Risk-first vs disease-first: where OELs, DNELs and kick-off values fit in proactive substance screening

Q: What is the risk-first approach to hazardous substances?

Identifying a hazardous substance or type of work that can cause an occupational disease before workers fall ill, and acting on that prediction. RIVM set it out in letter report 2025-0133 (2026) as the alternative to the disease-first approach, where the link to work is recognised only after illness.

Key Summary: The risk-first approach identifies a hazardous substance or task before workers fall ill, in contrast to the traditional disease-first approach where the link to work is recognised only after illness appears. RIVM set out this framework in its 2026 report Methods for early identification of new and emerging risks of chemicals (letter report 2025-0133), commissioned by the Ministry of Social Affairs and Employment under the Lexces programme, against a backdrop of an estimated 3,000 deaths per year in the Netherlands from workplace exposure to hazardous substances. RIVM judged the most promising methods to be high-throughput approaches that use large databases and combine hazard and exposure data. That is precisely the role curated occupational exposure data, including OELs, DNELs and kick-off values, plays as the data layer a risk-first screen reasons over.

In April 2026 RIVM published Methods for early identification of new and emerging risks of chemicals (risk-first-approach) (letter report 2025-0133, N. Palmen and J.A.B. Kettelarij), commissioned by the Ministry of Social Affairs and Employment as part of the Lexces research programme. It reframes a long-standing problem in Dutch occupational health: despite extensive regulation, an estimated 3,000 people die each year in the Netherlands from exposure to hazardous substances at work. This article explains the risk-first framework, the two methods RIVM is taking forward, and where occupational exposure benchmarks fit as the hazard and exposure data layer underneath any such method.

What the risk-first approach means

In the disease-first approach, the link between a substance or a type of work and a health effect is typically recognised only after someone has become ill. By the time the signal is visible, harm has already occurred. The report’s framing is blunt: there is often simply too little information about the substance or the working conditions to intervene in time.

The risk-first approach inverts this: predict whether a hazardous substance or a type of work can cause an occupational disease, and act on that prediction before exposure causes harm. RIVM compiled an overview of methods that can serve this purpose and noted that each has a different purpose, and that the effort and the information required differ per method.

The report’s assessment of what works is direct, and worth quoting: “The most promising would be high-throughput methods using large databases and combining hazard and exposure data. No single method meets all criteria, and all methods require some expert judgment for prioritization and follow-up.”

The two methods RIVM is taking forward

RIVM selected two methods for further development within the Lexces programme:

Lexces in silico prediction tools (the “Lexces AI tool”) — focused on prediction: estimating whether a substance or work situation could lead to an occupational disease from its properties.
The RIVM Endocrine Disruptor toolbox — focused on compiling information from various sources into a usable signal.

Both were chosen because they are high-throughput, draw on large databases, incorporate both hazard and exposure information, and are readily available. The report situates this work alongside European and international efforts including the Partnership for the Assessment of Risks from Chemicals (PARC) and Mistra SafeChem, and notes that elements from other methods may be integrated later to strengthen risk-first screening.

Where OELs, DNELs and kick-off values fit

A prediction is not yet a decision. An in silico model can flag that a substance is likely hazardous, but turning that flag into a defensible workplace action still requires an exposure benchmark to compare measured or modelled exposure against, and curated substance data to resolve identity, classification and regulatory status. That benchmark layer is what occupational exposure limits provide:

Legal OELs anchor compliance for the few hundred substances per country that have them.
REACH DNELs extend a substance-specific benchmark to several thousand more substances where no OEL exists. See How to find and apply a DNEL and DNEL vs OEL.
Kick-off values cover the long tail: a hazard-banding-derived screening benchmark for the very large number of classified substances that have neither an OEL nor a DNEL.

Kick-off values deserve emphasis in this context, because hazard-banding screening is itself a risk-first method in the sense the RIVM report describes: it produces an actionable benchmark from a substance’s hazard classification before an OEL has ever been set, which is exactly “identify the risk before people fall ill” applied to the no-limit majority of the chemical universe. It is peer-reviewed (see the validation of the DOHSBase methodology) and predates the “risk-first” label. RIVM’s 2026 overview does not include database-driven hazard-banding screening among its selected methods; that is a genuine gap in the inventory rather than a judgement against the method, and it is where curated exposure databases are complementary to the prediction-focused tools RIVM is developing.

DOHSBase as the data layer for risk-first screening

DOHSBase is not a prediction model and does not claim to be one. Its role in a risk-first workflow is the data layer: it is a high-throughput, readily available database that combines hazard and exposure information at scale, which is the exact profile RIVM identifies as “most promising” for the data substrate a risk-first method reasons over. Concretely:

15,000+ OELs from 30+ countries, with legal status and source made explicit.
5,300+ worker inhalation DNELs extracted from ECHA registration dossiers, with deep links back to the source record for verification.
100,000+ kick-off values so that a screen does not go blind on the substances that have no formal limit.
Hazard and exposure combined per substance in the DOHSBase Compare ranking, rather than held in separate silos.

In a risk-first programme, prediction tools and a curated benchmark database are complementary, not competing: the model proposes, the benchmark layer makes the proposal actionable and auditable. DOHSBase is built to be that benchmark layer.

Frequently Asked Questions

What is the risk-first approach to hazardous substances? It is identifying a hazardous substance or a type of work that can cause an occupational disease before workers fall ill, and acting on that prediction. RIVM set it out in letter report 2025-0133 (2026) as the alternative to the disease-first approach, where the link to work is recognised only after illness.

How does risk-first differ from disease-first? Disease-first is reactive: the work-illness link is established after harm, often via epidemiology or reported cases. Risk-first is proactive: it predicts whether a substance or task can cause disease and intervenes before exposure causes harm.

Why does the Netherlands need this? An estimated 3,000 people die each year in the Netherlands from workplace exposure to hazardous substances despite existing regulation, largely because there is too little information to intervene in time. Risk-first aims to close that information gap before harm occurs.

Where do occupational exposure limits fit in a risk-first method? A prediction has to be compared against something. OELs, DNELs and kick-off values are the exposure benchmark layer that turns a hazard prediction into a defensible, auditable workplace decision. RIVM itself identifies methods that combine hazard and exposure data over large databases as the most promising.

Do kick-off values count as a risk-first method? Conceptually yes. Hazard-banding-derived kick-off values produce an actionable screening benchmark from a substance’s hazard classification before any OEL exists, which is the risk-first principle applied to the substances that have no formal limit. The approach is peer-reviewed.