Two weeks ago, I was invited to speak on the use of big data to fight chronic diseases at the Asia IoT Business Platform, held in Kuala Lumpur on 3 – 4 Nov 2015. Although I was only able to attend on the 2nd day, it was an interesting gathering of technologists and opportunists who are all interested in the potential for M2M and IoT to change the world. IoT = Internet of Things.
It was my privilege to be able to share my personal viewpoints on how data science is going to transform medicine in the near future with an audience of over 200+ people, as well as connect with many new friends across the causeway who are tackling the same challenges. It was very heartening to have several doctors, including a few from the Malaysian Ministry of Health, come up to tell me how excited they are about this future too – thank you!
I hadn’t been back to KL for several years, and it has changed a fair bit. I was most impressed by the KLIA express train from the airport – comfortable, fast and has free wifi on board.
Zaf (my contact at the conference) shared some of these photos from the event.
So – how exactly is data science going to shift the way doctors practice medicine in the near future, especially for chronic diseases?
When you start looking at some of the key problems in healthcare today – the rising burden of chronic diseases, of which more than 30 million people die of thisyear, and a rapidly ageing population in many countries that mean a steep rise in healthcare demand in the next 10 years. As a doctor, you start to think – I can’t just treat just one patient at a time. Doctor’s aren’t scalable, our time is limited. I wanted to treat entire populations. and so today I work with a digital health and big data analytics company. I’m a big believer that there will be few industries that will be more disrupted, more reinvented than healthcare in this decade.
Vinod Khosla – many of you probably know him, he was the co-founder at Sun Microsystems and now runs a prominent venture capital firm, Khosla Ventures. He is quite a visionary. 2 years ago, he famously, or infamously said that data science will be the key driver for innovation in healthcare.
To begin, I want to tell you a story about this good-looking chap. Has anyone heard about John Wanamaker? I don’t blame you unless you lived in the 19th century, but John Wanamaker was a famous department store magnate in America, back then. He was a innovator, a merchandising genius and a proponent of the power of advertising. He said ‘Half the money I spent on advertising is wasted; the trouble is I don’t know which half’. And the problem is obvious – with the traditional forms of advertising available at that time, such as newspapers, billboards – he had no way of knowing how people were responding to them.
And then Google came along. Google AdWords and the pay-per-click model transformed advertising spending – changing it from a business where advertisers paid for ad impressions into one where they pay for results. Google was able to track when someone clicks a link. And by using data about a user’s behaviour, google was able to place advertisements where someone was likely to click – they were able to change the success rate. They knew ‘which half’ of their advertising was effective, and didn’t bother with the rest.
Data and predictive analytics have transformed finance, manufacturing and many consumer industries. Healthcare is slow to the game, but no exception
I hope this doesn’t come as a shock to you – but as doctors, we are experts at practising ‘trial-and-error’ medicine. A patient presents with symptoms ,we make the best diagnosis we can come up with based on our personal knowledge and experience, and prescribe drugs based on standard dosages. And we cross our fingers and hope for the best. I’m not kidding pause. After some time, if it doesn’t work, we change the dosage or the medication, or consider a different diagnosis. Until something works.
We’re treating patients as a sort of average. ‘Evidence-based medicine’ – a core tenet where doctors base our clinical decisions on well-conducted research – is often inefficient because the results in research are summarized and averaged out and we assume that our patient sitting in front of us is an ‘average patient’.
Take breast cancer as an example. It’s a nasty disease and is the second leading cause of cancer deaths in women today. Survival rates for advanced breast cancer are pretty dismal, as can be seen by this chart here – less than 30% live beyond 5 years. There is a drug for treatment of breast cancer, Tamoxifen. For a long time, based on clinical studies, we thought that Tamoxifen was 80% effective. But now we know that it is 100% effective in 70 – 80% of patients, and ineffective in the rest of them. And you don’t want to give it to someone it will be useless for, because it has side effects, like risks of blood clots and uterine cancer. Biochemical markers are the key – we can now tell in advance which patients will respond and which won’t. Availability of new kinds of data and new tools for data analysis underpins this new approach to medicine.
