Abstract
Background
Exposure to heat leads to physiological and cognitive impairments that increase the risk of workplace injuries. This study estimates the number and proportion of work injuries reported to the United States Occupational Safety and Health Administration (OSHA) that can be attributed to heat exposure. These estimates contribute to the calculation of the benefits of standards, policies, and programs that reduce workplace exposure to extreme heat.
Methods
We analyzed all 2023 injury cases reported to OSHA's Injury Tracking Application by establishments with 100 or more employees, primarily in high-hazard industries. Each injury was geocoded and matched with high-resolution weather data for the specific injury date. Using a case-crossover design, we compared heat index on each injury day (case) with matched non-injury control days for the same worker. Conditional logistic regression was applied separately for summer-only and year-round periods with a non-linear term for heat index to estimate the odds ratios for injury occurrence. We additionally examined heat-injury patterns by industry sectors and in states with/without workplace heat standards.
Results
The odds of work injury increased non-linearly with a rising heat index: the pooled national estimate showed a clear upward trend starting around 85°F and accelerating above 90°F. Our results were consistent across nearly all industry sectors, including those that are predominantly indoors. Using a heat index of 80°F as reference, odds ratios (OR) of injuries at or above 90°F, 100°F and 110°F were 1.03 (95% confidence intervals [CI]: 1.02, 1.04), 1.10 (1.07, 1.13), and 1.20 (1.13, 1.26), respectively. At a heat index of 110°F or higher, the odds increased by 22% in states without occupational heat rules (OR=1.22; 1.15,1.29) versus 9% in states with rules (OR=1.09; 0.84, 1.41), suggesting a protective effect, although confidence intervals overlapped. Overall, we estimate 1.18% (95% empirical CI: 0.92%, 1.45%) of all injuries were attributable to heat exposure on days exceeding a heat index of 70°F.
Conclusion
Heat exposure increases the overall risk of work injury, an effect consistent across nearly all major industries.
Supplementary Information
The online version contains supplementary material available at 10.1186/s12940-025-01231-1.
Keywords: OSHA, Heat rule, Heat standards, Heat related injuries, Climate change, Extreme heat, Occupational heat, Heat index
Introduction
Exposure to high temperatures can lead to heat illnesses and, in extreme cases, heat stroke and death. However, long before reaching such outcomes, heat exposure has been associated with subtler impairments in physiological and cognitive performance. Studies have shown that elevated temperatures can impair hand-eye coordination [1], cause postural instability [2], lead to muscular fatigue [3, 4], and reduce physical reaction times [5, 6]. Cognitive effects of heat include reduced attention [7, 8], memory [9], information processing [10], and overall judgment [11, 12]. Put together, physiological and cognitive impairments increase the risk of workplace injuries few of which are officially classified as “heat” injuries [13].
Efforts to measure these broader impacts of heat on overall workplace injuries in the U.S. have typically relied on natural fluctuations of workers’ compensation claims data linked with temperature data. Such studies have been conducted using data from individual states, including in California [14], Washington [15, 16], and Texas [17], or from multi-state claims datasets [18, 19]. We know of no national assessment that quantified the excess national burden of work injuries attributable to heat.
In 2023, the U.S. Occupational Safety and Health Administration (OSHA) implemented a final rule [20] requiring large establishments (with 100 or more employees) in certain high-hazard industries to electronically submit detailed records of every work-related injury and illness (regardless of whether or not they resulted in a workers’ compensation claim) through the agency’s Injury Tracking Application (ITA) [21, 22].
We geocoded every reported injury in the ITA dataset for the year 2023 and matched each case with high-resolution weather data across the contiguous United States. Our goal was to estimate the number and fraction of overall injuries that could be attributable to heat exposure. We focused specifically on the heat index and two thresholds (80 °F and 90 °F) identified by OSHA in its proposed (not yet finalized) federal rule [23] as regulatory triggers for workplace interventions such as rest breaks, water provision, and access to shade. In post-hoc analyses, we further examined injury patterns by industry sectors, states with/without heat rule, and by large U.S. census regions.
Methods
Injury data
Beginning in 2024, establishments began submitting detailed injury records for the prior year (2023) to OSHA’s Injury Tracking Application (https://www.osha.gov/injuryreporting/). Submission was mandatory for establishments with 100 or more employees in certain high-hazard industries [24], although some smaller establishments also submitted records voluntarily. An estimated 85% of covered establishments submitted the required data (personal communication with OSHA). The dataset includes the exact date of each incident, the establishment’s legal and company name, street address, industry classification, establishment size, the nature and outcome of the injury, and six narrative text fields filled by the employer describing each event. We restricted the dataset to incidents classified as “injury”. We treated each injury record as a distinct injury event, although the same worker could, in theory, contribute more than one record. To focus on acute outcomes, we excluded cases in which there was no single time-specific incident, such as hearing loss, skin disorders, COVID-19, and other illnesses. Vehicle crashes were not included in the dataset. Our analysis was limited to injuries reported by private industry; records from state and local government establishments were uneven across states and therefore excluded. Although the narrative text fields could be analyzed to further subclassify injuries (for example, traumatic vs. musculoskeletal), in this study we used any recordable injuries and did not subclassify by injury type.
To obtain a national estimate of all occupational injuries in private industry in the U.S. for the year 2023 (beyond those reported by large establishments with 100 or more employees in OSHA’s ITA), we extracted injury estimates from the Bureau of Labor Statistics (BLS) Survey of Occupational Injuries and Illnesses (SOII) [25]. The SOII provides annual national estimates of workplace injuries and illnesses, excluding only small farms with fewer than 11 employees, self-employed, and federal workers. However, SOII estimates are widely considered to undercount true injury totals by 20 to 60 percent [26, 27].
