New statistical model lets patient's past forecast future ailments

Kurzweil Accelerating Intelligence Blog, June 6, 2012

Analyzing medical records from thousands of patients, statisticians have devised a statistical model for predicting what other medical problems a patient might encounter called a “hierarchical association rule model.”

Like how Netflix recommends movies and TV shows or how Amazon.com suggests products to buy, the algorithm makes predictions based on what a patient has already experienced as well as the experiences of other patients showing a similar medical history.

“This provides physicians with insights on what might be coming next for a patient, based on experiences of other patients. It also gives a predication that is interpretable by patients,” said Tyler McCormick, an assistant professor of statistics and sociology at the University of Washington.

McCormick’s co-authors are Cynthia Rudin, Massachusetts Institute of Technology, and David Madigan, Columbia University.

McCormick said that this is one of the first times that this type of predictive algorithm has been used in a medical setting. What differentiates his model from others, he said, is that it shares information across patients who have similar health problems. This allows for better predictions when details of a patient’s medical history are sparse.

For example, new patients might lack a lengthy file listing ailments and drug prescriptions compiled from previous doctor visits. The algorithm can compare the patient’s current health complaints with other patients who have a more extensive medical record that includes similar symptoms and the timing of when they arise. Then the algorithm can point to what medical conditions might come next for the new patient.

“We’re looking at each sequence of symptoms to try to predict the rest of the sequence for a different patient,” McCormick said.      If a patient has already had dyspepsia and epigastric pain, for instance, heartburn might be next.

The algorithm can also accommodate situations where it’s statistically difficult to predict a less common condition. For instance, most patients do not experience strokes, and accordingly most models could not predict one because they only factor in an individual patient’s medical history with a stroke. But McCormick’s model mines medical histories of patients who went on to have a stroke and uses that analysis to make a stroke prediction.

The statisticians used medical records obtained from a multiyear clinical drug trial involving tens of thousands of patients aged 40 and older. The records included other demographic details, such as gender and ethnicity, as well as patients’ histories of medical complaints and prescription medications.

They found that of the 1,800 medical conditions in the dataset, most of them occurred fewer than 10 times. McCormick and his co-authors had to come up with a statistical way to not overlook those 1,400 conditions, while alerting patients who might actually experience those rarer conditions.

They came up with a statistical modeling technique that is grounded in Bayesian methods, the backbone of many predictive algorithms. McCormick and his co-authors call their approach the Hierarchical Association Rule Model and are working toward making it available to patients and doctors.

“We hope that this model will provide a more patient-centered approach to medical care and to improve patient experiences,” McCormick said.

The work was funded by a Google Ph.D. fellowship awarded to McCormick and by the National Science Foundation.

Ref.: Tyler H. McCormick, Cynthia Rudin and David Madigan, Bayesian Hierarchical Rule Modeling for Predicting Medical Conditions, to be published in Annals of Applied Statistics, 2012

Ref.: Tyler H. McCormick, Cynthia Rudin and David Madigan, Bayesian Hierarchical Rule Modeling for Predicting Medical Conditions, 2012, open access version [PDF]

The Intersection of Information and Energy Technologies

Why I think computational power will solve the world's energy problems in this century.

Bill Gross, Technology Review, April 27, 2012

See the rest of our Business Impact report on Computers Storm the Grid.

 

Two talks at the TED conference this year formed, back to back, a sort of debate about the future of our planet. First, Paul Gilding gave a talk entitled "The Earth Is Full," about how we are using up all Earth's resources, with possibly devastating consequences. Next, X Prize creator Peter Diamandis gave a presentation entitled "Abundance," about how we will invent innovative ways to solve the challenges that loom before us. 

I believe that we will need great ingenuity to enable our planet to provide successfully for more than seven billion human beings, let alone the nine billion that will probably inhabit it by 2050, and I believe that information technology will make this ingenuity possible. Because of fluid marketplaces and an ever more globalized economy, nearly every important resource is becoming scarcer and more costly. Evidence of this is seen in the price not only of oil but also of aluminum, concrete, wood, water, rare-earth elements, and even common elements like copper. Everything is getting more expensive because billions of people are trying creatively to repackage and consume these materials. But there is one resource whose price has consistently has gone down: computation. 

The power, cost, and energy use involved in one unit of computation is declining at a more consistent, dependable rate than we have seen with any other commodity in human history. That declining cost curve must be tapped to lower energy prices—and I believe it will be. This will happen as people ask: To achieve my purpose (in designing whatever device or system), can I use more "atoms" or more "bits" (computation power)? The choice will have to be bits, because atoms are going up in price while bits are going down. 

Here are a few examples. When designing a car, one can put a bit more effort into stronger, lighter-weight materials, which will increase energy efficiency but possibly drive up cost; or one can put a lot more effort into using computational power to run simulations that optimize the use of materials. Today, computational fluid dynamics allow a designer to accurately design a new shape of car, put it in a computer wind tunnel instead of a physical one, and test 1,000,000 body designs to improve fuel mileage by significant amounts. This was never before possible for those constructing vehicles. 

In solar energy, large fields of mirrors or photovoltaic panels can be optimized to be lighter, more reliable, and more power-efficient by putting a $2 microprocessor in every panel. An onboard computer that lets each panel track the sun independently replaces previous systems that used more steel, bigger gears, and bigger gearboxes—basically, more materials. 

As little as 10 years ago, the computing power and sensors needed to build a closed-loop, sun-tracking solar panel might have cost $2,000, or more than the panel itself, and thus the system would not have been cost effective. But with computing costs coming down by a factor of 1,000 every 15 years, all kinds of new opportunities arise to improve system design. 

