A comparative assessment of action plans on antimicrobial resistance from OECD and G20 countries using natural language processing.

ABSTRACT

Following the launch of the Global Action Plan on antimicrobial resistance (AMR-GAP) in 2015, most OECD and G20 countries developed their own national action plans (AMR-NAPs). This is the first paper that deploys natural language processing (NLP) techniques to systematically measure and compare the extent to which AMR-NAPs from 21 OECD and G20 countries align with the AMR-GAP in terms of the strategic objectives and interventions. We quantify the extent of alignment based on two NLP metrics term-frequency (TF) and term-frequency-inverse document frequency (TF-IDF). Quantifying TF allows us to compare the relative prominence of strategic objectives and interventions, whereas quantifying TF-IDF enables us to identify interventions that occur more frequently in each AMR-NAP. Similar to the AMR-GAP, in our sample, terms associated with optimizing antimicrobial use in human and animal health have the highest frequency (TF = 0. 287), whereas terms linked to raising AMR awareness and education have the lowest frequency (TF = 0.066). Substantial cross-country variation exists in the distribution of interventions that are distinctly frequent in each AMR-NAP. We also report new evidence on the selected policy design and monitoring and evaluation features of these documents. Our results suggest a high degree of congruence between the AMR-GAP and AMR-NAPs, with notable diversity in the spate of interventions that OECD and G20 countries discuss in their action plans.