Big data in medicine
On 17 October the ECCT's Healthcare Enhancement committee hosted an event on the subject of big data in medicine featuring guest speaker Martin Hiesboeck, International marketing expert and consultant at Geber Consulting.
The era of big data in medicine is still in its infancy but we already have smart devices that are able to monitor and collect information such as pulse and heart rate, blood pressure, glucose levels and temperature. Meanwhile there is plenty of other information such as genome sequencing, medical records (such as medications, test and surgery results and image data) that have been collected and stored digitally which could be very valuable for analysis.
If we take this a step further, if we know when a heart attack is likely to occur, we can fit devices to deliver medication to the patient when it is needed.Great advances are being made in passive sensors (which do not require a power source). These can be used, for example, to monitor glucose levels. Given the rapid developments in nanotechnology, we can envisage that it will soon be possible for everyone to swallow multiple kinds of miniature sensors that travel around digestive systems or can be injected into the bloodstream to collect and send valuable information for analysis on, for example, tumours or blood clots.
By making effective use of this information, warning signs will allow preventative action before diseases actually occur. This is one of the greatest areas of potential for big data medicine. Another great benefit of computer power and big data is that it could eliminate so much time spent analyzing data. For example, IBM's Watson can analyse two million medical image scans per second and predict with an accuracy rate of 99.5% which carcinomas are malign and which are benign. Watson can compare billions of DNA samples against a patient's genome sequence and cancer variant and sifts through millions of pages of new studies, medical journals and clinical records to make hypothetical diagnoses. While some medical practitioners are alarmed that such technology would make doctors obsolete, they are missing the important point that using new technology can free them up to focus on and solve the 0.5% of difficult cases.
Advances in robotics also raise the possibility of creating blueprints for automated surgeries. Given that many surgeries are routine, it is perfectly conceivable that a well-programmed robot will perform these surgeries with greater precision than human surgeons. Besides the ethical question of whether humans will accept being operated on by a robot, this also poses an existential question for surgeons if robots are to replace surgeons.
Big data is likely to be very useful for detecting, monitoring and predicting the spread of outbreaks. However, models will need to be fine-tuned. Hiesboeck noted that all the models used to predict the spread of the last Ebola outbreak gave incorrect results.
Big data will also make clinical trials much more accurate. Current clinical trials are a combination of guess-work and a bit of luck because of the random selection of patients and insufficient samples. Better use of big data will use analysis-based selection and thereby make clinical trial results much more accurate. In the same way, big data analysis will optimize candidate selection to speed up organ donor matching. Big data will also make it much easier to detect fraud and detect and correct errors.
But what about acceptance of artificial intelligence (AI) in medicine? According to statistics cited by Hiesboeck, doctors heed computer's advice about two thirds of the time, regard current algorithms as already very good while 90% of doctors believe that AI is smarter than their own intuition. For their part, 90% of patients would trust AI if it is faster and cheaper.
According to Hiesboeck, 85% of all procedures in a hospital are routine and can be automated. Big data-based automation could theoretically free up 50% of doctors' and 70% of nurses' time to allow them to take better care of patients.
Whether medical practitioners like it or not AI and start-ups will be disruptive. It would therefore be in their best interests to accept the trend and use it, where possible, to their advantage. For example, routine tests will soon be much cheaper and faster than traditional laboratories. While some incumbents will suffer, the reduced costs will be of great benefit to patients and cash-strapped healthcare systems globally.
However, there are several issues to resolve to realise the potential of big data medicine, most notably privacy, security and capacity. There is still a silo mentality in many healthcare institutions. For the full potential of big data to be realized, everyone needs to share their data. Most institutions are unwilling to do so. Moreover, for privacy and security reasons it is usually not legal to share patient data. Another problem is data analytical capacity. By some estimates there will be 5,200 gigabytes of data per person by 2020. There is currently not sufficient hardware and software capacity to properly analyse the data. While many companies are working on analytical tools, once again, they have different methods and are not sharing methods or information because they are competitors.
To survive in the new era, healthcare providers will have to invest in hardware and software, hire data analysis experts, hire computer security experts and develop privacy protection systems, all on a massive scale. According to Hiesboeck, big data offers a great solution for Taiwan where non-communicable diseases (NCDs) account for 79.3% of all deaths. Given that the three major NCDs are cancer, cardio- and cerebral-vascular disease, which can be detected and prevented by big data, by embracing the big data trend, Taiwan could literally save both lives and money by investing in big data solutions.