An infographic that appeared in Nature earlier this year very succinctly shows the problem with medications today. These are the top selling drugs in the US, and it shows how many people need to take a drug before one bad outcome can be prevented. For some of these, like Nexium, which is a drug for gastritis and heartburn, its something like 1 in 24 people. And so we know that our medicine doesn’t always work for our patients – we just don’t know which half, in advance. If we can collect enough data about medical treatments and use that data effectively, we can better predict which treatments will be effective for which patients
Healthcare is transitioning into a data rich environment today. In fact, healthcare is exploding with data. In 2011, 150 exabytes of data (that’s 150 billion gigabytes, don’t ask me how many zeros that is) was generated in US healthcare organizations alone. Alongside traditional sources of data like electronic health records, pharmacy records and imaging data, today we have sensors, genomics, mobile apps, social media. These sources of data that weren’t available a decade ago, but now have huge potential to help us better understand our bodies.
But data by itself is not very useful. The key enabler for this revolution is the availablilty of cheap computing power to harness these large datasets to generate new insights into disease. Today Anyone can gain access to a supercomputer using AWS, running some complex computations, all for a few hundred dollars.
There is another infographic that I really like. Each one of us here today is a unique individual – and we can think of ourselves, or our uniqueness, as a summation of various types of data. Like a google maps of the body. Sensors, nutrition, genes, microbiome (that’s an interesting one, I’ll come to that later), laboratory results. If we can make sense of all of that information, we’ll be able to figure out what makes us different from the average patient. This deeper level of personalization of diagnosis, treatments leads to better outcomes, reduced costs from unnecessary treatments and overall better cost-effectiveness
Our human bodies are becoming part of the internet of things too – “the internet of you”. All these wearables, they can glean useful information from our bodies, such as heart rate, respiratory rates, physical activity, sleep quality and more, and transmit them to networks aroudn us. In the past, you needed to be attached to big, expensive devices in the hospitals in order to obtain this kind of data. It is not difficult to imagine how this deluge of health-related data coming from these devices is going to be critically important to healthcare. What is really exciting to me, is not where we are today, but rather the pace of innovation that has happened in this space
Take a look at this sexy thing. This was my grandfather’s wearable tech – well, not really. 6 years ago, this was one of the first wearable computers, the W250 made by glacier computers.
Fast forward another 3 years – In 2012, the pebble smartwatch, raised $10 million on Kickstarter. Now it’s starting to look like something I might wear, sometimes.
Now 1 – 2 years ago : Fitbits. Not too shabby.
And today – we have this amazingly beautiful yet expensive Apple watch. Beautiful piece of technology that not only measures your physical activity and heart rate, but also allows you to be social.
So if you can see the amazing progress we’ve made in the last 5 years, it just thrills me to imagine how it will be like in the next 5 years
EEGs detects electrical signals on the scalp to measure brain activity. In the hospital EEGs are used to detect seizures and epilepsy. At the top left corner, this is how an EEG machine in the hospital looks like – kinda creepy. And at the bottom right, this is how a consumer grade EEG device looks today. You can buy it for all of $300.
Interestingly, this device is being marketed for 2 very important purposes – to help you meditate better, and to boost your golf game by training yourself to have a clearer, more focused mind. Apparently they have pro-golfers using this now. While these devices have not reached the level of accuracy and standards for medical diagnosis, it is very likely within a few years down the road, EEG data and sensors will be widely available, giving us information about our sleep, emotions and more.
Wearables are old. Today, still less than 20% of the general population uses wearables. But when you start going fashionable, it becomes a lot of acceptable to the mainstream audience. Especially when you have Ralph Lauren as a partner. This is a biometric shirt made by OMSignal, a great company. This shirt tracks your heart rate, breathing rate, exertion score and calorie counts. Sure, they still have that funky black box there, but it’s only going to get smaller.
