Margaret Horton, Paige’s Vice President of Clinical Partnerships and Evidence Generation, oversees the global external study program here at Paige, collaborating with diverse institutions and leaders in pathology to critically study the impact of AI-based digital pathology solutions on pathological diagnosis. One such leader is Monica Dolton, Research Project Manager in the Verrill Pathology Group, Nuffield Department of Surgical Sciences, University of Oxford. Together, Monica and Margaret have been working on the ARTICULATE PRO study, and at this year’s Pathology Visions, they joined forces to present an enlightening presentation on the study and how it has made patients a top priority, with the aim of impacting the patient experience both within the study and beyond.
ARTICULATE PRO is a multi-center prospective study examining the impact of introducing artificial intelligence (AI) software into the prostate cancer diagnostic pathway. Specifically, throughout the course of the study, 3 NHS Trusts across England – Oxford University Hospitals NHS Foundation Trust, North Bristol NHS Trust, and Coventry and Warwickshire Partnership NHS Trust – will implement Paige Prostate to assist pathologists in the detection of cancer and the grading and quantification of tumors. The study team is comprised of representatives from across multiple disciplines to ensure that one common thread is consistent throughout: prioritization of the patient.
Thus, the study includes 3 cohorts of patient and public stakeholders: patient representatives who work on the study team, patients in the clinical pathways being assessed in the study, and patients and public who are not directly involved in the study, but who contribute through focus groups and surveys. Contribution from all 3 of these groups will shed light on how patients feel about digital pathology and AI, how their trust in AI could be improved, and how effective AI governance can be established that protects patient interests.
One initial focus group of 14 men who have lived prostate cancer experience revealed 100% support for the use of AI in routine clinical practice, provided it is proven to improve the accuracy of pathologists. Similarly, the respondents indicated that they would feel more confident in their diagnosis knowing that a software system was used as an additional quality measure. The primary outcomes they expressed interest in seeing from a study of AI, such as ARTICULATE PRO, were whether it could enhance pathologist confidence and reduce the chance for missed cancers.2
Additionally, in a survey of 1,276 Prostate Cancer UK supporters, all of whom had a prostate biopsy at some time, many respondents expressed that they had a very limited or complete lack of understanding of histopathology. At the same time, however, when asked how they feel about digital pathology tools completely replacing microscopes, 87% gave a very positive or positive response, with only .3% indicating a negative response. The idea of using and studying AI in pathology was met with a similar response, with 83% of respondents agreeing that this is a good idea.3
Overall, these initial patient surveys and focus groups demonstrated positive support from the patient perspective for bringing new technologies into routine use. Yet they also shed light on a few key concerns and actions that must be taken to better educate and inform the public, improve their trust in AI, and establish practices that could improve their experience with AI’s use.
First, the responses prompted ARTICULATE PRO to establish a set of guiding principles, within which some key public concerns were addressed. For example, they established how patient data was allowed to be used and shared and confirmed that AI would never be replacing a pathologist, but instead only used with humans in the loop. Further, the team established a set of Quality Assurance requirements in line with the input of patients and national government requirements. One of their most important actions was to co-develop report-level language that will be added to patient biopsy reports, explaining that AI may have been used in their biopsies. Finally, ARTICULATE PRO endeavored to create patient facing collateral that could provide appropriately written information that directly addresses public concerns and engaged patient advocates in sharing their lived experience to offer insights and support to the community, from those who have undergone a similar journey. All of this open and clear communication about AI will help to educate on pathology and how AI may aid in its processes, improve public trust, and create a smoother interaction between patient and healthcare provider.
Importantly, the study homed in on a few additional factors that could impact patients, such as heath inequality and access to healthcare, due to geographical location for instance. To strengthen their understanding in these areas, ARTICULATE PRO are engaging with stakeholders including the NIHR Oxford Biomedical Research Centre Diversity in Research Group, whose members are drawn from minority ethnic communities, the transgender community, and from those with learning disabilities. They are also working with 2 Academic Health Sciences Networks in Oxford and Bristol, with plans to access data such as patient’s ethnicity on the study, to better understand and address related issues. Finally, as the study continues on, they plan to collaborate with other seldom-heard groups, all of which should help ensure that every patient from every background is considered and cared for in this study and beyond.
All of this interesting research and work, though it is still ongoing, has revealed 5 essential takeaways, which Monica and Margaret broke down at the end of the session:
- Involving the lived experience of patients in all aspects of the study shifts the focus from a technology-forward approach to human-centered approach, which is key to getting human-centered outcomes.
- Having a Multidisciplinary team is the only way to make the study as robust as possible. This includes patients, urologists, cancer specialist nurses, oncologists, computer scientists, statisticians, and the like.
- Patients are engaged, interested, and highly supportive of bringing AI into pathology in responsible way.
- Providing clear, transparent information to build public trust in AI in healthcare for patients and public who do not feel comfortable with its use is essential.
- As AI in healthcare use expands, and it is doing so quickly, there will be new opportunities for engagement with organizations and new standards for the benefit of patients.
To conclude, Monica and Margaret reiterated their hopes for the future of the study. With all of the work they have done to date, they were able to establish a strong understanding of patient and public views on the use of AI, and they will continue to use this to establish practices and set precedents that serve and protect patient needs, within the study framework and, hopefully, in ways that stretch beyond it.
1UK Standards for Public Involvement. NIHR. Accessed October 20, 2022. https://sites.google.com/nihr.ac.uk/pi-standards/home
2November 2020 Focus Group
3Rakovic K, Colling R, Browning L, Dolton M, Horton MR, Protheroe A, Lamb AD, Bryant RJ, Scheffer R, Crofts J, Stanislaus E, Verrill C. The Use of Digital Pathology and Artificial Intelligence in Histopathological Diagnostic Assessment of Prostate Cancer: A Survey of Prostate Cancer UK Supporters. Diagnostics. 2022; 12(5):1225. https://doi.org/10.3390/diagnostics12051225