

First, there should be validated sampling and analytical method. Field blanks, duplicates, or split  samples should be included for quality control. The recommended number of field blanks per sample set is 10% of the total number of  samples, or a minimum of one blank per sample set. Exposure monitoring should be performed using pre- and post-calibrated air sampling devices, direct-read instruments, noise dosimeters, or biological methods as applicable. Consideration should be given to collecting bulk and wipe samples to complement the air samples collected.

For example, bulk samples can verify the content of crystalline silica in the dust or product and wipe samples can be used to evaluate surface contamination in lunch and break rooms for ingestion hazards like inorganic lead or chemicals with skin notations. These chemical hazards are capable of significant dermal absorption where skin exposure can contribute as much as inhalation. Unlike airborne hazards, surface contamination is often unrecognized, leading to unintentional exposure through touch, cross-contamination, or ingestion. Also, immunoassay kits provide quick on-site detection for specific contaminants. Results can be interpreted against site-specific cleanliness criteria—based on toxicity, background levels, and remediation targets.

Contaminants like PCBs, dioxins, and TCE (trichloroethylene) may linger on surfaces from older industrial processes. These pose long-term risks—not just via airborne exposure but through direct contact. In short, toxic surface contamination is a pervasive but often overlooked occupational health hazard.

Dusts or residues containing cadmium, beryllium, or other metals can remain on work surfaces on street or protective work clothing—and transfer via touch or transfer to non-work areas. Without proper control, contaminants transfer from surfaces to hands, clothing, food areas, or even homes. Some low-volatility agents, like isocyanates and elemental mercury, can persist on work surfaces and react or be absorbed, posing risks over time. Mixing residues—such as ammonia-based cleaners with bleach—can form toxic gases like chloramines, causing respiratory distress. Addressing these concerns requires proactive sampling, thoughtful hygiene policies, effective controls, and health and safety culture reinforcement.

The number of collected samples should reflect a representative number of workers within the work area with similar exposures. It should also reflect the ambient air concentration to prevent overloading the filter cassette in the case of asbestos sampling. If there are too many fibers on the filter, the analyst cannot accurately count the number of fibers and sampling time may need to be repeated with a more frequent sampling interval. Another important consideration is the presence of contaminates or work practices that would interfere with the analysis.

Based on years of experience, some unscrupulous organizations have temporarily shut down or curtailed operations, changed work plans or production schedules for the day, transferred workers to other work assignments, opened windows and doors to increase the amount of ventilation to dilute the contaminants, and asked workers to refuse any cooperation of being monitored. If the work can be evaluated properly without inference, here are some of the strengths and limitations of random sampling. Selection of workers to be sampled should be based on visual observation of the work, toxicity of the material, potential exposure based on the operation or work task, and frequency and duration of the exposure.

There are seasonal times when work shifts are extended to meet production schedules. Sampling should continue to compare the results to an adjusted OEL. On the other hand, some workers do things differently such as place their head into printing press to check on the amount of ink in the trough rather than relying on the equipment. In such cases, workers may be exposed to elevated short-term or ceiling concentrations above the OEL. In another business case, a worker used toluene solvent without any respiratory protection to clean the dirt and grim off the of a school bus before being repainted. The painter was overexposed to the ceiling, short-term and full-shift time-weighted average (TWA) OEL.

Strengths

• Objective and statistically defensible.
• Calculates key metrics—mean, median, percentiles, variability, confidence limits.
• Delivers a clearer picture of exposure distribution and uncertainty levels.

Limitations

• More samples needed (AIHA recommends 6–10) for reliable estimates.
• Requires statistical know-how and planning—takes longer and costs more.

Data Analysis and Determining Risk

After sampling, results are compared to OELs and used to statistically evaluate exposure. Exposure assessments rely on central tendency (mean, geometric mean) or percentiles to determine if exposures are acceptable or require action. Based on the variability of the work task, work performance, and environmental controls, additional samples should be collected to determine risk.⁴

Table 1 – Direct Comparisons Between Worst-Case vs. Random Sampling

Statistically, sampling should be repeated several times at various times throughout the year to account for seasonal variability, changes in workload, number of workers, variations in work practices and the selection and use of engineering controls. A comparative analysis should be done to evaluate the sample results over time.

