Algorithms and Artificial Intelligence in Primary Care for conscious drug use: a systematic review

Primary care is a growing medical field willing to become a more integrated and technological asset. Although big changes already happened and more investments have been made, a limited amount of literature describes processes and technologies there applied or to be used. This paper aims to evaluate the efficacy and usability of different types of algorithms in primary care to improve drug safety by speeding up processes and achieving greater accuracy opening the path to providing better healthcare overall. The PICO model was adopted, three electronic databases (PubMed, Cochrane, Web of Science) were searched using appropriate keywords. Selected studies were assessed for quality and risk of bias using the National Institutes of Health Quality Assessment of Controlled Intervention Studies. Data were analysed using descriptive statistic, comparison of drug usage between algorithms or artificial intelligence application and usual care was performed using a χ2 test(α = 0.05). Out of 2207,19 studies were included,37% of them regarding error prevention,21% drug interactions,21% drug monitoring,16% drug prescription,5% drug administration. Results showed an easier and safer medication use in 74% of studies, a loss of safeness and accuracy in 16%; 10% of total studies did not come up with a valid esteem of results either for inadequate availability of the AI machine or because of the heterogeneity of the results in different settings. The evaluation of errors prevention, 40% of total studies, showed the most statistically significant results with 88% of positive outcomes from AI application. The results support that this technological approach to drugs management could contribute to the safety of treatment and to increase patients' and general practitioners' satisfaction. The application of AI or algorithms is significantly associated with a reduction of drug use errors (p < 0.05). Future studies should work toward establishing a gold standard to measure AI performances.