Can only AI can save us from the burden of administration?

As health systems adopt ever more expensive and complex electronic medical record (EMR) systems – clearly with minimal design input from doctors and with almost no thought to research and with limited focus on improving patient outcomes – we are left burdened with yet more and more menial tasks that are taking us away from face-to-face patient care [1, 2]. How do we re-balance “the system” that is often beyond our control? Perhaps AI may save us. Amongst many applications, AI is perfect for navigating repetitive processes and for being a scribe to assist us with our activities.

Opportunities exist in the AI space to make positive changes. Such technology should follow the “SAFE” mnemonic:

S-afe for patients and staff.

A-ffordable and accurate.

F-used with existing technologies.

E-asy to use.

This will ensure that they are not just expensive “plug-ins” and that there is integration that is smooth and reliable. Predictive tools for patient prognosis can be enhanced alongside simpler tasks of obtaining accurate patient histories and notes [3]. Education of patients about procedures could become more interactive and faster.

Diagnostics will also be aided by AI to not miss and guide interpretation whilst adding consistency in subjective reports [4]. Even more exciting are applications that may aid in the early diagnosis of malignancies, particularly rare ones like penile cancer, taking a de-identified picture on a smartphone, feeding through AI online and triaging back to patients and health workers are upon us [5]. The accuracy and educative power of such initiatives is potentially overwhelming, particularly when we consider the blockchain technology and how it may be utilised in urology and medicine [6].

However, there are potential pitfalls. AI can lead to lazy and indeed sloppy attention to detail where, in the health system, mistakes can be life threatening. “Cut and pasting” from notes to letters and discharge summaries is already becoming a problem, as is remembering doses of drugs, drugs to cease before surgery and common interactions of drugs that should not be discharged to AI without human input [7]. Furthermore, where free thought is abolished, and we become homogenised with few new ideas, we can become complacent. It may be easier for some to be comfortably safe in the knowledge that we have acceptable AI ideas to put forward, but this robs us of the innovation and free thought that we must foster to solve problems and improve patient outcomes.

A further danger is that AI is actually cheap but like everything else in healthcare, those in the space can only see the inflated prices for average systems to help repay investors and bring returns. Governments need to partner with universities to develop our own AI programmes that are cheap, reliable and can be rolled out across Australia and New Zealand, and then other regions. Whilst at it, we could coax the same academics and innovators to rebuild EMR that suit our region, our health workers and, most importantly, our patients' best interests. In the process we all benefit and can cease haemorrhaging billions of dollars offshore to prop up companies at the expense of our own jurisdictions.

But back to the initial question – is it only AI that can save us? Well, actually no, there are plenty of other levers that can be pulled. The first is a recognition by health systems of where EMR priorities should be: constructed with maximal design input from doctors, significant thought to research and with the main focus on improving patient outcomes.

Secondly, the underutilisation of local human resources full of contemporary and useful knowledge should be recognised and reversed. A critical example of failing to use such resources is the lack of debriefing of doctors who work overseas or even interstate. In particular, doctors who go on fellowships overseas are privy to other health systems, what works well and what does not. Health services seem incapable of recognising this, let alone harnessing such knowledge to improve their own systems. No forums, platforms or feedback systems are currently in place. We appear to be intent on either “reinventing the wheel” or more often “ignoring the wheel”.

Undoubtedly, innovation will lead to better outcomes for patients. AI will be part of this, but we need to act quickly to own it, manipulate it and then export it, otherwise it becomes yet another expensive “white elephant” in our health system, not built for purpose.

Nathan Lawrentschuk is developing an AI application for the early detection penile cancer (as referenced in this article).