Measuring drug name similarity to prioritise the application of tall-man lettering in a computerised pharmacy dispensing system
Bibliographic record
Abstract
Abstract Introduction Medicine name similarity is a contributory factor to medication errors.1 Published lists exists highlighting medicine pairs that are easily confused; locally a Look-Alike-Sound-Alike (LASA) list has expanded over time with no formal system of triage. A multipronged approach is required to address LASA risks. Tall-Man Lettering (TML) is one intervention that uses uppercase lettering for the dissimilarities in look-alike drug names to alert staff to the risk of error.1 For greatest impact, it should be reserved for pairs with the highest risk for patient safety. Aim This study aimed to use Levenshtein Distance (LD), Bigram (Bi) and Trigram (Tri) methods to prioritise medicine pairs for TML in the Pharmacy dispensing system. Objectives were to: Produce a comprehensive list of medicines pairs; Establish normalised thresholds from LD, Bi and Tri to prioritise medicine pairs for TML; and, measure drug name similarity using validated software2 applying LD, Bi and Tri to medication pairs as a method of triage for orthographic assessment. Methods Approval was obtained by the Trust Pharmacy Research Committee. The need for ethical submission was waived. A LASA list was developed combining medicine name pairs from National Pharmacy Association list3, historical local list and internal incidents where medicine name confusion was cited. Duplicated, branded and non-stock pairs were excluded. A literature search was undertaken to identify published thresholds for accuracy and sensitivity of the methods in the measure of medicine name similarity. LD measures the minimum number of edit operations needed to transform one string into another; Bi and Tri measures the frequency in which two/three similar sequential strings appear within a medicine name respectively. Two assessors independently entered medicine pairs through a validated computer program2 applying LD, Bi and Tri to measure orthographic similarity. Normalised computed similarity scores (between 0-1 where higher values represent increased drug similarity) were collated on Microsoft excel for comparison against thresholds. Results Two-hundred and twelve medicine pairs were identified for review. The literature defined Bi and Tri thresholds at ≥0.3 and ≥0.1 respectively; in absence of this for LD, in-house thresholds were assessed then defined at ≥0.6. LD identified 84 medicine pairs; Bi identified 144; and Tri identified 158; none were uniquely found by LD, four by Bi and 18 by Tri. A final look-alike list with 82 medicine pairs meeting all three thresholds was identified for TML. Discussion/Conclusion Screening using all three methods led to a 61% reduction in medicines pairs allowing prioritisation of TML as an effective intervention based on look-alike pairs with the highest risk of error. This study focused on identifying orthographic similarity in ‘look-alike’ medicine pairs only. With no single intervention available to prevent LASA errors, future work can explore other interventions. In the absence of literature around normalised LD, the definition of an in-house threshold posed to be another limitation and an area where further exploratory work should be considered. As new LASA incidents arise or the Trust catalogue increases, these methods should be applied to triage their look-alike potential, confirming if TML is an appropriate intervention. References 1. World Health Organisation. Look-alike, sound-alike medication names. patient safety solutions. 2007 May, Volume 1: Solution 1. 2. Strcmp2. [Internet] Department of Computer Science, University of Toronto: Aditya Bhargava [Date of publication unknown]. Available from: https://www.cs.toronto.edu/~aditya/strcmp2/, Work based on: Grzegorz Kondrak. N-gram similarity and distance. Proceedings of the Twelfth International Conference on String Processing and Information Retrieval. 2005 Nov; pp. 115—126. 3. The National Pharmacy Association Pharmacy Team. Look-alike sound-alike (LASA) items [Internet]. The National Pharmacy Association; 2021. Available from: https://www.npa.co.uk/information-and-guidance/look-alike-sound-alike-lasa-items/
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.007 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.002 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".