Spatiotemporal weather data
Establishment addresses from the OSHA ITA summary dataset were geocoded using the ArcGIS World Geocoding Service. We first merged individual address components (street, city, state, and ZIP code) into a single full-address string, which was then input into the geocoder to obtain latitude and longitude coordinates. Each result was assigned a match score from 0 to 100, with 100 indicating a perfect match. Of all geocoded records, 77% received a perfect score of 100, 19.5% scored between 95 and 99, and 1.2% and 1.8% fell into the 90–94 and 80–89 ranges, respectively. Only 0.1% of records had scores below 80. All records were retained for this analysis.
We used the Parameter-elevation Regressions on Independent Slopes Model (PRISM) to obtain rasterized daily weather data at 800-meter resolution across the contiguous United States for 2023 (https://prism.oregonstate.edu/). Daily weather values were taken from the centroid of the nearest grid cell to the geocoded latitude and longitude (no interpolation). Hawaii, Alaska and U.S. Island territories were excluded. Daily PRISM values for minimum and maximum temperatures (in °C), vapor pressure deficits (hPa), dew point temperatures (in °C), and precipitation (in millimeters) were extracted for each location in the OSHA dataset for the 6 days leading up to and including the reported accident using the ‘terra’ package in R. We calculated relative humidity from mean dew point temperature and mean ambient temperature using the August-Roche-Magnus approximation, as described by Alduchov and Eskridge (1996) [28] and McNoldy (https://bmcnoldy.earth.miami.edu). Heat index values (in °F) were then computed using the ‘weathermetrics’ package in R [29]using mean daily relative humidity and mean daily temperature. The heat index is a measure of how hot “it feels” when relative humidity is factored in with air temperature; for example, when the temperature is 86 °F with 70% humidity, the heat index is 95 °F [30]. We matched the daily mean heat index to each geocoded address based on exact longitudes and latitudes.
Study design
We applied a time-stratified case-crossover study design, an extension to the case-control design but with one key difference: each individual acts as their own control. This approach is useful when studying short-term or “transient” exposures (like a hot day) that might trigger sudden health events (such as an injury) [31, 32].
In our study, we looked at the weather on the day an injury occurred (the “case day”) and compared it to the weather on other days when no injury occurred for that same person (the “control days”). These control days were selected to match the day of the week, month, and year as the injury day. Because each person is compared to themselves, any individual and regional characteristics that do not change from day to day (such as where you live, where you work, age, sex, underlying health conditions, body weight, years of experience, skill level, etc.) are automatically controlled for, by design. We included control days both before and after the injury day to avoid bias that might occur when the weather follows a trend over time [33, 34]. Although severe injuries can remove some workers from risk on post-incident control days, the time-stratified case-crossover design has consistently produced unbiased estimates under exchangeability and limited bias from exposure time trends [34–36].
Statistical analysis
We fitted conditional logistic regression to estimate the odds ratio (OR) of injury across the distribution of the heat index. In general, the case-crossover OR approximates a rate ratio. Same-day heat index was modeled as a non-linear exposure using a natural spline with 3 degrees of freedom. The model adjusted for two independent variables: precipitation and public holidays in 2023. Restricting to summer months (June to August), odds ratios were estimated at extreme heat levels relative to 80 °F, which corresponds to the initial OSHA trigger threshold proposed in the federal heat rule. To estimate the full burden attributable to heat (number of injuries), we used pooled full-year data and calculated attributable injuries on all days where the heat index exceeded 70 °F, 80 °F, or 90 °F. Attributable injuries were computed by summing the daily contributions above a reference temperature [37]. We reported both population attributable fraction (PAF) and threshold-specific attributable fraction. For national burden estimates, we applied PAF derived from the ITA analysis to SOII national totals. Empirical 95% confidence intervals (eCI) around attributable numbers were calculated using 1,000 Monte Carlo simulations assuming a multivariate normal distribution. We also explored lagged effects of heat exposure using both moving averages (up to 7-days) and distributed lag non-linear models (DLNM) to assess cumulative impacts [38].
Post hoc stratifications
In addition to generating a national pooled estimate, we conducted several stratified analyses by U.S. Census regions (Midwest, Northeast, South, and West), by industry sector using 2-digit codes from the North American Industry Classification System (NAICS), and by comparing states with an occupational heat rule in place as of 2023 (California, Colorado, Minnesota, Oregon, and Washington) to those without. Colorado’s heat rule applied only to the agricultural sector, Washington’s and California’s heat rules applied only to outdoor workplaces, and Minnesota’s rule applied only to indoor workplaces. We included all five states together for our primary analysis. Because most heat rule states are in the West (except MN), lower risk in heat rule states could reflect regional differences (West vs. U.S.) rather than the effect of heat standards. To assess this, we repeated the analysis restricted to the U.S. Census West, comparing four heat rule states (CA, CO, OR, WA) with seven other Western states (AZ, ID, MT, NM, NV, UT, WY). Should a difference between these two groups persist within the West, a region-based explanation becomes unlikely. Finally, we stratified by serious injuries that resulted in days away from work, job restrictions, or transfer (DART).
All analyses were conducted using R [39] (version 4.4.2).
Results
A total of 845,014 occupational injuries were reported in OSHA’s ITA in 2023 (Table 1; summer injuries shown in Supplemental Table S1). Regionally, the highest number of injuries were recorded in the South (33.0%), followed by the Midwest (25.9%), West (24.3%), and Northeast (16.8%). The industry with highest count of injuries was Health Care and Social Assistance, accounting for 25.8% of all reported injuries, followed by Transportation and Warehousing (21.8%), Retail Trade (19.1%), and Manufacturing (18.1%). Manufacturing injuries were most frequent in the Midwest, while injuries in Agriculture, Forestry, Fishing, and Hunting were most common in the West. Injury counts distributed across months show a slight increase in the summer period (from June to August), compared to the rest of the years (Table 1). A time series of injuries, temperatures, relative humidity, and heat index are shown in Fig. 1. Occupational injuries were geographically widespread across the contiguous U.S. during the June to August period (Supplemental Figures A-B).
Table 1.
