Norman swan
Robots in healthcare can solve problems – but they can also create them.
Take the surgical robots which, using minimally invasive techniques, remove prostates, fix damaged heart valves, operate on the brain and remove bowel cancers. This list grows by the week. The robot is manipulated by a surgeon who is not even scrubbed up and could be across the world from the operating theatre.
It sounds great, and there’s high demand from patients who believe robot surgery must be safer and more effective. True, there is evidence that these patients bleed less, have smaller incisions and go home sooner. But there is little data to suggest that major complications, such as those arising from prostate removal, are reduced.
Cardiac surgery may be an exception. A human surgeon must fully open the chest – a major procedure in itself – but robotic surgery requires only small incisions. Since heart valve surgery often needs to be re-done because the repair fails or a synthetic valve wears out, it’s probably better to avoid opening the chest each time. Nevertheless, using robots in cardiac surgery does have its risks. For example, an important part of the operation is at the end when the heart must be emptied of air to avoid the risk of stroke. Robot surgery can do it, but it’s much easier when the heart can be manipulated by hand. It’s particularly hard to get rid of the air when a person has a stiff heart, say, as a result of multiple past operations. These patients would probably be poor candidates for robotic surgery.
All in all robot surgery has been marketed with great effectiveness by robot manufacturers and also by surgeons themselves who want to be seen to be cutting edge (pardon the pun). But behind the dazzle, there’s a downside for hospitals and patients. The machines are expensive, as are the consumables used in each operation. Then there’s training time and the inevitable number of patients who are exposed to risk before surgeons fully master the technology. In the US there’s evidence that the use of robots in procedures such as radical prostatectomy has driven some surgeons away, probably because of the effort and time required to re-train. That has narrowed the market and driven up the already high costs of healthcare for few added benefits.
But other examples of robotics do sound like good news. In Queensland country towns that are too small to support a paediatrician, a human-looking robot does ward rounds. It’s wheeled to kids’ bedsides by a nurse. Inside is a tele-health communications link to a paediatrician in Brisbane who can see the child, hear the story and advise on care. The “robot” is really just a cute trolley for a camera and a computer with Wi-Fi.
Chemical pathology testing is mostly done by robots these days, with barely a human hand touching the sample as the machine spits out data about your cholesterol level or the state of your liver after last night’s bottle of wine. Automation has lowered pathology costs and probably reduced unacceptable variation in test results. But some might argue it has also made doctors intellectually lazy since they are more likely to order a range of tests, rather than think through the various likely diagnoses.
Most doctors use software programs known as 'decision support tools'
In neurosurgery, robotic techniques place probes into the brain more accurately than the human hand can manage, or focus radiation beams to zap tumours – one version of which is the so-called cyberknife (though critics claim some of its benefits have been overstated).
And then there’s robotics of the intellect. Healthcare is becoming more complex because people live longer and acquire multiple conditions which will never go away. So it’s easy for physicians to miss something important, such as not considering an alternative diagnosis, or forgetting about interactions between different drugs.
Can machines improve the quality of doctors’ decision-making? Most doctors now use software programs known as “decision support tools” when prescribing medication, but the science of decision support – which I suspect fits the definition of robotics – is still immature.
Often the proven data is not sufficient to justify automating decisions in the complex real world. Randomised controlled trials tend to study people based on one condition and one treatment at a time. But real patients have multiple conditions, receive multiple treatments and have their own preferences about what they’re willing to accept. Until data that reflects this type of complexity starts appearing, the robotics of the medical mind will be limited.
It’s enough to give Marvin, the paranoid android in Hitchhikers’ Guide to the Galaxy another migraine.
