AI + Healthcare: Why is nobody working on operational improvement?
July 6, 2018 • Reading time 3 minutes
As AI research is making it mainstream there is a trend emerging in healthcare: hospitals are focusing on digital infrastructure and AI start-ups on improved diagnosis and delivery of care. As hospitals are resource constrained and struggle with more elderly patients and chronic illnesses however there is a crucial third dimension that is neglected: AI applied to operational improvement. The reason start-ups shy away from this one is that its tricky to identify the problem, understand the complex system in which it arose and get the behaviour change to integrate the AI solution on the ground.
AI in healthcare provision
The internet is being flooded by articles about AI + healthcare: Brain image recognition, Abdomen scan analysis, Chest abnormality detection, Breast cancer testing, Cardiac imaging, CT Scan analysis, ultrasound analysis, Histopathology, MRI Scan analysis. The goal being to improve diagnosis and treatment.
Meanwhile hospitals try to do their part by focusing on infrastructure and manpower: Chief data officers, chief information officers, cloud based architecture, digital transformation etc..
But while improved infrastructure, improved diagnosis and treatment are both fantastic, a third component is missing: applied AI to operational improvement.
The missing third component: AI for operational improvement
The biggest issues hospitals face currently are rising cost of treatment and increased number of elderly patients with long length of stay, chronic conditions and many comorbidities. Coupled with tighter funding this leads to bed shortages, staff shortages and makes efficient use of resources ever more pressing. The problem is that even with more funding many of these problems do not go away so other means of using resources more efficiently have to be found.
AI algorithms can help with this: Planning (predicting bed requirements, arrivals, cancellation rates), scheduling (theatres, outpatients, staff rotas) and other logistics (ordering equipment, right-sizing new buildings, demand and capacity alignment) are greatly helped by more powerful algorithms that predict and crunch ever more data.
Proper application of these algorithms allows doctors to spend more time on the ground, less time being administrators and the available resources to be used more efficiently.
How come AI start-ups are not jumping on these opportunities? There are three reasons: Problems need to be clearly identified, the complex system in which they arise understood and behaviour patterns adapted on the ground as the systems are implemented.
Integrating a deep learning algorithm in a CT scanner is just a marginal change to the way things operate. Integrating a new scheduling process might disrupt a whole department of booking managers, surgeons, managers and others who have done things a certain way for many years.
But for those who succeed there is a prize to be had – Doctors waste their time administrating systems, scheduling handling data entry all the while hospitals use their theatres inefficiently, have AEs filled to bursting, waste money when ordering hospital supplies, need to plan again and again and again for different types of beds, need to submit data to NHSI for reporting, schedule staff rotas, or do 100s of other things. Improve how these things are done and it might just bring healthcare into the 21st century…
To read more on this topic:
How we overcame the hurdles mentioned above when integrating an AI scheduling system on the ground in a major tertiary hospital: https://www.linkedin.com/pulse/how-stop-operating-theatres-from-wasting-two-hours-each-moroy