Distribution of occupational injuries recorded in OSHA’s injury tracking application for the year 2023
| Region* | Overall (N = 845,014) |
||||
|---|---|---|---|---|---|
| Midwest (N = 219,156) |
Northeast (N = 141,919) |
South (N = 278,637) |
West (N = 205,302) |
||
| NAICS 2-digit Sector | |||||
| Health Care and Social Assistance | 55,434 (25.3%) | 49,592 (34.9%) | 71,153 (25.5%) | 42,155 (20.5%) | 218,334 (25.8%) |
| Transportation and Warehousing | 39,762 (18.1%) | 29,327 (20.7%) | 64,214 (23.0%) | 50,656 (24.7%) | 183,959 (21.8%) |
| Retail Trade | 33,829 (15.4%) | 27,240 (19.2%) | 57,521 (20.6%) | 42,584 (20.7%) | 161,174 (19.1%) |
| Manufacturing | 62,099 (28.3%) | 18,836 (13.3%) | 47,513 (17.1%) | 24,374 (11.9%) | 152,822 (18.1%) |
| Wholesale Trade | 9,324 (4.3%) | 7,122 (5.0%) | 11,766 (4.2%) | 8,273 (4.0%) | 36,485 (4.3%) |
| Accommodation and Food Services | 2,668 (1.2%) | 3,764 (2.7%) | 5,879 (2.1%) | 9,673 (4.7%) | 21,984 (2.6%) |
| Arts, Entertainment, and Recreation | 1,359 (0.6%) | 1,069 (0.8%) | 6,863 (2.5%) | 5,165 (2.5%) | 14,456 (1.7%) |
| Agriculture, Forestry, Fishing and Hunting | 1,724 (0.8%) | 814 (0.6%) | 1,952 (0.7%) | 8,238 (4.0%) | 12,728 (1.5%) |
| Public Administration | 4,104 (1.9%) | 615 (0.4%) | 3,157 (1.1%) | 4,689 (2.3%) | 12,565 (1.5%) |
| Construction | 3,140 (1.4%) | 1,314 (0.9%) | 2,952 (1.1%) | 3,902 (1.9%) | 11,308 (1.3%) |
| Waste Management and Remediation Services | 1,952 (0.9%) | 1,118 (0.8%) | 2,296 (0.8%) | 2,406 (1.2%) | 7,772 (0.9%) |
| Educational Services | 1,879 (0.9%) | 572 (0.4%) | 1,298 (0.5%) | 723 (0.4%) | 4,472 (0.5%) |
| Utilities | 627 (0.3%) | 151 (0.1%) | 799 (0.3%) | 1,239 (0.6%) | 2,816 (0.3%) |
| Other | 1,255 (0.6%) | 385 (0.3%) | 1,274 (0.5%) | 1,225 (0.6%) | 4,139 (0.5%) |
| Month | |||||
| Jan | 18,490 (8.4%) | 11,817 (8.3%) | 22,515 (8.1%) | 17,105 (8.3%) | 69,927 (8.3%) |
| Feb | 17,719 (8.1%) | 11,008 (7.8%) | 21,460 (7.7%) | 15,569 (7.6%) | 65,756 (7.8%) |
| Mar | 19,331 (8.8%) | 12,200 (8.6%) | 23,931 (8.6%) | 17,762 (8.7%) | 73,224 (8.7%) |
| Apr | 16,873 (7.7%) | 11,393 (8.0%) | 22,360 (8.0%) | 16,317 (7.9%) | 66,943 (7.9%) |
| May | 18,602 (8.5%) | 12,159 (8.6%) | 23,989 (8.6%) | 17,578 (8.6%) | 72,328 (8.6%) |
| Jun | 18,677 (8.5%) | 12,373 (8.7%) | 24,414 (8.8%) | 17,557 (8.6%) | 73,021 (8.6%) |
| Jul | 18,796 (8.6%) | 12,521 (8.8%) | 24,869 (8.9%) | 18,411 (9.0%) | 74,597 (8.8%) |
| Aug | 20,157 (9.2%) | 12,720 (9.0%) | 25,898 (9.3%) | 19,100 (9.3%) | 77,875 (9.2%) |
| Sep | 18,072 (8.2%) | 11,839 (8.3%) | 23,478 (8.4%) | 17,090 (8.3%) | 70,479 (8.3%) |
| Oct | 18,944 (8.6%) | 12,210 (8.6%) | 23,893 (8.6%) | 17,676 (8.6%) | 72,723 (8.6%) |
| Nov | 17,640 (8.0%) | 11,302 (8.0%) | 21,859 (7.8%) | 16,174 (7.9%) | 66,975 (7.9%) |
| Dec | 15,855 (7.2%) | 10,377 (7.3%) | 19,971 (7.2%) | 14,963 (7.3%) | 61,166 (7.2%) |
*Regions are based on U.S. Census Bureau groupings
Fig. 1.
Time series of daily injuries and environmental exposures for the study period (Jan 1, 2023 to Dec 31, 2023)
On average, there were 3.44 controls per injury case. The control days were well-matched to case days and represented the distribution of the exposure in ambient temperature, relative humidity, and heat index (Table 2). Within matched sets, there was considerable exposure variability (Supplemental Table S2). Only 2.1% of location-days had clusters of more than two injuries on any given day.
Table 2.
