Independent pharmacist prescriber in critical care: One step closer with artificial intelligence

Abstract

ObjectivePrescribers usually authorize 80% of clinical pharmacy interventions (CPI). Current study aimed to determine if full authorization of CPI (i.e. 100% acceptance) would rather reduce the length of intensive care stay and readmissions. Data source/study settingThis is usual care CPI data obtained in a 100-bed, tertiary, acute care private hospital with 12-bed intensive care unit. Study designObservational artificial neural network model of CPI data. Data collection/extraction methodsClinical pharmacists documented CPI data using a special Excel sheet on a daily basis. Data analyst consolidated CPI to one accepted (124 of 167) and/or one rejected (43 of 167) per patient per admission to intensive care unit (ICU). Then analyst compared the two groups at two alpha levels (0.10 and 0.05) of significance. Artificial neural networks (ANN) were trained and validated for length of intensive care stay after the index CPI (LOSICUA, primary outcome) as well as for the secondary outcomes of mortality, readmissions and costs. Finally, we used the best ANN to compare the outcomes with 80% vs. 100% approval of CPI. Principal findingsModels had significant reduction in LOSICUA with 100% versus 80% authorization. At alpha of 0.05 (8 variables model), we observed ∼0.8 days reduction in LOSICUA (1.4 days (0.3–22.9) to 0.6 days (0.0–39.0), respectively, P-value = 0.001). Readmissions were less likely with 100% (16 of 167) versus 80% (22 of 167) approval (P-value = 0.041). Costs saved with 100% approval in this model would be almost 100 US dollars and would help hire 1.6 full time equivalents of competent doctor of pharmacy. ConclusionsMaximizing acceptance of CPI in this model reduced LOSICUA, readmissions, and costs. Pharmacists qualify for independent prescribing with further privileging and definitions of scopes of practice.