The National Death Service

There was a disturbing letter in the Telegraph last week about the latest bureaucratic practices in the NHS. According to the authors, the Government is rolling out a new treatment pattern of palliative care into hospitals, nursing and residential homes. They say is based on experience in a Liverpool hospice. It sounds like it is based on a GCSE multiple choice paper

If all the boxes are ticked in the Liverpool Care Pathway, the inevitable outcome of the consequent treatment is death. In the year 2007-2008, 16.5% of deaths in the NHS came about after "terminal sedation". Whether or not a patient qualifies for such treatment appears to be a simple matter of ticking boxes rather than the decision of an expert.

As the authors point out, algorithmic decision making has caused problems in the financial world. I would go further and say that whenever a decision process is based on a "bright-line" measure where falling one side of the line or the other can have enormous consequences distorts behaviour - witness tax avoidance and structured finance arrangements. In medicine, the tick-box approach can be used to excuse incorrect decisions, or even a policy to shorten bed occupancy.

It wasn't always thus. About 20 years ago, I was called into a presentation to hear about an "Expert System" that had been developed at great expense to make corporate lending decisions. Expert Systems were all the rage in the 1980's, powered by mysterious "neural networks", "inference engines" and "fuzzy logic". Somehow they seem to have drited off into obscurity, but at the time they were going to take over the world. Perhaps they did but we never noticed.

Anyway, having sat through the 30 minute presentation as a junior banker (albeit with some knowledge of software) surrounded by senior credit officers more accustomed to fountain pens than computer keyboards, I asked the obvious questions: What happened when the credit applications for previous corporate loans were fed into the system? Did the system reject the same loans as the credit committee, and more importantly, did it reject loans that were passed by the credit committe but which subsequently went bad?

"Um, no" was the reply, "we haven't tried that." And that was the last I ever heard of that particular system.