Barriers to the widespread adoption of diagnostic artificial intelligence for preventing antimicrobial resistance

Currently, antimicrobial resistance (AMR) poses a significant challenge and severely threatens public health, primarily due to the overuse of antimicrobials. In this survey, conducted across eight countries/areas, we assessed public preferences for two hypothetical types of artificial intelligence (AI): one prioritizing individual health (Individual precedence AI: abbr. Individual-AI) and the other considering the global threat of AMR (World precedence AI: abbr. World-AI). Our focus here is on the conflict (social dilemma) between recognizing the importance of AMR and desiring personalized treatment (for more details on the social dilemma, see Ito et al. 2022. Sci Rep. 12: 21084). The first question asked respondents about the desired adoption preference rates of World-AI versus Individual-AI. The second question inquired whether respondents were agree or disagree with the standardization of a single AI. The third question asked which type of AI, World or Individual -AI, should be chosen if standardization to a single AI was necessary. It is important to note that the only ideal response combination to resolve the issue of AMR would be a 100% adoption rate for World-AI (0% for Individual-AI), in favor of standardization, and choosing World-AI as the unified standard. This survey thus explores how the social dilemma inherent in AMR may impede the adoption of diagnostic AI.

Methods
The ‘Survey on Medical Advancement’ was conducted across 8 countries/areas. The survey was conducted in four phases: the first phase was conducted in 2020/Jan/8~10 in Japan; the second phase was conducted in 2020/Jul/1~7 in Japan, the US, and the UK; the third phase was conducted in 2021/May/18~26 in Sweden, Taiwan, and Australia; and the fourth phase was conducted in 2021/Jun/23~30 in Brazil and Russia.

For the two surveys conducted in Japan, Cross Marketing, Inc. (https://www.cross-m.co.jp/en/), an internet survey company, created questionnaire webpages in accordance with our study design. Cross Marketing, Inc., also handled the data collection process. As of April 2020, Cross Marketing Inc. maintained an active panel (survey participants who registered in advance) of 4.79 million individuals, defined as survey participants who had been active within the last year. The questionnaire and response section were hosted on a website, allowing respondents to complete the survey and submit their responses. We obtained 500 submissions for each gender and age group by random sampling from all surveys collected during the survey periods.

The surveys conducted in the other 7 countries/areas (i.e., the US, the UK, Sweden, Taiwan, Australia, Brazil, and Russia) were executed by Cint (https://www.cint.com/). Cint is the world’s largest consumer network for digital survey-based research and is headquartered in Sweden. Cint maintained a survey platform comprising more than 100 million consumer monitors in over 80 countries as of May 2020. For surveys conducted in the US, the UK, Sweden, Taiwan, Australia, Brazil, and Russia, Cint Japan (https://jp.cint.com/), the Japanese distributor of Cint, created translated questionnaire webpages following our study design. Cint Japan also managed the data collection process. We obtained at least 500 (the US, the UK, Sweden, Brazil, Russia) or 250 (Taiwan, Australia) submissions for each gender (male and female) and age group (20s, 30s, 40s, 50s, and 60s) by random sampling from all surveys collected during the survey periods.

Note that both companies employed a sampling process to exclude inconsistent or apathetic respondents. For instance, individuals with discrepancies in their responses (e.g., those whose registered age did not match their reported age at the time of the survey) were excluded before the data reached the authors. Furthermore, respondents with notably short response times (i.e., less than 1 minute) were excluded from the analysis, as their quick responses may have indicated a lack of careful consideration of the survey questions.

This study was approved by the Ethical Committee of the Institute of Tropical Medicine, Nagasaki University (Approval No. 190619217).