ERMI and HERTSMI-2, explained

The Environmental Relative Moldiness Index was developed by EPA researchers as a way to compare the mold burden of groups of homes in research studies. It works from a dust sample analysed by mold-specific quantitative PCR, which counts the DNA of named species rather than growing anything on a plate.

The index measures 36 species. Twenty-six are Group 1, associated with water-damaged buildings; ten are Group 2, common background organisms found in most homes regardless of condition. The score is calculated as the difference between the two groups on a logarithmic basis, which is why it can be negative, and it was normed against a national sample of roughly 1,096 homes drawn to be representative of US housing.

That normative sample is the thing that makes it an index rather than a measurement. An ERMI value is a position relative to that housing stock, not a quantity of mold. It is a percentile in disguise.

HERTSMI-2 — Health Effects Roster of Type-Specific Formers of Mycotoxins and Inflammagens, second version — was developed outside EPA by Ritchie Shoemaker with Mycometrics. It takes five of ERMI's 36 species and scores them on a simple points system rather than as a relative index.

The five are Aspergillus penicillioides, Aspergillus versicolor, Chaetomium globosum, Stachybotrys chartarum and Wallemia sebi. Each contributes a fixed number of points according to its concentration in spore equivalents per milligram of dust, and the points are summed.

Because it uses a subset of the same species, a HERTSMI-2 score can be calculated from an existing ERMI report without new sampling. That convenience is a large part of why the two circulate together.

EPA has said that ERMI has been peer reviewed for research purposes but has not been validated for non-research purposes, and that for this reason the agency does not recommend routine public use of it in homes, schools or other buildings. EPA also does not recommend that homes be routinely tested or sampled for mold at all.

The agency's own Office of Inspector General examined the question and raised the concern directly: that members of the public may make inappropriate decisions or take unnecessary actions about indoor mold in the belief that the tests they bought rest on tools fully validated and endorsed by EPA.

This is worth separating from a broader claim. EPA is not saying the mycology is wrong, that MSQPCR does not work, or that the 36 species are badly chosen. It is saying a research instrument built to compare populations of homes has been repurposed to judge individual ones, and that the validation supporting the first use does not transfer to the second.

It can tell you which of the named species are present in the dust you sampled, and in what quantity. That is genuine information, and the composition is informative in the way the species pages describe: the five span the water activity range, so their relative levels sketch what kind of moisture history a building has had.

It cannot tell you where the growth is, whether it is still active, where the water came from, or whether anyone has been affected. It is also sensitive to things that have nothing to do with the building's condition — how long dust has accumulated, whether the house was recently cleaned, which room was sampled, and what came in on shoes and pets.

So the defensible use of a panel is as one input into a picture that also includes the building itself. The indefensible use is as a single number that decides whether a family should leave a house. The distance between those two is where most of the harm in this market happens, in both directions: people frightened out of sound homes, and people reassured about homes with a real problem the sample happened to miss.

A modelled estimate and a dust panel answer different questions, and neither substitutes for the other. A panel samples what is in one building's dust today. A model estimates, from records, whether an address has the conditions that make growth likely — climate load, terrain, construction era, water event history.

The model's honest limitation is the mirror image of the panel's: it cannot find mold, because it never enters the building. Its advantage is that it describes the conditions, and conditions are both what caused any growth and what has to change for the problem to stop recurring.

The reasonable sequence for most people is the cheap, non-invasive question first and the sample second, directed at what the first step raised. Sampling without a hypothesis is how a homeowner ends up with a number and no decision.