Collecting a single sample is a poor representation of the exposure. Occupational exposures are highly variable and skewed (typically lognormal), meaning most values are low, but occasional high exposures occur—the ones most hazardous and important to detect. One data point provides no statistical reliability and it doesn’t detect errors such as contamination, flow rate deviations, or lab analysis mistakes. It may underestimate the risk, especially for short or ceiling exposures. Taking multiple samples gives a clearer picture of the mean, spread, distribution, and outliers—enabling more informed exposure evaluation.

Exposure Profiling

First, a SEG may be a single worker performing a single task; however, it is often impractical to perform random sampling for each and every worker. So, a more practical approach is to include multiple employees in a SEG who have similar exposures. For example, employees propane-powered lift trucks in a warehouse may be grouped together as having similar potential exposures to carbon monoxide (CO) or welders working on the same base metals would have similar exposures. However, as noted, the difference in selecting a worker to be sampled involves work practice observations and the use of engineering and administrative controls.

A random sample is one where each worker and time period has an equal probability of being selected for sample collection. Careful observations can help identify the worker at greatest risk rather than using random selection. Another consideration in the random selection process is worker experience. Younger workers may not be familiar with or use the controls in the same way as a more experienced worker.

A random number table and/or the random number function in a Microsoft Excel spreadsheet computer program are useful tools in the random selection process. Another consideration is how many samples should be collected in order for the exposure profile to be useful? The answer depends on many factors, including the variability of the sample. The American Industrial Hygiene Association (AIHA, 2006) recommends between 6-10 samples are needed in order to perform a baseline exposure profile.⁵

Breathing zone samples are preferred over area samples to properly evaluate worker exposure. The breathing zone is the hemisphere around the face about 6-9 inches around the worker’s face. With welders, air sample cassettes placed under the hood may interfere with the work. Investigators conducting the exposure monitoring should have at least an understanding of the nature of the job or process in which the agent(s) is used or generated and also have a basic understanding in field sampling methods and techniques. All sampling pumps should be properly calibrated and the sampling media should be fresh from the analytical lab or within the timeframe provided by the manufacturer, such as the case of mold spore sampling cassettes.

Descriptive statistics include the mean, median, percent above the OEL, range, and standard deviation that characterize the sample’s distribution such as the central tendency and the variability in the data. The mean and median are used to measure the central tendency of the data, whereas, the range and standard deviation are measures of variability. By looking at the data from a number of points of view, information, and patterns in the data may be discovered. Many data sets can be interpreted simply by comparing the OEL with descriptive statistics. When most of the data are clustered together well below or well above the OEL, a decision can generally be made on workplace acceptability by using descriptive statistics.

Upper and lower confidence limits (UCL and LCL) and upper tolerance limits are calculated based on knowing (or assuming) a certain underlying distribution of the data set. The type of distribution (i.e., normal or lognormal) will generate different confidence intervals and tolerance limits. Federal and state OSHA programs utilize the upper confidence limit to validate a citation for overexposure to an airborne contaminant. This information can be found in the Index of OSHA Sampling and Analytical Methods.⁶

A random variable is called normally distributed if the distribution (as plotted on a histogram) looks like a bell-shaped curve. Often times, occupational hygiene sampling data is often “skewed to the right” since the exposure values have a lower boundary (i.e., the measured exposure value cannot be less than zero). Taking the log of the variable often mitigates such skewness. In such cases, the distribution is then considered lognormally distributed, or lognormal, if the log of the variable is normally distributed. A lognormal distribution is often applied to occupational exposures, yet the assumption of lognormality is seldom verified. If the data follows neither a lognormal nor a normal distribution, it may not represent a single SEG.