Distribution of environmental exposures on case days vs. matched control days in time-stratified case-crossover design
| Cases | Control Days* | |
|---|---|---|
| Mean ± Standard Deviation [Min, Max] | ||
| All Months | (N= 845,014) | (N= 2,866,090) |
| Ambient Temperature (°F) | 59.1 ± 17.0 [−18.9, 105.5] | 59.2 ± 16.9 [−23.9, 106.1] |
| Relative Humidity (%) | 66.3 ± 16.8 [6.3, 100] | 66.3 ± 16.8 [6.3, 100] |
| Heat Index (°F) | 58.8 ± 18.2 [−19, 110] | 58.9 ± 18.1 [−24, 110] |
| Summer Months (Jun-Aug) | (N = 225,493) | (N = 1,008,987) |
| Ambient Temperature (°F) | 75.0 ± 8.5 [35.4, 105.5] | 75.0 ± 8.5 [34.0, 106.1] |
| Relative Humidity (%) | 65.2 ± 15.6 [6.3, 100] | 65.2 ± 15.6 [6.3, 100] |
| Heat Index (°F) | 76.4 ± 10.7 [35, 110] | 76.3 ± 10.7 [34, 110] |
*Each case day is matched to control days for the same individual on the same day of the week within the same month and year, when no event occurred
A dose-response relationship was observed between higher temperatures and injuries. The pooled national estimate increased non-linearly starting around 85 °F and accelerating above 90 °F (Fig. 2). Compared to 80 °F, the OR for injury at 90 °F was 1.03 (95% CI: 1.02, 1.04), rising to 1.06 (1.04, 1.08) at 95 °F, and 1.10 (1.07, 1.13) at 100 °F. The odds continued to rise at extreme heat levels, reaching 1.15 (1.10, 1.19) at 105 °F and 1.20 (1.13, 1.26) at 110 °F, compared to an 80 °F reference (Table 3).
Fig. 2.
Odds ratio of occupational injury by heat index across regions, sectors, and states with heat rules
Table 3.
Odds ratios (with 95% confidence intervals) for occupational injury at selected heat index levels (90, 95, 100, 105, and 110 °F) compared to heat index of 80 °F
| 90 °F (vs. 80 °F) |
95 °F (vs. 80 °F) |
100 °F (vs. 80 °F) |
105 °F (vs. 80 °F) |
110 °F (vs. 80 °F) |
|
|---|---|---|---|---|---|
| OR (95% CI) | OR (95% CI) | OR (95% CI) | OR (95% CI) | OR (95% CI) | |
| National | |||||
| Overall | 1.03 (1.02, 1.04) | 1.06 (1.04, 1.08) | 1.10 (1.07, 1.13) | 1.15 (1.10, 1.19) | 1.20 (1.13, 1.26) |
| Region | |||||
| Midwest | 1.05 (1.02, 1.08) | 1.09 (1.05, 1.14) | 1.15 (1.08, 1.22) | 1.21 (1.11, 1.32) | 1.28 (1.14, 1.43) |
| Northeast | 0.97 (0.88, 1.06) | 0.95 (0.82, 1.10) | Temp not reached | Temp not reached | Temp not reached |
| South | 1.01 (1.00, 1.03) | 1.04 (1.02, 1.06) | 1.08 (1.05, 1.11) | 1.13 (1.07, 1.19) | 1.18 (1.09, 1.28) |
| West | 1.03 (0.99, 1.08) | 1.06 (0.99, 1.14) | 1.09 (0.98, 1.21) | 1.12 (0.97, 1.29) | 1.15 (0.96, 1.38) |
| Heat Rule* | |||||
| States without Heat Rule | 1.02 (1.01, 1.03) | 1.05 (1.04, 1.07) | 1.10 (1.07, 1.13) | 1.16 (1.11, 1.21) | 1.22 (1.15, 1.29) |
| States with Heat Rule | 1.03 (0.97, 1.1) | 1.04 (0.94, 1.16) | 1.06 (0.91, 1.23) | 1.08 (0.88, 1.32) | 1.09 (0.84, 1.41) |
| NAICS 2-digit Sector | |||||
| Health Care and Social Assistance | 1.00 (0.97, 1.03) | 1.01 (0.97, 1.05) | 1.02 (0.97, 1.08) | 1.04 (0.96, 1.13) | 1.06 (0.95, 1.19) |
| Transportation and Warehousing | 1.05 (1.02, 1.07) | 1.09 (1.05, 1.13) | 1.15 (1.09, 1.21) | 1.21 (1.12, 1.32) | 1.29 (1.16, 1.44) |
| Retail Trade | 1.02 (0.99, 1.05) | 1.04 (1.00, 1.08) | 1.05 (0.99, 1.12) | 1.07 (0.98, 1.18) | 1.09 (0.97, 1.23) |
| Manufacturing | 1.05 (1.02, 1.09) | 1.11 (1.06, 1.16) | 1.17 (1.10, 1.26) | 1.25 (1.14, 1.38) | 1.34 (1.19, 1.52) |
| Wholesale Trade | 1.01 (0.95, 1.08) | 1.05 (0.96, 1.14) | 1.11 (0.98, 1.26) | 1.18 (0.98, 1.41) | Temp not reached |
| Accommodation and Food Services | 0.98 (0.89, 1.07) | 0.96 (0.85, 1.07) | 0.93 (0.79, 1.1) | 0.91 (0.72, 1.14) | Temp not reached |
| Arts, Entertainment, and Recreation | 0.96 (0.86, 1.07) | 0.98 (0.87, 1.12) | 1.04 (0.87, 1.23) | 1.11 (0.86, 1.43) | Temp not reached |
| Agriculture, Forestry, Fishing and Hunting | 1.14 (1.00, 1.30) | 1.23 (1.00, 1.51) | 1.33 (0.98, 1.81) | 1.45 (0.94, 2.23) | 1.55 (0.90, 2.65) |
| Public Administration | 1.05 (0.93, 1.19) | 1.09 (0.89, 1.33) | 1.13 (0.84, 1.53) | 1.18 (0.77, 1.79) | Temp not reached |
| Construction | 1.07 (0.97, 1.19) | 1.12 (0.97, 1.31) | 1.19 (0.95, 1.49) | 1.26 (0.91, 1.73) | 1.27 (0.90, 1.79) |
| Waste Management and Remediation Services | 1.11 (0.97, 1.27) | 1.23 (1.02, 1.48) | 1.40 (1.07, 1.83) | 1.61 (1.11, 2.35) | 1.82 (1.13, 2.92) |
| Educational Services | 1.27 (0.99, 1.62) | 1.59 (1.06, 2.37) | 2.08 (1.12, 3.87) | 2.80 (1.16, 6.74) | Temp not reached |
| Utilities | 0.97 (0.78, 1.2) | 0.96 (0.71, 1.30) | 0.96 (0.60, 1.51) | 0.95 (0.50, 1.83) | Temp not reached |
| Other | 1.08 (0.92, 1.28) | 1.19 (0.96, 1.49) | 1.35 (0.96, 1.89) | 1.54 (0.94, 2.53) | Temp not reached |
*States with heat standards as of 2023 are: CA, CO, MN, OR, and WA
Stratified analyses showed that nearly all industry sectors experienced increased odds of injury with rising temperatures. Among industries that were exposed to very hot temperatures (105 to 110 °F), the largest ORs were observed in Waste Management and Remediation Services, followed by Agriculture, Manufacturing, Transportation and Warehousing, and Construction. Regionally, the Midwest and South exhibited steeper increases in injuries compared to the Northeast, which showed a flatter curve; however, heat index values above 96 °F were not observed in the Northeast, and odds ratios were not statistically significant at higher temperatures (confidence intervals included 1).
