OBJECTIVE: To improve the scoring system of a newly developed ADR algorithm to measure the probability of ADR causality. METHOD: Several ADR cases obtained from Pharmacoviligance Unit at Ministry of Health with known causality probability values were used as reference points for the development of the scoring system. Based on a review of ADR reports with definite causality assessment, several rules were developed to define all possible combinations of criteria for ‘Definite' ADR cases and some combinations for ‘Probable' ADR cases. These parameters were used to determine the new scoring system with the help of genetic algorithm. Testing of the new scoring system was performed on 37 ‘Definite' ADR cases. In addition, sensitivity and specificity analysis were performed to allow a comparison of performance between our algorithm and the algorithm used by ADRAC. RESULTS: When this new scoring system developed by using genetic algorithm was applied to the 37 ‘Definite' ADR reports, 83.8% were identified as ‘Definite' compared to 21.6% by guidelines from ADRAC. Our new algorithm gave a sensitivity of 83.8% and specificity of 71.0% (versus 21.6% and 98.4% respectively for ADRAC). Hence, our algorithm had more cases being classified correctly. CONCLUSIONS: The refining of the scoring system to reflect a quantitative scale has helped make this algorithm more sensitive and increased its useful index, especially when used by clinicians, regulatory agencies or drug companies to generate ADR alert signals. Using a quantitative method of assessing causality also mean that rare ADRs and new ADRs can be detected since a quantitative score can give more precisely the degree of ADR causality.