Artificial intelligence and the NHS: a qualitative exploration of the factors influencing adoption

Article Figures & Data

Table 1.

Further information on the diffusion of innovations framework²²

Attribute	Description	Impact on rate of adoption
Relative advantage	Degree to which an innovation is perceived as being better than the idea it supersedes	Increases rate
Compatibility	Degree to which an innovation is perceived as consistent with the existing values, past experiences and needs of potential adopters	Increases rate
Complexity	Degree to which an innovation is perceived as relatively difficult to understand and use	Decreases rate
Trialability	Degree to which an innovation may be experimented with on a limited basis	Increases rate
Observability	Degree to which the results of an innovation are visible to others	Increases rate

Table 2.

Key informants sought and rationale

Group of key informants	Rationale
Individuals of influence	Influential power and a breadth of experience in the field
Individuals from royal colleges	Knowledge on specialty-specific issues
Individuals from governmental and regulatory bodies	Knowledge on the regulation and policy aspects of AI
Researchers	In-depth understanding of the subject area

Box 1.

Description of data analysis

Familiarisation with the data was achieved initially through transcription. To be immersed in the data, the researcher listened to the audio-recording of each interview, re-read the transcripts and made notes of first impressions.
Initial codes were generated. The researcher coded all transcripts, and 69 codes were identified.
Themes were searched for. Codes were collated in three prospective themes, with 11 candidate sub-themes.
Themes were reviewed. Several sub-themes merged to produce nine sub-themes within the three main themes.
Themes and sub-themes were defined and named: system, people and technology.
Finally, this paper was produced. Themes were interpreted using the diffusion of innovations theoretical lens.

Table 3.

Themes and sub-themes

Theme	Sub-themes
System	Socio-political context of the NHS in 2020 Regulatory landscape Fit within the puzzle
People	What actually is AI? People powered transformation It's not going to be a robo-doc
Technology	Data driven nature Challenges ahead Evaluation

Table 4.

Key findings mapped to the diffusion of innovations framework

Element of diffusion of innovations theory	Impact on rate of adoption	Findings
Relative advantage	Increases rate	AI offers a relative advantage by improving the working lives of clinicians Risk of bias in AI tools reduces this relative advantage The degree of relative advantage needed for adoption of AI in the NHS has not been agreed: absence of a gold standard
Compatibility	Increases rate	NHS IT infrastructure may not be compatible with AI Regulatory landscape is not compatible with AI Certain specialties are more compatible with AI Transferability of AI tools may be poor: they may only be compatible with a single NHS site
Complexity	Decreases rate	Improved language clarity around AI could reduce its perceived complexity Education about AI could reduce its perceived complexity
Trialability	Increases rate	High up-front costs of AI, combined with the existing financial pressures facing the NHS, limit its trialability
Observability	Increases rate	Black box AI reduces the observability of the decision-making process
Time: adopter categories	n/a	Some healthcare professionals will be more or less resistant to adopting AI: this reflects the five adopter categories
Social system: opinion leaders	Increases rate	Champions could be used as facilitators of AI adoption; these reflect the opinion leaders described by Rogers