Measurement Accuracy and Precision Quiz

Bias

• Bias is the difference between the measured value and the true value of a concentration.
• Bias can be positive (measured result is too high) or negative (measured result is too low).

Accuracy

• Accuracy is how close a measurement is to the true value.
• It is estimated from the recovery of known standards.

Recovery of Standards

• To estimate accuracy, a standard solution with a known concentration of analyte (expected value) is measured.
• The ratio of measured concentration to expected value is then calculated and expressed as a percentage, known as % recovery.
• By regularly measuring known standards and recording % recovery, an analyst understands the general accuracy of a method.

Precision

• Precision is the degree of agreement among replicate measurements, reflecting the scatter of results.
• For multiple measurements (greater than 3), precision is often expressed as the standard deviation (SD).
• Precision can also be expressed as the percent relative standard deviation (%RSD), calculated as SD x 100% / mean.
• More precise results have a lower standard deviation.

Detection Limits

• Every analytical method has a detection limit, the smallest concentration that can be detected.
• Detection limit is a yes/no decision: is the analyte present or not?
• The quantitation limit is the smallest concentration that can be reliably measured and reported as a meaningful number.
• Quantitation Limits are always higher than Detection Limits.

Method Detection Limit (MDL)

• MDL is the minimum concentration that can be measured with 99% confidence that the analyte concentration is greater than zero.
• Also known as Limit of Detection (LOD).
• It's the smallest concentration where you can confirm the presence of something, but not necessarily quantify it.

Estimated Quantitation Limit (EQL)

• The lowest concentration that can be reliably measured within specified precision and accuracy limits during routine laboratory operations.
• Also known as Limit of Quantitation (LOQ).
• The smallest concentration at which a number can be reported, usually 5 to 10 times the MDL.

Reporting Limit (RL)

• The lowest concentration that a laboratory reports for an analyte.
• Concentrations below the RL are considered "not detected".
• Laboratories typically use their EQL as the reporting limit.
• RL must always be above the MDL.

Why MDLs and RLs Matter

• Environmental regulations often set limits on the maximum allowable concentrations of chemicals in samples (e.g., drinking water).
• Laboratories must be able to reliably measure concentrations below these regulatory limits.

Method Detection Limit (MDL) Determination

• MDL depends on the entire analytical method, not just the measurement step.
• It's determined from multiple replicate analyses of a sample matrix, calculated from the precision of the measurements.
• MDL is not related to the accuracy of the method.

Steps for MDL Determination

• Estimate the MDL.
• Prepare a solution with a concentration 5 to 10 times the estimated MDL.
• Analyze 8 portions (aliquots) of the solution using the entire analytical method.
• Calculate the standard deviation of the 8 results.
• Use Student's t distribution to calculate the MDL: MDL = t x SD (for n = 8, t = 2.998; for n = 7, t = 3.143).

Example of MDL Determination for Total Copper

• Analyst estimates the MDL to be 0.0007 mg/L.
• Analyst prepares a solution with 0.005 mg/L copper.
• 8 aliquots of the solution are analyzed, resulting in a mean of 0.0051 mg/L and SD of 0.000196 mg/L.
• MDL is calculated as 2.998 x 0.000196 = 0.00059 mg/L or 0.59 µg/L.

Spiked Blank or QC Standard

• A spiked blank is a sample prepared by adding a known concentration of analyte to a blank matrix.
• It is crucial that the spiked blank be prepared independently from calibration standards, using a different bottle of reagent (second source).
• Spiked blanks are used to monitor the accuracy of measurements and check for matrix effects.

What is an Acceptable Spiked Blank?

• Acceptance limits for spiked blanks are based on historical data (at least 25 data points collected over time).
• A result within 3 standard deviations of the historical mean is considered acceptable.

Matrix Effects

• Real samples are often complex mixtures, and components in the sample matrix can sometimes interfere with the measurement of the target analyte.
• These interferences, which can suppress or enhance the measurement, are called matrix effects.

Matrix Spike or Spiked Sample

• To assess matrix effects, a matrix spike is prepared by adding a known concentration of analyte to a sample instead of pure water.
• It is used to determine the effect of the sample matrix on the method's recovery efficiency.

Laboratory Certification/Accreditation

• Accreditation is formal recognition by an outside agency of a laboratory's competence.
• Accreditation agencies include CALA (Canada), SCC (Canada), and NELAP (US).

Laboratory Accreditation - ISO Standard 17025

• ISO Standard 17025 specifies "General Requirements for the Competence of Testing and Calibration Laboratories".
• This international standard is obtainable from the International Organization for Standardization (ISO) or the Standards Council of Canada (SCC).

Laboratory Accreditation - Requirements

• Accreditation is granted on a parameter and matrix basis.
• Laboratories are required to participate in performance evaluations (PE) twice a year and undergo site assessments every two years.
• Fees are associated with accreditation.

Standard Operating Procedures (SOPs)

• SOPs are written procedures detailing how analyses must be conducted (analytical methods).
• SOPs provide clear, unambiguous instructions for analysts to follow.
• They include validation data (MDLs, accuracy, precision, linearity) and are controlled documents.

Control Charts

• Control charts are graphs that show how measured results change over time (days, weeks, months).
• They typically monitor measurements on known standard solutions or spiked blanks.
• Control charts are used to monitor the "in-control" status of a process.

What is a Control Chart?

• Control charts graphically display a quality characteristic measured or computed from a sample over time.
• They include a center line representing the average value during the in-control state, an upper control limit (UCL), and a lower control limit (LCL).