SETTING THE REPORTING LEVEL

The USEPA MDL procedure does not address the issue of setting reporting levels. Both the USEPA MDL and the LT–MDL focus exclusively on minimizing the risk of reporting a false positive. At the MDL concentration, however, the risk of a false negative is not adequately limited. A sample with a true concentration equal to the USEPA MDL or LT–MDL has a 50-percent chance of not being detected (Keith, 1992). This is shown in figure 8, where the frequency distribution is centered on the calculated MDL. Assuming that the MDL concentration does, indeed, represent a detection “limit” (that is, the analyte cannot be detected reliably at less than this concentration), then up to 50 percent of the measurements made of a sample having a true concentration equal to the MDL would be less than the MDL (shaded region in fig. 8) and, thus, would result in a false negative. The NWQL views a 50-percent probability of a false negative as unacceptably high for use of the MDL as a reliable reporting level.

Figure 8. 50-percent chance of false negative.

image size 14KB-jpg

Figure 8. False negative probability when a sample contains the analyte at the method detection limit (MDL) concentration.
Recognizing the inadequacies of the MDL as a reporting level, laboratories often set quantification limits (operationally minimum reporting levels) at concentrations greater than the determined MDL’s and in a region that supports quantitative determination. For example, the reporting levels might be set at practical quantitation limits (PQL’s) that are 5 or 10 times the MDL (U.S. Environmental Protection Agency, 1985), or at the limit of quantitation (LOQ), which is a concentration 10 standard deviation units above the average blank response (Keith, 1992). More recently, the USEPA has suggested the use of a minimum level (ML), which is 3.18 times the MDL (for n = 7) (U.S. Environmental Protection Agency, 1993). Gibbons and others (1997a, b) recommend use of an alternative minimum level (AML) that is derived from a multiple-concentration calibration procedure that eliminates or minimizes many of the assumptions and limitations of the USEPA MDL and ML procedures.

In establishing the reporting level, the NWQL has set the acceptable rate of false negatives at no more than 1 percent. This requires the use of a different value from the LT–MDL as the reporting level. The laboratory reporting level (LRL) has been devised to meet this requirement and is comparable to the reliable detection level of Keith (1992) when the false positive and false negative rates are set at <1 percent. The LRL is calculated from the LT–MDL, as follows:

LRL = 2 x LT–MDL.

As shown in figure 9, multiple measurements of a sample having a true concentration at the LRL should result in the concentration being detected and reported 99 percent of the time. Note that the reported concentration can be lower or higher than the true concentration at LRL. One percent of the measurements of this sample would result in a non-detection (shaded region in figure 9), because the measurements fall below the LT–MDL (again assuming that the LT–MDL represents a true “detection limit”).

Figure 9. The risk of a false negative (not detecting an analyte when it is present).

image size 18KB-jpg

Figure 9. The risk of a false negative (not detecting an analyte when it is present) at the laboratory reporting level (LRL) is no more than 1 percent. Note: The reported concentration might be less than or greater than the true concentration at the LRL. [LT–MDL, long-term method detection level]
For non-detections, reporting <LRL is more reliable than if a lower concentration is used such as <LT–MDL, because of the increasing risk of a false negative report at decreasing concentration. For any given sample in which the analyte is not detected, there is no indication as to what concentration the analyte might be “less than” for reporting purposes. Therefore, <LRL is reported for non-detections because the risk of a false negative at LRL is no greater than 1 percent. The LRL was chosen as the default “less than” reporting level to protect against false negatives while simultaneously being set as low as practical to provide low-concentration environmental data that are needed for most studies conducted by the USGS.

An additional feature of the new reporting conventions is that, unlike the MRL convention that was used historically by the NWQL, positive detections below the LRL are not censored. Detected analytes with concentrations between the LT–MDL and the LRL are reported as estimated. This is because a detection in this region should have a <1-percent probability of being a false positive. Even less censoring occurs in information-rich methods where estimated concentrations are provided for analytes detected below the LT–MDL.

next arrow New Reporting Conventions
back arrow The U.S. Geological Survey's New Reporting Procedure Based On Long-Term Method Detection Levels
up arrow U.S. Geological Survey Open-File Report 99-193