Most CIOs get asked questions about how quickly a new software program will be deployed or what it will cost to outsource parts of their IT infrastructure. For Jim Tom, it’s more like this: How long before we can expect to see a decline in the rates of sexually transmitted infections?
The IT executive at Public Health Ontario may have joked about it at the Big Data Innovation Summit in Toronto this week, but it was a good example of how high the expectations have risen around the power of analytics to not only offer insight but meaningful action. Anyone leading a project in this area, he said, had better get pretty good at managing those expectations.
“The path from getting the information and affecting real change is a long path. It’s not six months,” he said. “You have to talk about probabilities, but people like certainty.”
Big data has been touted as a way for firms in financial services, retail and other sectors to get a better handle on their customer’s behaviour and, ideally, increase the volume of business they get from them. Tom pointed out that the situation is a lot messier in an organization like Public Health Ontario, an arms-length provincial agency which is mandated to test and assess potential risks across a wide array of problems.
“Mostly the health system talks about health conditions. Once you get sick, we kind of know what to do,” he explained. “Public Health is about input — it’s about getting at prevention, at control, what happens before you get the condition. As your grandmothers all told you, that ounce of prevention thing really is important.”
It’s also a particularly complex task, because health can be adversely affected by so many different factors. These include genetics, environmental conditions, socio-economic conditions such as income and more. The link between cause and effect, Tom said, can be considerably long.
“I would say that (in the case of marketing analytics), there’s a fairly good idea of what the difference between the signals versus the noise,” he said. “When your’e tracking what a customer’s doing on your Web site, you know what they’re doing. There’s not a lot of noise there. When you’re trying to track what causes someone’s cancer, there’s a lot of noise.”
That doesn’t mean Public Health Ontario isn’t forging ahead anyway. Tom said the agency recently launched a series of dashboards that analyze trends around sexually transmitted infections (STIs) by putting them into data marts which are exposed on the Web and made available to public health units. The STI Dynamic Reports, as they’re called, can help those working in public health units to determine, for instance, who they should be testing for certain STIs, whether a particular age group might be at risk, or how their results compare with other public health units or regions, as well as the level of antibiotic resistance.
Although proponents of big data sometimes paint a picture of harnessing unstructured information and turning it into vital knowledge, Toms said the path at Public Health is much more back and forth. The agency conducts more than four million tests a year for communicable and infectious diseases, and multiple tests can be applied to each sample.
“There is an iterative process where we implement data transformations and present summary totals and test cases for verification, then go back and revise. The ETL process is not so straightforward as to say, we have a set of business rules and then we just apply them against a set of transactions,” he said. “In some cases we had to go all the way back to the lab testing and the raw data and say, ‘What’s going on here?’”
Public Health Ontario is conducting similar analysis in areas like HIV testing. Over time, Tom said the agency may be doing big data work that resembles more of what happens in other sectors, in terms of assessing what marketing techniques work to change behaviours around health.
“The old days when you could give everyone a penicillin shot are kind of over,” he said.
The Definitive Guide to Architecturing End-User Computing Solutions
Determining the right infrastructure is one of the biggest decisions for application and desktop virtualization projects. Without a stable, highly available and high performance infrastructure underlying the design, IT will face any number of challenges during the deployment and operational phases of your End-User Computing (EUC) project.