How the lab turns your dust into 36 numbers.

The steps at a glance

1. Collecting the dust

In EPA's research, dust was vacuumed from the floors of the rooms people used most. In two of its studies, the protocol on carpet covered 2 square meters and vacuumed for 2 minutes per square meter (Meklin et al. 2004; Vesper et al. 2007a). The national survey behind ERMI vacuumed 2 square meters in a bedroom and 2 in the living room (Vesper et al. 2007b).

Labs that sell ERMI tests use their own kits. Some collect dust with a vacuum attachment and a filter. Others use a dry electrostatic dusting cloth.

The collection method matters more than most people expect. Scherer et al. (2014) loaded six brands of electrostatic cloth with known amounts of spores. Then they washed the cloths the way a lab would and measured the DNA that came off. Every cloth gave back only a few percent, and the brand alone changed the result.

That study used cloths left open on a shelf for ten weeks, not ERMI kits. For ERMI itself, EPA and HUD researchers tested the question directly. They sampled 151 homes both ways: the standard vacuum sample, then an electrostatic cloth wiped over surfaces in the same rooms. A commercial lab ran ERMI on both. For 33 of the 36 molds, the average levels were not significantly different, and the ERMI values went up and down together (Vesper et al. 2021).

The cloth samples picked out 96% of the homes that the vacuum samples put in the top quarter of the ERMI scale. They also flagged some homes the vacuum samples did not. The researchers concluded that a cloth sample is a useful practical alternative when the standard vacuum sample isn't practical (Vesper et al. 2021).

If you test again later, use the same lab and the same kind of kit. That keeps the two results as comparable as they can be.

2. Sieving and weighing

Most of what is in household dust isn't mold. It is hair, fibers, grit and skin flakes. So the dust is sieved, and only the fine dust is kept. EPA's studies used sieves with openings of 300 to 355 micrometers, about a third of a millimeter (Meklin et al. 2004; Vesper et al. 2007a, 2007b). Another study using EPA's tests used a finer 150-micrometer sieve (Nonnenmann et al. 2012).

The lab then weighs out a small portion for each test. Every study we checked used 5 milligrams (Meklin et al. 2004; Vesper et al. 2007a, 2007b; Nonnenmann et al. 2012). That is roughly a quarter of the weight of a grain of rice.

Results are reported per milligram of dust. So this one weight sets the scale for all 36 numbers.

3. A reference mold goes in

Before the DNA comes out, the lab adds a known number of spores of a reference mold, Geotrichum candidum. The studies use the same mold in different amounts: about 2 million spores per sample in Meklin et al. (2004), and 1 million in Nonnenmann et al. (2012).

The reference goes through every later step with the dust. If an extraction loses part of its DNA, the reference loses about the same share. If something in the dust slows the reaction, it slows the reference too.

Each mold's result is read against the reference from the same extract. That cancels most of the differences from one sample to the next. It is a correction, not a guarantee: it works as long as the 36 molds and the reference behave alike.

Scherer et al. (2014) used the same mold as a warning light. A weak reference signal would flag a sample with something in it that blocks the reaction.

4. Getting the DNA out

Spores have tough walls. The lab shakes the dust in a tube with tiny beads, hard enough to crack them open. The DNA is then washed out and cleaned (Meklin et al. 2004; Nonnenmann et al. 2012).

Not all of it comes out. In Meklin et al. (2004), the extraction recovered on average 43% of the mold DNA from dust, compared with clean control samples. That loss is the reason for the reference mold in step 3.

5. One DNA test per mold

The test is quantitative PCR, usually shortened to qPCR. Each of the 36 molds has its own test. It uses short pieces of DNA, called primers, and a probe that match a stretch of that mold's DNA. EPA researchers designed these tests in the late 1990s and early 2000s (for example, Haugland et al. 1999 and 2004).

A few of the tests can't tell two close relatives apart. The report then names both, as in Aspergillus flavus/oryzae.

The reaction copies the matching DNA over and over, in cycles. Each cycle roughly doubles it. In Meklin et al. (2004), the tests multiplied the DNA by 1.81 to 2.00 per cycle.

The probe gives off light as the copies build up. The lab records the cycle at which the light rises clearly above the background. That cycle number is called the Ct (cycle threshold) or Cq (quantification cycle).

Fewer cycles means more DNA to start with. The table shows what a difference in cycles means when the DNA doubles every cycle.

6. From cycles to spore equivalents

A cycle count on its own isn't a quantity. The lab turns it into one with a standard curve.

To build the curve, the lab runs the same test on samples with a known number of spores. It plots the cycle results against those known amounts. A dust sample's result is then read off that line (Meklin et al. 2004; Vesper et al. 2007a).

The result is reported as spore equivalents per milligram of dust. EPA's papers call the same unit cell equivalents.

A spore equivalent is the amount of DNA that one spore of that mold carries. It is not a count of spores. Broken spores, dead spores and bits of mold threads (hyphae) all carry DNA, and all of it adds to the number. Our methodology page explains why that is useful rather than a flaw.

Other researchers convert differently. Scherer et al. (2014) turned the measured DNA weight into cells using the size of each organism's genome. Both approaches estimate the same thing. The numbers from one still can't be swapped for the other.

What “not detected” means

Every test has a floor. Below it, the signal is too weak to count reliably.

Scherer et al. (2014) set their floor at the smallest amount the test found in seven out of seven repeat runs, with a spread of under one cycle between runs. Anything weaker was reported as negative.

So “not detected” means below the floor. It does not prove the mold is absent from the home.

Labs set and report their floors in their own ways. A “not detected” from one lab and from another aren't strictly the same result.

Studies also count a “not detected” differently when they average results. One EPA study set the floor at 1 cell equivalent per milligram and counted a non-detect as half of that (Vesper et al. 2007a). Another counted it as zero (Nonnenmann et al. 2012).

How a “not detected” enters the ERMI score matters too. The methodology page explains the two conventions and how far they move the score.

Checks built into the method

The published methods include several checks.

• Blanks. A tube of sterile water with no dust goes through every extraction and every run. Any signal there means contamination (Scherer et al. 2014).
• The reference mold. It is in every sample, to catch DNA losses and blocked reactions (Meklin et al. 2004; Scherer et al. 2014).
• Cross-checks. Each test is run against the other molds, so one mold isn't counted as another. Scherer et al. (2014) found no cross-reactions among their ten tests.
• Standards. Every test is tied to a standard curve built from known amounts (Meklin et al. 2004).

These checks come from research papers. A commercial lab's exact procedure is its own. If you want to know how yours works, these questions are worth asking:

• How was the dust collected, and with what?
• What is the lowest amount each test reports?
• Are the results per milligram of sieved dust?

What this means when you read your report

• Each number estimates the mold DNA in a few milligrams of your dust, scaled to spores. It is not a count of spores.
• Near the floor, repeat runs of the same sample can differ by up to about one cycle (Scherer et al. 2014). That is about a factor of two. A small gap between two low numbers may be noise.
• “Not detected” means below the test's floor. It does not mean absent.
• If you test again, use the same lab and the same kind of kit. A change of either can move the numbers by itself.
• One result is one sample of dust from one point in time. The methodology page covers how much a single result can move.

Further reading

• Methodology: why dust, why qPCR, and how we read it
• ERMI limitations: what dust testing does and doesn't measure
• The 82-species story: why ERMI tests 36