The rise in odds of injury at extreme heat levels was higher in states without occupational heat rules. On a 105 °F day (vs. 80 °F), the odds of injury increased by 16% in states without heat rules (OR = 1.16, 95% CI: 1.11–1.21), compared to 8% in states with heat rules (OR = 1.08, 95% CI: 0.88–1.32) (Fig. 2; Table 3). At 110 °F, the odds rose by 22% (OR = 1.22, 95% CI: 1.15–1.29) in states without rules, versus 9% (OR = 1.09, 95% CI: 0.84–1.41) in states with rules. Heat rule states in the West (CA, CO, OR, WA) still showed attenuated odds ratios at high heat relative to states without rules in the same region (AZ, ID, MT, NM, NV, UT, WY) (Supplemental Figure C).
Results from models using up to 7-day moving averages (Supplemental Figure D) and distributed lag non-linear models were consistent with the main findings. DART (Days Away, Restricted, or Transferred) injuries showed a similar dose-response relationship to non-DART injuries (Supplemental Figure E).
We estimate that 1.18% (95% eCI: 0.92–1.45%) of all injuries were attributable to days with heat index ≥ 70 °F (Table 4). The population attributable fraction at ≥ 80 °F was 0.69% (95% eCI: 0.52–0.85%) and 0.33% (95% eCI: 0.24–0.41%) at ≥ 90 °F. When applying the same fractions to BLS’s national injury totals (2,368,900 injuries in 2023), the estimated burden was 27,953 (95% eCI: 21,794 − 34,349) injuries attributed to heat index ≥ 70 °F (Supplemental Table 3). Full exposure range relationship across all temperatures in summer and year-round are shown in Supplemental Figure F. There was a U-shaped pattern with increases in injuries at both high and low temperatures.
Table 4.
Estimated heat-attributable injuries and fractions were calculated using the 2023 OSHA ITA dataset (empirical 95% confidence intervals [eCI] were calculated using 1,000 Monte-Carlo simulations)
| Estimated Attributable Injuries to Heat | Total Injuries ≥ Threshold |
Estimated Threshold-specific Attributable Fraction (%) ‡ | Total Injuries in Dataset | Estimated Population Attributable Fraction (PAF%) † | |
|---|---|---|---|---|---|
| N [95% eCI] | N | % [95% eCI] | N | PAF % [95% eCI] | |
| Threshold (Heat Index) | |||||
| ≥ 60 °F | 11,238 [8,788 − 13,858] | 430,679 | 2.61% [2.04% − 3.22%] | 845,014 | 1.33% [1.04% − 1.64%] |
| ≥ 70 °F | 9,971 [7,774 − 12,253] | 262,123 | 3.80% [2.97% − 4.67%] | 845,014 | 1.18% [0.92% − 1.45%] |
| ≥ 80 °F | 5,831 [4,394–7,183] | 100,633 | 5.79% [4.37% − 7.14%] | 845,014 | 0.69% [0.52% − 0.85%] |
| ≥ 90 °F | 2,789 [2,028 − 3,465] | 36,313 | 7.68% [5.58% − 9.54%] | 845,014 | 0.33% [0.24% − 0.41%] |
‡Threshold-specific Attributable Fraction = (Estimated attributable injuries/injuries occurring at or above threshold) × 100
†Population Attributable Fraction = (Estimated attributable injuries/total injuries in dataset) × 100
Discussion
This study provides the first national assessment of the work injury burden attributable to hot temperatures in the contiguous United States. We found that even moderate daily heat can subtly increase the risk of workplace injuries that are not thought of nor classified as “heat-related”. A clear dose-response relationship was observed where odds of injury began rising around a heat index of 85 °F and accelerated at more extreme heat. By 90 °F, the odds of injury were a few percent higher, and on a 105 °F and 110 °F days the odds were about 15% and 20% higher compared to an 80 °F day, respectively. Across all industries and regions, we estimate roughly 1.18% of occupational injuries in 2023 were attributable to days with heat index above 70 °F. In absolute terms, if we use BLS national injury totals for the same year, we arrive at 27,953 extra injuries nationwide attributed to heat. Notably, we found these injuries occur across all major industry sectors, including predominantly indoor workplaces. The overall risks of injury were lower in states with workplace rules protecting workers from exposure to extreme heat.
