Arizona HB 2684 Forces WC Actuaries to Price Heat Exposure

From tracking NCCI's emerging issues reports across filing cycles, the 2026 edition elevates heat-related workplace injuries to a top emerging legislative trend for the first time, and the timing forces a pricing conversation the workers compensation system has deferred for years. Arizona HB 2684 would require employers to develop written heat-illness mitigation programs for any employee exposed to temperatures exceeding 80°F. OSHA's proposed Heat Injury and Illness Prevention standard, published in the Federal Register on September 25, 2025, would impose nationwide prevention obligations triggered at an 80°F heat index. New Hampshire HB 1451 proposes a Workplace Extreme Temperatures Protection Standards Act covering both heat and cold stress. At the federal level, HR 3702 directs NOAA to study extreme heat's economic costs using workers' compensation claims data. These are not abstract regulatory proposals. They are converging on a single pricing question: what is the explicit loss cost of occupational heat exposure, and how should it flow through classification ratemaking?

The Data Problem: Heat Losses Hide in Aggregate Class Experience

Heat-related injuries span dozens of NCCI classification codes. Construction classes (5XXX series), agriculture (0XXX), landscaping, utilities, and transportation all carry material outdoor heat exposure. But no single classification code isolates heat as the defining hazard. A roofer in class 5551, a landscaper in class 0042, and a utility lineworker in class 7539 all face heat risk, but their class loss experience aggregates heat injuries with falls, lacerations, strains, and every other cause of loss in the code.

This aggregation creates a structural identification problem. BLS data documents approximately 34,000 heat-related injuries and 479 worker fatalities between 2011 and 2022. But these figures almost certainly understate the true exposure. Many heat-triggered claims, including cardiac events, heat stroke, and rhabdomyolysis, are coded to the underlying medical diagnosis rather than to heat as the cause of injury. A construction worker who suffers a heat-induced cardiac arrest gets coded as a cardiac claim. A warehouse employee with exertional rhabdomyolysis gets coded as a musculoskeletal injury. The heat-causal link is real, but the coding system obscures it, diluting the pricing signal within each classification's aggregate loss experience.

NCCI's cause-of-injury coding does include Nature of Injury codes for heat prostration and environmental exposure. These codes capture the claims that are explicitly identified as heat-related. But the miscoded claims, where heat was the precipitating factor but not the recorded cause, sit outside this subset. Research on 11 million California workers' compensation claims from 2011 to 2018 found that workers face up to a 9% greater chance of injuries on days above 90°F compared to temperate days in the 50s and 60s, a figure that captures the all-cause frequency elevation attributable to heat, including the injuries that never get coded as heat-related.

Step 1: Isolating Heat-Attributable Losses from Class Experience

The first step in building an explicit heat-exposure loading is separating heat-attributable losses from the aggregate class loss experience that currently absorbs them. Start with NCCI's Nature of Injury codes for heat prostration (code 070) and environmental exposure. This provides the directly coded baseline. Then apply a supplementary allocation to capture the miscoded claims.

Two external benchmarks can calibrate the supplementary allocation. OSHA 300 log data, which employers are required to maintain and which OSHA collects during inspections, provides establishment-level heat injury counts by industry. BLS Survey of Occupational Injuries and Illnesses (SOII) publishes heat-specific injury counts by industry sector annually. The ratio of BLS SOII heat injuries to NCCI Nature of Injury heat claims, by classification grouping, estimates the miscoding factor. If BLS records 500 heat injuries in construction nationally but NCCI's cause-of-injury data shows only 300 heat-coded claims in construction classifications, the implied miscoding factor is 1.67x, meaning direct heat claims should be grossed up by 67% to approximate total heat-attributable losses.

This approach has obvious credibility limitations. The grossing-up factor varies by classification, by state, and by the stringency of local reporting requirements. But even an imprecise isolation is better than the current default of zero explicit heat loading, which implicitly assumes heat losses are adequately captured in aggregate class experience despite systematic undercounting.

Step 2: Building a Climate-Adjusted Frequency Trend

Standard NCCI loss cost filings use historical loss frequency to project prospective frequency trends, typically through a least-squares regression on 10 years of accident year data. For most perils, this approach works because the underlying hazard is approximately stationary: slip-and-fall frequency does not exhibit a secular trend driven by an external physical variable. Heat exposure is different. The frequency of extreme heat days is accelerating, not mean-reverting, and the acceleration is documented by NOAA's Climate Extremes Index and annual heat-day counts at the regional level.

