Human errors in an analytical chemistry laboratory - Implementation of ordinal analysis of variation for risk assessment

• New approach to assessment of risks caused by human errors is described. • Recently developed two-way ordinal analysis of variation ORDANOVA is implemented. • New multinomial score of risk reduction by a laboratory quality system is proposed. • Examples related to pH measurements, GC–MS and ICP-MS analyses are discussed. • Expert judgements on an ordinal scale related to different error scenarios are treated. Decision-making risks caused by human errors in performing a chemical analysis are assessed using laboratory expert judgments (responses) on a specified ordinal scale. In the present paper, a new approach to assessment of risk is described based on implementation of the recently developed two-way ordinal analysis of variation – ORDANOVA. This approach calculates the number of expert responses related to the same category for each ordinal characteristic and then analyzes their relative frequencies as fractions of the total number of responses (of all categories) obtained for this characteristic. It does not violate the properties of ordinal data and allows for the correct interpretation of expert responses. Previously published expert responses on the risks in pH measurements of groundwater, in gas chromatography–mass spectrometry multi-residue pesticide analysis of fruits and vegetables, and in inductively coupled plasma–mass spectrometry analysis of geological samples, are analysed as examples. The datasets prepared for ORDANOVA calculations with the freely available software tool are provided in supplementary materials to the paper. The reduction of risk by different components of the laboratory quality system (QS) are estimated under several error scenarios. New multinomial scores characterizing risk reduction by the laboratory QS as a whole are proposed.