Monday, 6 July 2020

Assisting doctors

Not too rocky a ride on the IBM share price

Prompted by an accident backing up some of the stuff on my laptop, I took another look at a short book on the Kindle, that is to say reference 1. Where there is talk of IBM building on its Watson platform to provide a cognitive assistant to cancer doctors.

Some readers may remember the days when IBM was the unchallenged leader in the IT industry, the days when Microsoft and Google were yet to be invented. But they made good use of the cash pile they built in their glory days and they are still a thriving company. Others may remember a contest in 2011 between an IBM computer called Watson (reference 2) and a champion at a game called Jeopardy!, a contest which IBM won.

I don’t know much about Jeopardy! except that it is a very successful general knowledge quiz show on television in the US. I comprehensively failed the candidate player test thoughtfully provided on the web site (reference 3), partly because the questions had a US flavour, partly because my general knowledge was not up to it and the mental agility was not there. So getting a computer to play this game and win was a major achievement by IBM.

They are now leveraging their investment by turning the Watson software into something more general purpose, something that can be tailored to help professionals in all kinds of demanding fields – including cancer doctors, oncologists, in serious hospitals. The relevant bit of the Wikipedia article at reference 4 suggests that this particular project has had a chequered history, and a cursory search of the IBM site fails to find it – although it does find Watson helping with the coronavirus (reference 5). There is also the point that IBM made a major investment in Watson and would be looking for a return. And at some point any such investment would be reviewed by the board – who might well chose to pull rather than to sink more money into the project. You don’t get to be a big and powerful company by pouring good money after bad. The same issue cropped up in the world of big pharma, as touched on at reference 7.

But that is not the present point. Let us suppose instead that IBM succeeded in building a cognitive assistant for oncologists. An assistant which had hoovered up all the scientific papers about cancer, had access to huge numbers of (anonymized) medical records and to the medical records of the current patient, whoever that might be.

An assistant which provided help and support for the oncologist which might not otherwise be available to him on a time scale which was relevant to the patient. The example given in reference 1, written at a time when IBM were more gung-ho about all this than they appear to be now, was a non-smoking lady of Japanese extraction with lung cancer who was worried about losing her hair during treatment. Watson was able to dig up information which was very relevant to this particular lady. He was able to talk to the medical insurance computer about the treatment that he was proposing. All of which sounds entirely plausible and I think it likely that such systems will exist in the near future, even if they are not in routine use just now.

But how would we feel about a corporation like IBM owning a world beating system of this sort? They might be getting most of the data it uses for free – free-loading, as it were, on today’s open access science – but they will certainly be charging to use their system. Reference 2 is all about getting paying customers.

How do doctors learn to trust such a system? There is talk at reference 1 of Watson being good at telling its customers why it has come to the view that it has, at setting out its lines of reasoning, which customers can check for themselves. Bearing in mind, that once such a system is being used routinely, such checking would probably not be practical. Much better to get on with treatment and move onto the next patient.

Would there be problems with getting patients to trust such a system? Would they need to know? Would they need to be told? How would they feel about Watson negotiating with the computer systems operated by their medical insurers about the treatments both appropriate and available to them?

Now my lefty view is that our National Health Service offers a far better model for the delivery of health services than the insurance based model which rules in most of the US. The ‘most’ here prompted by the thought that Veterans’ Affairs (at reference 6) runs a large health service, of the same order of size as our own, despite such things being anathema to most Republicans. Notwithstanding, I was impressed by the thought that there is merit in a busy consultant oncologist having his recommendations for a patient checked by the medical insurance people. They do provide a second opinion, they do stop doctors setting themselves up as infallible gods. It would be interesting to find out how our NHS addresses the issue.

Then how would this affect medical negligence insurance, another big industry in the US and a growing one here in the UK. Who gets blamed when our cognitive assistant makes a mistake, as it will from time to time?

One supposes that anyone, in this case IBM, selling such a system, would need to include insurance cover in their offering. And they might well need to negotiate with the insurance providers about their quality control. How careful had they been to make sure that Watson picked up some obscure paper from some obscure corner of the world – a paper which would have been very relevant to a case which has gone badly wrong? Would it make a difference whether the error was a software error during construction or a clerical error during operation? Perhaps a clerk failing to follow up the non-arrival of any material from said obscure corner? Perhaps a corner which had forgotten to pay its subscription to something or other.

Are there circumstances in which the insurance provider would withdraw cover temporarily, rather in the way that Boeing 737 MAX jets have been grounded – near 400 aircraft for more than a year now. Not to mention the similar number in the production pipeline. In what circumstances might a health service provider be obliged to turn off such a support system, so inflicting massive damage to the service it is able to offer its customers, its patients?

I supposes these same or comparable questions have already been asked about the provision of expensive bits of hardware, like scanners, now an integral part of health service provision. In any event, there are lots of questions.

References

Reference 1: Smart Machines: IBM's Watson and the era of cognitive of computing - John E. Kelly III and Steve Hamm – 2013.





Reference 6: https://www.va.gov/

 
Reference 8: https://www.macrotrends.net/stocks/charts/IBM/ibm/stock-price-history. The source of the snap above. They also gave me daily price information since 2nd January, 1962. Near 15,000 lines of it.

No comments:

Post a Comment