RESUMO
Most Legionella culture tests are performed on building water samples that have been shipped to analytical laboratories for analysis. Significant (≥ 1 log10 unit) changes in results were observed in 52% of held samples (6 h or longer, ambient temperature) drawn from building water systems in a 42-sample initial survey. It was not practical to use the spread plate protocol for on-site "t = 0" cultures in a larger, more diverse survey of thousands of building water systems. Two thousand four hundred twenty-one (2421) building water samples were split for on-site analysis using a field culture protocol and then also cultured after overnight shipment to the lab for analysis with the standardized spread plate method. Legionella test results from building water system samples are usually interpreted as ≥ a numerical detection or action limit. Therefore, binary statistical analyses were calculated by setting t = 0 culture results to "true". Overall in this survey, 10.4% of water samples sent to the laboratory for analysis returned either false-positive or false-negative results. The overall positive predictive value of results was poor (36%). Most (83%) false-positive results were returned from utility water systems. Most (74%) false-negative results were returned from potable water systems. These inaccuracies have serious implications in regard to interpretation and use of Legionella test results. The overall negative predictive value of results was excellent (99%) and also it was good (92%) for results from a polymerase chain reaction (PCR) assay that can be therefore used as a negative screening method.