Integration of ISO15189 and external quality assurance data to assist the detection of poor laboratory performance in New South Wales

A systematic survey of the peer-reviewed literature was conducted to identify the key international themes that govern laboratory quality management. Informed by the survey results, predictive models utilising assessment data against the ISO 15189 standard, and external quality assurance programme (EQA) data, were assessed. Via PubMed, a systematic survey (SS) of the international pathology quality literature identified more than 100 articles, which were subjected to text-mining and meta-analyses via R statistical programming. Word patterns were examined for indicators of current best practice in quality assurance. Random Forest and ANCOVA models were subsequently developed with data obtained from twenty-one anonymous pathology laboratories in NSW. The SS and text-mining did not show a consistent international consensus for laboratory quality; however, approximately 15% of articles suggested root cause analysis as a means to investigate quality problems. Using the Random Forest algorithm, an integrated ISO 15189 EQA model was developed, with results further supported by ANCOVA. The combined Random Forest ANCOVA method succeeded in identifying EQA markers (e.g., serum potassium) that correlated with ISO 15189 audit results, providing a robust predictive model of laboratory quality monitoring superior to that proposed for root cause analyses.