David Buckeridge is a health-care detective. A public health physician and an associate professor in the Department of Epidemiology, Biostatistics and Occupational Health at McGill University, Dr. Buckeridge leads the Surveillance Lab, part of the McGill Clinical and Health Informatics Research Group. He and his 15-member team make automated software systems that detect patterns of disease and evaluate public health interventions.
For instance, the Surveillance Lab has created an alert for the prescription of opiates and other drugs that make elderly patients more likely to fall. It built this software with Dr. Buckeridge’s departmental colleague Robyn Tamblyn, who drew upon her research network of about 100 physicians.
Underpinning the alert is a statistical algorithm that pinpoints the probability of a patient falling within the next year, based on factors such as age and medication. Dr. Tamblyn and the Surveillance Lab have embedded the algorithm in physicians’ Electronic Medical Record (EMR) systems. When one of those doctors opens a patient’s record to prescribe a new medication or refill a prescription, a thermometer graphic shows the risk of falling.
“Physicians pay a lot more attention to that kind of alert than to an alert that just says, ‘Opiates are a problem,’” explains Dr. Buckeridge, the holder of a Canada Research Chair in Public Health Informatics. “It’s actionable information right now, about something quite specific and quite relevant.”
A vast field that encompasses everything from patient record systems to telehealth and computer-aided diagnosis, the field of health informatics could transform the medical system by making physicians many times more efficient. But there are ethical concerns about giving computers greater power in a clinical setting.
At Toronto’s Mount Sinai Hospital, internal medicine specialist Matthew Morgan knows the power of health informatics. Dr. Morgan can use the hospital’s Electronic Health Record (EHR) system to review patient information, see lab results, order lab tests, and safely prescribe medications.
“Practising medicine in the absence of an EHR is something that I couldn’t even consider today, because of the advantages it provides in providing better care for patients,” he says.
Dr. Morgan believes health informatics help the hospital see how well it’s managing the patient population: “We use information made available through the EHR to help us understand whether we’re delivering care that is meeting the standards of evidence-based and other guidelines.”
To create their systems, health informatics researchers make extensive use of artificial intelligence (AI). Dr. Buckeridge, who has a PhD in biomedical informatics from Stanford University, combines existing data with powerful statistical techniques capable of recognizing patterns.
Health informatics is also helpful for its computer simulations of disease outbreaks. The Surveillance Lab is currently studying the detection of water-borne illness through a 3D model of the Montreal drinking water system.
The researchers feed infectious organisms into the model to see how they move through it, Dr. Buckeridge says. Working from census data, he and his colleagues have also simulated the local population. By keeping track of how these “people” use the health-care system when they get infected, the Surveillance Lab can determine from where signals of such an outbreak might come. “Then we can evaluate the best way to control the outbreak and the best kind of surveillance approaches,” Dr. Buckeridge says.
Scientists have applied a host of AI technologies to diagnosis and decision support. All such systems have their problems, says Dominic Covvey, president of the National Institutes of Health Informatics (NIHI), based in Waterloo, Ont. For example, making sure the thing works is no easy task.
“All you can do is throw problems at it and see how it does,” says Dr. Covvey. “But trying to show the logical limits of the system and what it can and can’t do properly? That’s a scientific problem. I don’t think anybody knows the exact answer.”
Bruce Buchanan is a professor of computer science, philosophy and medicine at the University of Pittsburgh. For him, one of the most exciting recent developments is IBM’s plan to use Watson – the supercomputer of Jeopardy! fame – for commercial medical applications.
Watson is so powerful because it learned how to integrate the information from tens of millions of documents and make associations in ways that are far from obvious, notes Dr. Buchanan. Likewise, it could help physicians make a quick and accurate diagnosis by combing through medical literature.
“A program that reads everything and integrates it, and then reminds the decision-makers that yes, there are some additional possibilities, can be extremely helpful,” Dr. Buchanan says.
Still, medical software has its risks. As McGill’s Dr. Buckeridge points out, machines can be programmed to do things wrong – and then they do them systematically wrong. “People are increasingly aware of that, and there’s even discussion about whether these software [products need]to be regulated like other medical devices to make sure we know they’re safe.”
Will computers ever replace medical professionals? Instead of that rise-of-the-robots scenario, NIHI’s Dr. Covvey envisions a world where technology enables a physician to handle 10 times more patients. Among other things, that will mean using automated diagnosis and caring for the sick at home through chronic-disease management systems.
“Right now we’ve got the robotic physician – doc in a box, locked in the office,” Dr. Covvey argues. “What we really need to do is free up the physician to deal [with]those situations where real humanity is required.”
For at least a decade, Dr. Buchanan predicts, computers will continue to play a supporting role in medicine. “There are a lot of reasons, one of which is our own preference to deal with a person in a white coat who has the capacity to understand our pain,” he says. “That’s a very strong preference, and one I don’t think we want to give up lightly.”