At eSolar, one of our companies, we designed and built a utility-scale solar-thermal power plant with a huge amount of computation embedded into the field of mirrors. We reduced the size of the components, cut the installation expense, and drove the cost of the system down to nearly half what had been achieved before. This experience proved to me the feasibility of replacing atoms with bits. 

The price reduction curve for computing is not over—it's continuing, and each year will open up further avenues for ingenuity. That is important because our current energy resources are not at all easy to compete with. Fuels that we dig out of the ground and burn are extremely cheap. They are, in effect, the concentrated storage of millions of years of sunlight falling on Earth. Ironically, the biggest component of energy costs is the expense of moving the fuel to consumers from where it's obtained—and transportation costs are mostly fuel, too. So we are in a kind of vicious cycle. The way to break free of fossil fuels is to introduce something new to our energy equation that isn't fuel. 

I believe ingenuity in the form of information technology is the only variable that offers sufficient leverage. We need to replace a cheap, unsustainable form of energy with sustainable forms of energy that are equally cheap. The only way to compete with cheap fuels is to be more clever with computation; that is, to use as little of anything else as possible. 

Bill Gross is a lifelong entrepreneur and CEO of Idealab, an incubator for ideas and prototypes that has spun off more than 75 operating companies.

10 ways big data is remaking energy

Katie Fehrenbacher,  Gigaom,  Jan. 29, 2012, 9:30pm

One of the most obvious trends from the big smart grid conference DistribuTECH last week was how much analytics and big data tools will be used to try to remake energy in 2012, from curbing energy consumption, to reducing energy loss, to adding in more clean power to the grid. Here’s 10 ways that analytics and big data will start to shape the production and consumption of energy in the world:

1). Weather data: Having a finger on the pulse of constantly changing weather data on a micro and macro level can help utilities, building owners and consumers optimize their energy consumption habits and promote energy efficiency. Startup EnergyHub recently partnered with sensor network player Earth Networks to use weather data to make a more efficient form of demand response (utilities controlling power consumption). Other startups like EcoFactor, Opower and Tendril also use weather data as part of their energy behavioral analytics.

IBM has long sold a weather prediction service called Deep Thunder to municipalities, organizations and utilities, which use it to do things like tailor their services, change routes, or generate more or less power. I think weather data could some day provide a platform for some very important next generation services and applications for energy efficiency, much in the way that location data is used as a platform for a variety of services.

2). Cell phone data: Cell phones in our pockets are essentially palm-sized sensors and computers sending a constant stream of information to the cloud where companies could one day use that data to create energy efficiency and better energy products. And yes, a lot of that data is private information, but after that data is anonymized it can be used for the greater good of the community — particularly via the billions of cell phones in developing countries. A startup called Jana does research projects around cell phone data in developing countries, and looks to work with NGOs on programs to create better infrastructure, energy infrastructure and resources.

3). Connected thermostat data: One of the biggest trends from DistribuTECH this year was the overwhelming amount of smart thermostats that are now being sold and marketed. Companies can incorporate that thermostat data into data bases that can be used to promote energy efficiency. EcoFactor’s service remembers every time a home owner overrides the automated smart thermostat system and changes the personalized service to accommodate that manual override. Using 100,000 connected thermostats (which produce 5 billion data points each month) EnergyHub found some interesting statistics like folks in cold climates have a lower average heating temperature set point than households in warmer states.

4). Hadoop & energy databases: The open source data base tool Hadoop is well known — and oft used — in the computing worlds. But in the energy and utility worlds it’s quite rare. However, as the amount of energy data has started to rapidly grow from the smart grid, some companies are embracing Hadoop as a key way to manage energy info. Opower tells me it’s using Hadoop (and the company commercializing Hadoop, Cloudera) as an important way to manage its massive energy data streams. Likewise PJM has turned to Hadoop as a way to organize the energy data coming off of a synchophaser sensor project.

5). Clean power data: One of the main goals for the smart grid is to enable the addition of more variable clean power, which is far more unreliable than fossil fuels (the sun doesn’t shine and the wind doesn’t blow 24/7). Analytics crunching the data from a utilities’ energy supply and demand can help make clean power a little less variable, by being able to more accurately predict the environmental conditions, as well as more accurately assess demand from energy users.

6). Electric car data: Electric cars will by their nature be connected cars, using information technology to manage the vehicle charge and location. Utilities will be closely tracking the charging habits of electric car owners in order to make sure that the grid isn’t overloaded in some early adopter neighborhoods.

7). Power line sensors: One of the areas of low hanging fruit for the power grid is the simple task of helping utilities find blackouts more easily and be able to monitor and manage grid outages. That’s partly where sensor systems called synchophasers come in, which can in real time monitor the health of power lines, collecting multiple data streams per second. Expect all major networks to have synchophaser systems installed over the coming years.

8). Real estate data: Startups like First Fuel Software can use big data to make super accurate assessments about buildings and ways to reduce the energy consumption of buildings — without any extra hardware or monitoring software being installed at the building. Things like weather around the building, demographics of the people in the building, and the building’s historical energy consumption can be used to create an accurate projection. The best way to make a building more energy efficient is by getting as much data about the building;s energy use as possible.

9). Variable pricing: Some day when electricity is sold throughout the world at different prices dependent on supply and demand, massive data bases will be needed. This type of variable pricing is offered in some places in the world, but if it ever becomes ubiquitous it will help curb consumption, by offering high prices when energy is being over used.

10). Using behavioral analytics to curb energy consumption: Getting into the brains of energy users is the job of startups like Opower and Tendril (after it acquired Gr0unded Power.) Essentially these companies have collected data on consumers and demographics and they are using it to try to guess the best way to influence the consumer to do things like upgrade their home appliances and lights to more efficient ones.