Our policy for manual hemacytometer counts states that the count from the two sides of the hemacytometer must agree within 10%, or the count must be repeated. However, if the count is very low (i.e. a clean CSF), it may be very difficult to get 10% agreement (e.g. side one has 2 WBCs, side two has 1 WBC- this is a 50% difference, but not clinically significant). How do other labs define their agreement criteria on samples with low counts?

I had to check our procedure-we have it written like yours too! However, since counts below 5 per ul are considered normal, it perhaps is not important. In you example, whether you report a 1 or 2 makes little difference.

We do have a thing in our procedure for counting bone marrow diffs where we allow for more variance on very low counts instead of just a straight percentage for all amounts.

Our policy states that the allowable difference is 10% or 2 cells, whichever is greater. My last employer had an allowable difference of 10% or 5 cells, whichever is greater. This should cover those fluids with low cell counts.

Our policy is "two sides must agree within 4 cells if the count is <50 or within 10% if > 50."

At 50 cell 10% is 5 cells, if less than 50 we revert to agreement +/- 4 cells.

Our policy states both sides should agree within 20%

Does anyone have a reference for the agreement percentage? From what I can tell, the 10% is a rule of thumb from the manual WBC, RBC, and Platelet counts. It is sometimes a bit tight when counting white cells on bloody pleural fluids. I would like to have ours set at 20% as well.

You may be able to finding something in a textbook, but I think this is just the kind of thing you set up and follow your procedure (approved by pathologist). I don't think you are going to find any manufacturer or regulator recommendation. Here, the pathologist is ultimately responsible for dictating when something is or is not significant regarding precision. So if you want to write into the procedure "replicates acceptable when within 10% or 5 cells" or something like that, you should be OK I think.

Scott