For a normal distribution, the estimated arithmetic mean is the same as the sample mean. However, if the sample data is lognormally distributed, there are several methods for estimating the arithmetic mean and for calculating confidence limits. Other preferred methods for estimating the arithmetic mean and computing the UCL and LCL are described in AIHA’s A Strategy for Assessing and Managing Occupational Exposures, 3rd Edition (2006) but estimating the arithmetic mean and computing the confidence limits using such preferred methods are difficult to compute without the use of a computer and/or specialized software.

If the UCL at 95% results in a value greater than the long-term OEL. It suggests the exposure profile is unacceptable; whereas, a UCL at 95% that results in a value below the long-term OEL suggests that the exposure profile is acceptable. For chemicals with acute (or short-term) effects, the upper tolerance limit of the 95th percentile should be examined. A UTL at 95% that results in a value below the short-term exposure level and ceiling limit suggests that the exposure profile is acceptable, but large numbers of air samples are needed in order to identify “acceptable” environments (Selvin, Rappaport, Spear, Schulman, & Francis, 1987).⁷

Finally, if the air sample results of the exposure profile inconclusive, the SEG may need to be further reexamined for certain workers seem with elevated exposures. To statistically test the significance of the sample set variation, an analysis of variance (ANOVA) may be performed. An ANOVA is a statistical test that compares two or more means to determine if the means are significantly different. If the means are statistically different, the SEG should be reevaluated.

Conclusion

Worst-case sampling is an ideal first-pass tool—great for quickly ruling out extreme exposures. But to truly understand exposure patterns, quantify risk, and make defensible decisions, random sampling is the go-to method. Many occupational hygiene strategies use both: start with worst-case to spot red flags, then follow with random sampling to validate and quantify risk.

Historically, more attention has been given to airborne exposures rather than physical agents and dermal exposures. However, there are many situations in which random sampling strategies may be applied to data other than airborne samples. For some chemicals, skin absorption may be the predominant route of exposure and airborne samples would not be the most appropriate variable to study. Biological monitoring may be more appropriate to study for such circumstances.

Random sampling approaches may be applied to physical agents. For noise data and/or other variables measured in a logarithmic scale, analyzing the allowable dose (i.e., percent of dose), rather than decibels, should be considered so that statistical tools can be applied to such exposure measurements.

Both worst-case sampling and random sampling strategies are useful in assessing exposures. It is important to understand the limitations of each and to correctly apply the chosen air sampling strategy. A primary benefit of a random sampling strategy is that it allows SEGs to be profiled with a known level of certainty, which makes it a more defensible and objective sampling strategy. Conversely, the primary benefit of a “worst-case” sampling approach is that fewer samples are needed (thereby less costly and less time-consuming) to make an exposure judgment. In some cases, both a combination of a “worst-case” and a random sampling strategy may be beneficial.

References

1. Jerome Spear, Industrial Hygiene Exposure Assessments: Worst-Case vs. Random Sampling, https://jespear.com/industrial-hygiene-exposure-assessments-worst-case-vs-random-sampling/

2. Waters M, McKernan L, Maier A, Jayjock M, Schaeffer V, Brosseau L. Exposure Estimation and Interpretation of Occupational Risk: Enhanced Information for the Occupational Risk Manager. J Occup Environ Hyg. 2015;12 Suppl 1(sup1):S99-111.

3. Karrie Ishmael, GHS SDS Ingredient Disclosure, ICC Compliance Center (ICC), https://www.thecompliancecenter.com/ghs-sds-ingredient-disclosure/

4. Waters, M., Selvin, S., & Rappaport, S. (1991). A Measure of Goodness-of-Fit for the Lognormal Model Applied to Occupational Exposures. AIHA Journal, 52, 493-502.

5. (2006). A Strategy for Assessing and Managing Occupational Exposures, 3rd Edition. (J. Ignacio, & W. Bullock, Eds.) Fairfax, VA: AIHA Press.

6. OSHA Index of Sampling and Analytical Methods, https://www.osha.gov/chemicaldata/sampling-analytical-methods

7. Albright, S., Winston, W., & Zappe, C. (1999). Data Analysis and Decision Making with Microsoft Excel. Pacific Grove, CA: Brooks/Cole Publishing Company.