If you are old enough, you might remember Star Trek. Anyone here follows Star Trek? Spock uses the tricorder to examine the vitals of a comrade. And today, we have a real-life tricorder – the Scandu Scout that measures heart rate, temp, pulse oximetry, BP, all just by pointing it to your forehead, and it sends that information to your smartphone. It’s currently still under going FDA review but will be released soon, probably in January next year
And we have the continuous glucose monitors. These devices are implanted just underneath the skin, and sends readings of your glucose levels every 5 minutes to a receiver, 24 hours a day. They have made diabetes care easier, especially for people with type 1 diabetes where its important to titrate your insulin doses carefully.
Genomics: I’m not going to go into the nitty gritty details here, but if you want a key takeaway point – this is it. I’m sure most of you are familiar with Moore’s law, where processing power doubles approximately every 2 years. The cost of genome sequencing has beaten Moore into a pulp – from $100M 15 years ago, it now costs almost $1k to sequence the human genome. Soon, it’ll be the price of a McDonalds Happy meal. And its not only about the cost. The first human genome took 6 – 8 years and thousands of researchers to do it. Today, it can be done in 2 – 3 days.
Within that human genome is a wealth of information, containing several billion base pairs and more than 20,000 genes. Soon there will be genomic data available for millions of people. Combining this with data we generate from wearable and sensors which we talked about before, and you gain the ability to rapidly analyze the correlation between genes, habits and disease. This allows us to develop individualised treatments for disease.
Microbiome: Each of us harbours 100 trillion bacteria in us right now. Our human cells are outnumbered 10 to 1 by the bacteria we carry – we are a walking bacteria colonies. These bacteria help you digest your food, provide essential nutrients and modulate your immune system. Today, we can learn what these microbes can do by extracting and analysing the DNA from them. We have a second genome active in our bodies. More and more diseases are associated with disturbed gut flora – obesity, diabetes, autism. One note of caution : Antibiotics can permanently alter your gut flora. Think again when you take that antibiotics when your doctor tells you you don’t need it.
We’re really only at the cusp of the big data revolution in healthcare, the very beginnings of this. This gives you an idea of the landscape. Most of these are based in the US, where there is governmental support for unlocking the value of data, for example the precision medicine initiative which aims to sequence the genomes of 1 million people. I’m going to run you through 2 companies that are doing interesting work in this space in different ways
Ginger.io. They are data scientists from MIT and are developing a much more fine-grained analysis of mental health disorders.
What they do is install an app on your phone that monitors your behavior. They are able to adapt sensor data via algorithms to better understand depression and mental health issues. Like how many hours you slept last night, how many text messages you sent today, did you go out of the house today or just stayed in your room…whether anything appears out of the ordinary.
It can alert a nurse if some one needs help and allows health providers to call in and check on you. They work with psychologists and when a psychologist uses their software he’ll know which of of his 300 patients are ”in trouble” today. There’s no way to know otherwise. That’s a really significant contribution. Over time, nurses can monitor and take care of their patients 100 times more effectively than current standards.
Alivecor allows anyone to take ECGs from your smartphone and it interprets them automatically via algorithms. All you have to do is place your 2 fingers on the phone casing. These algorithms can constantly monitor and detect atrial fibrillation in cardiac patients. Every month they collect close to a million ECGs and have one of the largest longitudinal ECG databases. They follow-up the evolution of your heart disease. By following that, hopefully we will be able to predict before serious things happen.
Zebra Medical Vision is a new startup based out of Isreal, which is very interesting to me personally because of my background. Diagnostic scanners rapidly getting more precise, being able to provide better resolution and thinner cuts of the human body. A single CT scan of the chest and abdomen that literally takes 5 minutes can generate thousands of images.
Advances in machine learning and computer vision have made it possible to create diagnostic quality algorithms based on big data that surpass current accuracy rates. These algorithms will be able to provide earlier diagnosis of cancer and other disease and unlock incidental findings hidden in the vast amounts of imaging data that we have
I believe in open source, so you can find a copy of my slides here from the conference on Slideshare:
Note: Most of the infographics and photos were not made by me, they were taken from reputable sources and I have tried my best to attribute them accordingly. Apologies if any were missed.