Our estimate of 27,953 excess injuries (95% eCI: 21,794 − 34,349) in 2023 was derived by applying effect estimates from OSHA’s large-establishment injury data (ITA) to the national total of workplace injuries (BLS). This is likely a significant undercount of the actual number of injuries that were associated with work on hotter days. BLS commissioned a series of studies to estimate the completeness of the employer recorded work injury and illness logs that serve as the basis of the SOII, and found the size of the undercount varied, with estimates ranging from 20 to 70% of cases depending upon the research methodology, the types of injuries examined, and state studied [26, 27]. In California, Heinzerling et al. [40] found that the number of heat-related injuries, in particular, were three to six times lower in BLS SOII compared to what was identified in the state’s Workers’ Compensation Information System. Park et al. [14] found that in California alone, high ambient temperatures (above 60 °F) likely caused about 20,000 extra injuries annually. Using a heat index (which incorporates humidity) of ≥ 60 °F we arrived at 31,506 annual injuries (95% eCI: 24,637 − 38,850), nationally in 2023. OSHA’s own regulatory economic analysis for its proposed heat rule applied a 7.5-fold underreporting factor to the BLS SOII count of non-fatal heat injuries (3,288 per year) reaching a tally of 24,656 injuries attributable to heat per year (OSHA applied a range from 6,575 (2-fold) to 32,875 (10-fold)) [41]. Our analysis agrees with OSHA’s central estimate. However, we find OSHA’s lower estimate of 6,575 injuries per year to be extremely unlikely.
Our results were remarkably consistent across different industries. The impact of heat is pervasive, with nearly every sector we examined showing increased injury risk on hotter days. In our data, industries with substantial outdoor work (e.g. waste management, agriculture, construction) unsurprisingly showed large heat-related effects. However, even industries with predominantly indoor work such as manufacturing and warehousing showed injury increases at high heat levels. Only two sectors (Utilities, and Accommodation & Food Service) did not exhibit a clear heat-injury association, and in those two sectors the confidence intervals were wide due to small sample sizes. The overall pattern suggests a fairly universal physiological response to heat stress: when workers become thermally uncomfortable, their motor skills, balance, and cognitive function can degrade, leading to mistakes or accidents across diverse job types and industries. Park et al. in California reported similar magnitudes of risk increase (about 10–15% increase in injuries on 100+°F days) across a wide range industry sectors beyond traditionally hot outdoor jobs [14]. Similarly, two multi-state workers’ compensation claims analyses showed increases in injuries with extreme heat across all sectors examined [18, 19]. These studies also suggested that the heat-injury association is stronger for traumatic injuries than for musculoskeletal injuries such as sprains and strains [18, 19].
We also found suggestive evidence that existing heat-safety regulations mitigate some of the risk. In states with occupational heat standards in place (California, Colorado, Minnesota, Oregon, and Washington), the heat–injury relationship appeared attenuated compared to states without such rules. For example, on a 105 °F heat index day, we estimated the odds of injury rose by roughly 16% in states lacking a heat rule, versus about 8% in states with a rule in place. At 110 °F, it was 22% vs. 9%, respectively. Underestimation is also possible because some state rules were recently adopted (WA, OR) and others have sector-limited scope (CO), which can dilute pooled effects. Although the confidence intervals overlapped, given only one year of data, the trend was consistent with a protective effect of heat regulations. This observation aligns with California’s experience: after implementing its outdoor heat Illness Prevention standard in 2005, the rate of heat-related work injuries in California declined by an estimate of 30%14, and following additional regulatory enhancements in 2015, heat-related deaths may have been reduced by as much as 43% [42]. In the sensitivity analysis, we examined California alone. We found the estimated increase in injury odds at 110 °F (vs. 80 °F) was 14% in California, compared to 22% in states without heat rules, representing a ~ 36% relative reduction (Supplemental Table S4). Although some of these rules were limited to certain indoor or outdoor settings, they may influence broader workplace practices or encourage increased attention to heat precautions. A longer follow-up and formal policy evaluation (including for specific provisions: e.g., acclimatization, trigger thresholds) would be useful to confirm the extent of injury risk reduction achievable from heat regulatory measures.
Our analysis has several limitations. First, our study covered only a single year. This one-year snapshot may not capture atypical conditions or longer-term trends. Future research should incorporate multiple years of data as they become available through the ITA. Second, the OSHA injury data we used are not a complete census of all workplaces. Reporting was mandatory only for establishments with ≥ 100 employees in designated high-hazard industries. BLS employment totals for 2023 shows that only 42% of U.S. employment in 2023 was in large establishments with ≥ 100 employees (analysis not shown). Smaller businesses (which often lack extensive safety resources) were not required to submit injury data. If heat disproportionately affects workers in small establishments because these firms lack safety personnel or other resources typically available in larger workplaces, our results would underestimate the consequences of working in hot environments. Similarly, not all qualifying establishments reported as required in 2023 (despite the mandate), leading to underreporting of injuries in the dataset. Third, our exposure metric was the daily heat index at an 800-meter grid cell for each establishment’s location. While this high-resolution gridded weather data provides state-of-the-science information about environmental exposure, it cannot account for all workplace microclimate variations or individual exposures. Some workers might have experienced higher heat (e.g., working next to heat-generating machinery or in direct sun), whereas others (like those in air-conditioned or cooled environments) had lower personal exposure than the ambient heat index suggests. We also matched injuries to the establishment address, which in some cases might differ from the exact incident location (for example, a delivery driver’s injury might occur off-site). We also used daily mean heat index and did not account for hour-specific exposure or for time of day or shift at the time of injury. These exposure misclassification issues are likely non-differential, which typically biases risk estimates toward the null rather than exaggerating them. In other words, if we had precise, individual-level heat exposure data, the true associations might be even stronger. All these limitations suggest that our estimates may underestimate the true burden. Finally, although the case-crossover design inherently controls for all time-invariant confounders, we lacked data on certain potentially important modifiers such as worker characteristics like age, experience, sex, or use of medications, as well as workplace characteristics such as provision of shade, breaks, and hydration.
Conclusion
The main findings from this analysis are: (1) work injury risks begin to rise when the heat index reaches around 85 °F and increase sharply after 90 °F; (2) the effect is seen across nearly all industry sectors, (3) these industry-wide impacts translate to an estimated 1.18% of all injuries in 2023 (approximately 28,000 cases, if we use BLS denominator) being attributable to heat; and (4) workplace state rules protecting workers from exposure to extreme heat appear to attenuate the risk.