A pricing actuary constructing a prospective heat-frequency trend should contrast two approaches. The first is the standard regression on 10-year historical heat claim frequency within the relevant classification group. This produces a trend, but it systematically understates the prospective hazard because the observation period reflects a climate that was cooler than the climate the prospective rate period will experience. Arizona recorded its fourth-hottest summer on record in 2025, with nearly 500,000 workers in heat-vulnerable occupations exposed weekly during peak months.

The second approach supplements the historical loss frequency with a climate-adjusted projection. Using NOAA's regional extreme-heat-day forecasts (days exceeding a specified heat index threshold), the actuary can construct an exposure-adjusted frequency trend that accounts for the increasing number of heat-risk days per year. If NOAA projects a 3% annual increase in days exceeding a 100°F heat index in Arizona over the next decade, the prospective heat-claim frequency trend should incorporate that acceleration rather than relying solely on backward-looking loss data.

The gap between these two approaches is the non-stationarity adjustment. For classifications with heavy outdoor exposure in states experiencing the sharpest climate warming, the gap can be material, potentially adding 1 to 3 percentage points of annual frequency trend above the historical regression estimate. This is the same conceptual challenge that NCCI's medical price-utilization gap presents in severity trend selection: the historical data is measuring the right thing, but not enough of it.

Step 3: Modeling the Presumption-Law Frequency Shift

Arizona HB 2684's 80°F mitigation trigger and OSHA's proposed 80°F heat index threshold create a bright-line compliance standard with direct compensability implications. Once enacted, claims currently denied under a general-duty defense could become presumptively compensable where employers lack documented heat-illness prevention programs. This is a frequency step-change, not a trend change: the pool of compensable claims expands discretely when the legal standard shifts.

California's experience provides the closest available calibration. When California adopted its 2005 heat illness prevention standard, the first state-level heat regulation in the country, heat-related workers' compensation claims became more clearly identifiable, and approximately 1,000 California workers per year now submit heat-related WC claims. The frequency loading factor for a presumption-law shift can be modeled as the ratio of post-adoption compensable heat claims to pre-adoption claims within the same classification group, adjusting for secular frequency trends and employment growth.

The California data suggests the frequency effect operates through two channels. First, claims that would have been denied under the prior legal framework become accepted when employers cannot demonstrate compliance with explicit prevention standards. Second, awareness of the regulatory requirement increases reporting rates among workers who previously would not have filed. Both channels increase measured frequency. For states considering legislation similar to Arizona HB 2684, the pricing actuary should apply a presumption-law frequency adjustment as a multiplicative factor on the climate-adjusted trend from Step 2, calibrated using jurisdictions where similar standards have been in force long enough to produce credible pre/post comparisons.

California's vetoed AB 1336 in 2025 would have gone further, creating a presumption that heat-related injuries in agriculture following employer noncompliance arose from employment. Had it passed, the frequency shift would have been even larger. The veto signals political limits, but the legislative direction is clear: more states will adopt explicit heat standards, and each adoption event will produce a discrete upward shift in heat-claim frequency.

Step 4: Severity Adjustment for Outdoor Classifications

Heat exposure does not just increase claim frequency; it extends disability duration in ways that differ from the typical workers' compensation claim. When a construction laborer or agricultural worker suffers a heat illness severe enough to generate a lost-time claim, the return-to-work timeline is constrained by the same environmental condition that caused the injury. An office worker recovering from a sprain can return to light duty at a desk. An outdoor worker recovering from heat stroke cannot return to light duty in a 105°F field.

This structural limitation inflates indemnity severity for heat claims in outdoor classifications relative to the class average. WCRI disability duration benchmarks by occupation group document the differential: outdoor trades with limited light-duty alternatives consistently produce longer disability durations for equivalent injury severity levels. The severity trend adjustment for the heat-exposure loading should reflect this differential using WCRI's occupation-specific duration data rather than the class-average severity trend.

The compounding effect matters. When both frequency (more claims accepted under presumption laws) and severity (longer durations due to limited return-to-work options) increase simultaneously, the impact on class loss costs compounds multiplicatively. A 15% frequency increase combined with a 10% severity increase does not produce a 25% loss cost increase; it produces a 26.5% increase. For classifications where heat exposure is a material portion of total losses, the compounding effect can shift the loss cost indication by several percentage points.

