Let's look at a 2nd example where big data can have a big impact on saving lives. I mean, literally saving lives one life at a time. I collaborated with a number of world-class researchers in San Diego, and an industrial group who are dedicated to improving human health through research and practice of precision medicine. What is precision medicine? It is an emerging area of medicine targeted toward an individual person. Analysing her genetics, her environment, her daily activities so that one can detect or predict a health problem early, help prevent disease and in case of illness provide the right drug at the right dose that is suitable just for her. Very recently the White House and the National Institute of Health here in the U.S. have declared it to be a top priority area for research and development for the next decade. The expected learning outcome of this video is for you to give example of sensor, organizational and people-generated data used in precision medicine. And, explain why the integration of different kinds of data is critical in advancing healthcare. For any technology to succeed in real life we need not only a certain level of maturity of the technology itself, but a number of enabling factors including social economic environment, market demands, consumer readiness, cost effectiveness, all of which must work together. Why is big data for precision medicine important now? Let's see. An important aspect of precision medicine is to utilize an individual's genetic profile for his or her own diagnoses and treatment. Analyzing the human genome, which holds the key to human health is rapidly becoming more affordable. Today's cost to sequence a genome is less than 10% of what it cost just back in 2008. But, human genomic data is big. How big? In a perfect world, just the three billion letters of your genome would require about 700 megabytes to store. In the real world, meaning the kind of data generated from genome sequencing machines, we need 200GB to store a genome. And it takes now about a day to sequence a genome. We are finally beginning to create more electronic records that can be stored and manipulated in digital media. Most doctors offices and hospitals now use electronic health record systems which contain all details of a patient's visit and lab test. How big is this data? As a quick example The Samaritan Medical Center Watertown New York at 294 that Community Hospital reported 120 terabytes as of 2013. The data value more than double in just the last two years. So clearly just in a past two years dramatic changes have prepared the health care industry to produce and analyze larger mounts of complex patient data. To summarize what we have seen so far the key components of these changes are: Reduced cost of data generation and analysis, increased availability of cheap large data storage, and they increased digitization of previously paper records. But we need one more capability to advance toward the promised land of individualized health care practices. We need to combine various types of data produce by different groups in a meaningful way. Let's look at this issue from the same point of view as Ilka did. With her discussion of how big data can help with wildfire analytics. The key is the integration of multiple types of data sources. Data from sensors, organizations and people. In the next few slides, we look at each of these, and then I'll share a story about some of the new and really exciting ways people data especially has the potential to change healthcare big data landscape. Let's start with sensor data. Sure, digital hospital equipment have been producing sensor data for years, but it was unlikely that the data was ever stored or shared, let alone analyzed retrospectively. These were intended for real-time use, to inform healthcare professionals, and then got discarded. Now we have many more sensors and deployment. And many more places that are capturing and explicitly gathering information to be stored and analyzed. Let's just take a new kind of data that's increasingly becoming common in our daily lives. Fitness devices are everywhere now their sales have skyrocketed in the last few years. They are in wristbands, watches, shoes and vests, directly communicating with your personal mobile device, tracking several activity variables like blood pressure, different types of activities, blood glucose levels, etc at every moment. Their goal is to improve wellness. By having you monitor your daily status and hopefully improve your lifestyle to stay healthy. But the data they generate can be very useful medical information because this data is about what happens in your normal life and not just when you go to the doctor. How much data do they generate? The device called FitBit can produce several gigabytes a day. Could this data be used to save healthcare costs, effect a healthier lifestyle? That's a question mark. It's safe to guess that this data alone wouldn't drive the dream of precision medicine. But what if we consider integrating it with other sources of data like electronic health records or a genomic profile? This remains an open question. This is an open arena for research that my colleagues at scripts are doing. It's also a potentially significant area for product and business development. Let's look at some examples of health related data being generated by organizations. Many public databases including those curated and managed by NCBI, the National Center for Biotechnology Information, had been created to capture the basic scientific data and knowledge for humans and other model organisms at the different building blocks of life. These databases carry both experimental and computed data that are necessary to observations for unconquered diseases like cancer. In addition, many have created knoweledge-bases like the Geneontology and The Unified Medical Language System to assemble human knowledge in a machine processable form. These are just a few examples of organizational data sources and governmental data gathered by health care systems around the world could also be used as a massive source of information. But really some of the most interesting and novel opportunities seem likely to come from the area of people generated data. Mobile healths apps is an area that is growing significantly. There are apps now to monitor heart rates, blood pressure, and test oxygen saturation levels. Apps, we might say, record data from sensors but are also obviously generated from people. But there is more people generated data that's interesting beyond censure measurements. In 2015 the Webby People's Voice Award went to an app which supports meditation and mindfulness. Rather than an electronic sensing device, a human would indicate how many minutes per day they spent meditating. If they interact with the app which reminds them to be mindful, then we have human generated behavior that we couldn't get from a sensor. There are well over 100,000 health apps today in either iTunes or Google Play. And by some estimates the Mobile Health App market may be worth 27 billion dollars by 2017. So really we are just seen the beginning of what data might be generated here from what's being called human sensors, but to really understand where the power of people generated data might take us in the era of big data for healthcare. Let's imagine how things stand now. In general, a patient goes to see their doctor and maybe their doctor asks if they have had any side effects from their medications. The accuracy and hence the quality of data patients provide in this kind of setting is very low. Not that it's really the patients fault. It might have been days or weeks ago that they experienced something. They may be unsure whether something they experienced was actually a reaction to then report it. And there might be details about exactly when they took a medication that are meaningful, but they've forgotten it after the fact. Today, people are self reporting reactions and experiences they are having. We're on Twitter, on blog sites, online support groups, online data sharing services: these are sources of data that we've never had before that can be used to understand in a far more detailed and personal rate the impact of drug integrations are responses to certain If applications were designed to integrate doctor and hospital records with information on when drugs were taken and then to further mine social media or collect self reports from patients. Who knows what kinds of questions we will be able to answer? Or new questions we may be able to ask?