The findings demonstrate that hot temperatures are not just a health concern for heat illness or heat stroke, but also a significant risk factor for a wide array of workplace injuries. These injuries are typically not labeled as “heat-related”, reflecting a subtle yet pervasive impact of heat on worker vigilance, coordination, and judgment. Occupational safety training and education requirements should explicitly warn about the role of heat in such injuries. Heat safety measures by employers (acclimatization, rest, water, shade, written heat plans, etc.) will be critical for safeguarding worker health, and reducing the 'hidden' economic costs of heat-related injuries. As climate change intensifies and extreme heat events become more frequent and pervasive, adopting and enforcing comprehensive heat safety regulations will become increasingly important. Our study provides a quantitative baseline for these efforts and highlights the need for continued surveillance of injuries and research to monitor the effectiveness of heat interventions that best protect the workforce.
Supplementary Information
Authors’ contributions
Conceptualization, supervision, and results interpretation: GRW, DM - Manuscript first draft: BA - Study design: BA, JS - Data analysis: BA - Injury dataset geocoding: BA, YA - Weather data curation and matching: BA, WK - Manuscript review and approval: All authors.
Funding
Financial support for this study was provided by the McElhattan Foundation.
Data availability
No datasets were generated or analysed during the current study.
Declarations
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Piil JF, Lundbye-Jensen J, Trangmar SJ, Nybo L. Performance in complex motor tasks deteriorates in hyperthermic humans. Temp (Austin). 2017;4:420–8. 10.1080/23328940.2017.1368877. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Distefano LJ, Casa DJ, Vansumeren MM, Karslo RM, Huggins RA, Demartini JK, et al. Hypohydration and hyperthermia impair neuromuscular control after exercise. Med Sci Sports Exerc. 2013;45:1166–73. 10.1249/MSS.0b013e3182805b83. [DOI] [PubMed] [Google Scholar]
- 3.Chen M-L, Chen C-J, Yeh W-Y, Huang J-W, Mao I-F. Heat stress evaluation and worker fatigue in a steel plant. AIHA J (Fairfax Va). 2003;64:352–9. 10.1080/15428110308984827. [DOI] [PubMed] [Google Scholar]
- 4.Smallcombe JW, Foster J, Hodder SG, Jay O, Flouris AD, Havenith G. Quantifying the impact of heat on human physical work capacity; part IV: interactions between work duration and heat stress severity. Int J Biometeorol. 2022;66:2463–76. 10.1007/s00484-022-02370-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Yeoman K, Weakley A, DuBose W, Honn K, McMurry T, Eiter B, et al. Effects of heat strain on cognitive function among a sample of miners. Appl Ergon. 2022;102:103743. 10.1016/j.apergo.2022.103743. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Mazloumi A, Golbabaei F, Mahmood Khani S, Kazemi Z, Hosseini M, Abbasinia M, Dehghan F, S. Evaluating effects of heat stress on cognitive function among workers in a hot industry. Health Promot Perspect. 2014;4:240–6. 10.5681/hpp.2014.031. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Sun G, Yang X, Jiang Q, Liu K, Li B, Li L, et al. Hyperthermia impairs the executive function using the attention network test. Int J Hyperthermia. 2012;28:621–6. 10.3109/02656736.2012.705217. [DOI] [PubMed] [Google Scholar]
- 8.Sun G, Li L, Li M, Jiang Q. Hyperthermia impaired pre-attentive processing: an auditory MMN study. Neurosci Lett. 2011;502:94–8. 10.1016/j.neulet.2011.07.029. [DOI] [PubMed] [Google Scholar]
- 9.Racinais S, Gaoua N, Grantham J. Hyperthermia impairs short-term memory and peripheral motor drive transmission. J Physiol (Lond). 2008;586:4751–62. 10.1113/jphysiol.2008.157420. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Cedeño Laurent JG, Williams A, Oulhote Y, Zanobetti A, Allen JG, Spengler JD. Reduced cognitive function during a heat wave among residents of non-air-conditioned buildings: an observational study of young adults in the summer of 2016. PLoS Med. 2018;15:e1002605. 10.1371/journal.pmed.1002605. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Ramsey JD, Burford CL, Beshir MY, Jensen RC. Effects of workplace thermal conditions on safe work behavior. J Saf Res. 1983;14:105–14. 10.1016/0022-4375(83)90021-X. [Google Scholar]
- 12.Heyes A, Saberian S. Temperature and decisions: evidence from 207,000 court cases. Am Economic Journal: Appl Econ. 2019;11:238–65. 10.1257/app.20170223. [Google Scholar]
- 13.Spector JT, Masuda YJ, Wolff NH, Calkins M, Seixas N. Heat exposure and occupational injuries: review of the literature and implications. Curr Environ Health Rep. 2019;6:286–96. 10.1007/s40572-019-00250-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Park J, Pankratz NMC, Behrer A. Temperature, workplace safety, and labor market inequality. SSRN J. 2021. 10.2139/ssrn.3892588. [Google Scholar]
- 15.Spector JT, Bonauto DK, Sheppard L, Busch-Isaksen T, Calkins M, Adams D, et al. A case-crossover study of heat exposure and injury risk in outdoor agricultural workers. PLoS ONE. 2016;11:e0164498. 10.1371/journal.pone.0164498. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Calkins MM, Bonauto D, Hajat A, Lieblich M, Seixas N, Sheppard L, et al. A case-crossover study of heat exposure and injury risk among outdoor construction workers in Washington state. Scand J Work Environ Health. 2019;45:588–99. 10.5271/sjweh.3814. [DOI] [PubMed] [Google Scholar]
- 17.Dillender M. Climate change and occupational health. J Hum Resour. 2021;56:184–224. 10.3368/jhr.56.1.0718-9594R3. [Google Scholar]
- 18.Negrusa S, Fomenko O, Thumula V. Impact of excessive heat on the frequency of Work-Related injuries. Workers Compensation Research Institute (WCRI); 2024. https://www.wcrinet.org/images/uploads/files/wcri3312.pdf.