Experience Rating Feedback: When Heat Claims Hit Primary Territory

NCCI's experience rating plan underwent a significant structural change in November 2023, moving from a uniform $18,500 countrywide split point to state-specific split points. As of 2026, these split points range from approximately $9,500 in Oregon to $38,000 in Louisiana, with most states falling between $15,000 and $25,000. This creates a state-dependent interaction between heat claim severity and the experience modification.

Heat illness claims tend to cluster in a specific cost range. A heat stroke requiring emergency treatment and two weeks of lost time might produce a $12,000 to $20,000 total incurred cost. In a state with a $15,000 split point, the bulk of that claim lands in primary losses, which receive disproportionate weight in the experience modification formula. In a state with a $35,000 split point, the same claim falls entirely in primary territory.

The consequence for outdoor employers is significant. A landscaping company or roofing contractor with three or four heat claims in a single experience period could see its experience modification increase by 15 to 25 points, depending on the state split point and the employer's expected losses. Because primary losses carry more weight than excess losses in the mod formula, a cluster of moderate-severity heat claims has a larger impact on the modification than a single large claim of equivalent total cost.

This experience rating dynamic creates a potential feedback loop. As heat-claim frequency rises due to climate trends and legislative expansion, outdoor employers accumulate more primary-layer losses. Their experience modifications increase, pushing premiums higher. Marginal employers, those already near the threshold of voluntary market insurability, may be pushed into the residual market. The residual market absorbs the risk, but at administered rates that may not reflect the climate-adjusted loss cost the voluntary market is now pricing for. Nevada's recent 21.6% loss cost increase demonstrated how structural pricing gaps in one dimension can cascade through the experience rating system.

OSHA Enforcement Data as an Emerging Actuarial Source

Between April 2022 and December 2024, OSHA conducted approximately 7,000 heat-related inspections under its National Emphasis Program, issued 60 citations, distributed almost 1,400 hazard alerts, and removed roughly 1,400 employees from hazardous heat conditions. This represents a 35-fold increase from the pre-NEP average of 200 heat-related inspections per year between 2015 and 2020. In April 2026, OSHA revised and extended the NEP for five more years, updating the target industry list to 55 high-risk sectors based on 2022 through 2025 enforcement data.

For pricing actuaries, this enforcement data is a nascent but potentially valuable supplement to traditional loss cost data. OSHA inspection findings, citation frequency by industry sector (mapped to NCCI classification codes via NAICS crosswalks), and penalty data provide a leading indicator of heat-exposure risk that precedes the claims data by one to three years. An industry sector with rising OSHA heat citations is accumulating compliance risk that will eventually manifest as compensable claims, particularly as state presumption laws tighten the link between employer noncompliance and claim compensability.

Why This Matters

Workers compensation remains profitable at the system level. The 2026 State of the Line report confirmed a calendar year combined ratio of 91% and a reserve redundancy of $14 billion. But the accident year combined ratio stands at 102%, lost-time claim frequency decline has slowed to just 2%, and both medical and indemnity severity are running at 4%. The margin between profitability and deficit depends on which emerging cost pressures the pricing actuary anticipates versus which ones arrive as surprises in loss development.

Heat exposure is an emerging cost pressure with a clear trajectory. The legislative direction is toward more explicit standards, lower compensability thresholds, and broader employer obligations. The climate direction is toward more extreme heat days, higher peak temperatures, and longer heat seasons. The actuarial data infrastructure is not yet calibrated to capture these trends: heat losses are diluted in aggregate class experience, historical frequency trends understate the prospective hazard, and the interaction between presumption laws and experience rating amplifies the pricing impact in ways the current methodology does not model.

The four-step framework outlined here, isolating heat-attributable losses, building climate-adjusted frequency trends, applying presumption-law frequency factors, and adjusting severity for outdoor return-to-work constraints, provides a structured approach to constructing an explicit heat-exposure loading. It is necessarily approximate given current data limitations. But an approximate, documented loading is a more defensible actuarial position than the implicit assumption embedded in the current system: that heat losses are adequately priced within aggregate class experience when the data, the legislation, and the climate all indicate otherwise.