Comparison of responses between Chatgpt and clinical pharmacist

The emergence of electronic databases and information technology has reshaped drug information (DI) is accessed and provided. Generative Pre-trained Transformer (GPT)-4, has drawn attention for its potential application in clinical decision-making and medical research. Thus, the study compares responses from GPT-4 to those of clinical pharmacists in handling DI requests. Sixty DI requests based on common diseases in the UAE involving queries of different complexities were developed. Responses from GPT-4 were compared to those clinical pharmacists using cosine similarity method. Independent t-tests were used to compare them for accuracy and relevance. Two-way ANOVA was used to test the interaction between case complexities and language used in study outcomes. The mean similarity between pharmacists and GPT-4 was 0.88 (SD=0.05), with more than 80% cases almost identical. GPT-4 was as accurate as humans (4.23 vs. 4.13; P=.56). However, answers from GPT-4 were more comprehensive and patient-centered (4.25 vs. 3.72; P=.0006 and 4.35 vs. 3.75; P=.0008; respectively). GPT-4 had a significantly longer responses (270.2 vs. 150.3; P