Artificial intelligence Part 6: specific industry impacts - healthcare

ChatGPT image

The impact of AI on healthcare is more nuanced and varied than most other white collar sectors. Healthcare is more complex due to need to retain a strong physical human presence in the medical and care functions that cannot be automated or replaced by digital technology. Healthcare has a complex set of regulations, legal liability and an irreducable human dimension with doctors, nurses, allied health professionals required for direct patient face-to-face contact inclusive of the use of tele-health services.

In contrast, administrative and diagnostic supportive functions are highly exposed. A summary is provided below and is not exhaustive -

Clinical diagnosis and decision support

• Radiology already has AI systems that match or exceed radiologists in detecting certain cancers (breast, lung, skin). The aspect of concern is potential over diagnosis due to the sensitivity of the digital systems used. The radiologist role is shifting towards oversight, complex cases handling and AI exception management. The potential risk in this field being the volume of radiologists needed to undertake radiology functions may reduce.
• Pathology has a similar pattern to radiology as AI can analyse tissue samples at scale. The role of pathologists however is not removed at this time but is being augmented.
• Diagnostic support to General Practitioners can be provided through AI tools that synthesise patient medical history, symptoms and test results. These AI tools are being deployed already however the intention is to support the medical service provided by doctors to patients not substitute it. A secondary intention is to reduce the need for specialist referrals however this has yet to be achieved.
• Dermatology and ophthalmology are two specialties that are heavily dependent on pattern recognition and will face some AI encroachment, however as with other diagnositic tools it may be a supportive function not a medical role replacement one. 

Clinical administrative functions and documentation

• Medical transcription is already largely automated with voice-to-text using clinical AI being widely used.
• Clinical note writing is being addressed by ambient AI scribes such as Nuance DAX. Documentation can consume 30-40% of physician time and assists medical practitioners to achieve quality of life improvement however it reduces medical transcription services significantly, if not in some cases, entirely.
• Prior authorisation, coding and billing  are very large cost centres and are being progressively automated and threatening large administrative workforces in hospitals and insurance companies.

Nursing and Allied Health

• Triage and patient monitoring can monitor patient vital signs, identity and alert to deterioriation and prioritise nursing care. AI provides service augmentation but not replacement of front line nursing care which must be physically provided.
• Care coordination roles do face pressure from AI that can track patient journeys, identify gaps and schedule follow-ups. At this time however this remains an augmentation tool rather than a job replacement one. 
• Bedside care, emotional support and physical nursing are strongly human services and cannot be replaced by AI. It remains one of the most protected areas across all industries. 

Pharmaceuticals and medical science/research

• Drug discovery timelines are being compressed by AI which reduces some research roles but creates new roles in AI-guided drug design. An example of AI impact is Alphaford's protein structural predictions which transformed structural biology.
• Clinical trial design and patient matching is being assisted by AI but not replaced by it. 

In healthcare, it is the administrative organisational pyramid that is being compressed with headcount reduction. Clinical roles continue with signifcantly increased volumes of patients possibly over time.