- 19.Coate P, Scott C, Colon D. Adverse Weather and Workers Compensation Claims (National Council on Compensation Insurance (NCCI). 2024. https://www.ncci.com/Articles/Documents/Adverse-Weather-and-Workers-Compensation-Claims.pdf.
- 20.OSHA. Final Rule: Improve Tracking of Workplace Injuries and Illnesses. 2023. https://www.osha.gov/recordkeeping/final-rule. [PubMed]
- 21.Michaels D, Wagner GR. OSHA injury data: an opportunity for improving work injury prevention. Am J Public Health. 2025;115:588–95. 10.2105/AJPH.2024.307934. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.OSHA. Injury Tracking Application (ITA) Data: Establishment and Case Detail Work-related Injury and Illness Data. Occupational Safety and Health Administration. 2025. https://www.osha.gov/Establishment-Specific-Injury-and-Illness-Data.
- 23.OSHA. (2024). Notice for Proposed Rulemaking (NPRM): Heat Injury and Illness Prevention in Outdoor and Indoor Work Settings Rulemaking. https://www.osha.gov/heat-exposure/rulemaking.
- 24.OSHA. Appendix B to subpart E of part 1904 — Designated industries. Occupational safety and health administration. 2023. https://www.osha.gov/laws-regs/regulations/standardnumber/1904/1904SubpartEAppB.
- 25.Bureau of Labor Statistics. 2.6 million workplace injuries and illnesses in private industry in 2023, down 8.4% from 2022. U.S. Bureau of Labor Statistics (BLS). 2024. https://www.bls.gov/opub/ted/2024/2-6-million-workplace-injuries-and-illnesses-in-private-industry-in-2023-down-8-4-percent-from-2022.htm.
- 26.Wiatrowski W. Examining the completeness of occupational injury and illness data: an update on current research. MLR. 2014. 10.21916/mlr.2014.24. [Google Scholar]
- 27.Spieler EA, Wagner GR. Counting matters: implications of undercounting in the BLS survey of occupational injuries and illnesses. Am J Ind Med. 2014;57:1077–84. 10.1002/ajim.22382. [DOI] [PubMed] [Google Scholar]
- 28.Alduchov OA, Eskridge RE. Improved Magnus form approximation of saturation vapor pressure. J Appl Meteor. 1996;35:601–9. 10.1175/1520 – 0450(1996)035 < 0601:IMFAOS > 2.0.CO;2. [Google Scholar]
- 29.Anderson GB, Bell ML, Peng RD. Methods to calculate the heat index as an exposure metric in environmental health research. Environ Health Perspect. 2013;121:1111–9. 10.1289/ehp.1206273. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.National Weather Service. Heat Index Calculator: How do we calculate the heat index? NOAA/ National Weather Service, National Centers for Environmental Prediction, Weather Prediction Center. 2023. https://www.wpc.ncep.noaa.gov/html/heatindex.shtml.
- 31.Maclure M. The case-crossover design: a method for studying transient effects on the risk of acute events. Am J Epidemiol. 1991;133:144–53. 10.1093/oxfordjournals.aje.a115853. [DOI] [PubMed] [Google Scholar]
- 32.Maclure M, Mittleman MA. Should we use a case-crossover design? Annu Rev Public Health. 2000;21:193–221. 10.1146/annurev.publhealth.21.1.193. [DOI] [PubMed] [Google Scholar]
- 33.Lee JT, Kim H, Schwartz J. Bidirectional case-crossover studies of air pollution: bias from skewed and incomplete waves. Environ Health Perspect. 2000;108:1107–11. 10.1289/ehp.001081107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Janes H, Sheppard L, Lumley T. Case-crossover analyses of air pollution exposure data: referent selection strategies and their implications for bias. Epidemiology. 2005;16:717–26. 10.1097/01.ede.0000181315.18836.9d. [DOI] [PubMed] [Google Scholar]
- 35.Carracedo-Martínez E, Taracido M, Tobias A, Saez M, Figueiras A. Case-crossover analysis of air pollution health effects: a systematic review of methodology and application. Environ Health Perspect. 2010;118:1173–82. 10.1289/ehp.0901485. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Tobias A, Kim Y, Madaniyazi L. Time-stratified case-crossover studies for aggregated data in environmental epidemiology: a tutorial. Int J Epidemiol. 2024. 10.1093/ije/dyae020. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Gasparrini A, Leone M. Attributable risk from distributed lag models. BMC Med Res Methodol. 2014. 10.1186/1471-2288-14-55. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Gasparrini A. Distributed lag linear and non-linear models in R: the package Dlnm. J Stat Softw. 2011;43:1–20. [PMC free article] [PubMed] [Google Scholar]
- 39.R Core Team. R: A Language and environment for statistical computing. R Foundation for Statistical Computing; 2024.
- 40.Heinzerling A, Laws RL, Frederick M, Jackson R, Windham G, Materna B, et al. Risk factors for occupational heat-related illness among California workers, 2000–2017. Am J Ind Med. 2020;63:1145–54. 10.1002/ajim.23191. [DOI] [PubMed] [Google Scholar]
- 41.OSHA. Federal Register - Heat injury and illness prevention in outdoor and indoor work settings notice of proposed rulemaking (NPRM) and request for comments. Labor: AGENCY: Occupational Safety and Health Administration (OSHA); 2024. [Google Scholar]
- 42.Dean A, McCallum J. The Effect of Heat Standards on Deaths Related to Outdoor Work (Center of Labor & Just Economy - at Harvard Law School). 2025. https://www.afscme.org/Dean%20Adam%20and%20McCallum%20Jamie%20Heat%20Study%20v3%202025_formatted.pdf.
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
No datasets were generated or analysed during the current study.