Developing signal triage algorithm for Thai national adverse drug reaction database

Abstract

Since there was an increasing number of ADR reports submitted from healthcare professionals to the Health Product Vigilance Center under Thai FDA, the Thai Signal Detection Program was developed to filter the potential signals. Due to a large number of disproportionate reportings (SDRs), any measures cannot be managed in time. The objective of the study is to develop a signal triage algorithm that can prioritize SDRs in order to assign in-depth assessment, further investigation or regulatory action. Multicriteria decision analysis (MCDA) was chosen to apply to the triage algorithm. The proposed triage algorithm was tested by comparing the result of triaging SDRs with those triaging by SDAWG and by individual experts. Two main steps were carried out: selection of key attributes and assignment of relative importance weight. Six selected key attributed were %serious cases, the fatal outcomes, new drugs, positive re-challenge, changes in reporting and multiple sources of reports. Then the relative importance weights were assigned by the experts to calculate the importance score of SDRs. The high score means priority for further investigation. Comparing the result of signal triaging by the signal triage algorithm with triaging by individual experts revealed the agreement of 69% whereas comparing with signal triaging by the collective judgment from SDAWG revealed the agreement of 32% since there are other factors influencing experts’ decisions such as comorbidity and multiple medication. For example, serum lipid reducing drugs were often prescribed to comorbid patients. When they experienced ADRs, more weight can be given to some specific concomitant drugs. Drug or ADRs in current interest and unfamiliar ADRs should be additionally taken in the criteria. The signal triage algorithm can enhance the efficiency of the triage method by experts because it is systematic, transparent, timely, repeatable and also developed on the scientific basis.