Development and exploratory analysis of software to detect look-alike, sound-alike medicine names

Background: ‘Look-alike, sound-alike’ (LASA) medicines may be confused by prescribers, pharmacists, nurses and patients, with serious consequences for patient safety. The current research aimed to develop and trial software to proactively identify LASA medicines by computing medicine name similarity scores. Methods: Literature review identified open-source software from the United States Food and Drug Administration for screening of proposed medicine names. We adapted and refined this software to compute similarity scores (0.0000–1.0000) for all possible pairs of medicines registered in Australia. Two-fold exploratory analysis compared: • Computed similarity scores vs manually-calculated similarity scores that had used a different algorithm and underpinned development of Australia's 2011 Tall Man Lettering List (‘the 2011 List’) • Computed risk category vs expert-consensus risk category that underpinned the 2011 List. Results: Screening of the Australian medicines register identified 7,750 medicine pairs with at least moderate (arbitrarily ≥0.6600) name similarity, including many oncology, immunomodulating and neuromuscular-blocking medicines. Computed similarity scores and resulting risk categories demonstrated a modest correlation with the manually-calculated similarity scores (r = 0.324, p < 0.002, 95 % CI 0.119–0.529). However, agreement between the resulting risk categories was not significant (Cohen's kappa = −0.162, standard error = 0.063). Conclusions: The software (LASA v2) has potential to identify pairs of confusable medicines. It is recommended to supplement incident reports in risk-management programs, and to facilitate pre-screening of medicine names prior to brand/trade name approval and inclusion of medicines in formularies