Why “big data” needs to be made small

Why “big data” needs to be made small

Before we can realistically prevent chronic diseases, we first need to measure the variables that count, argues Nightingale Health CTO Antti Kangas
Nightingale- Antti Kangas - Why “big data” needs to be made small

Wearables, such as smartwatches and fitness trackers, have spread quickly, generating an enormous amount of biometric data for us to interpret. With a plethora of sensors available on the market and personal health data stored in the cloud, you’d think it would be easy for us to take more control over our health. However, global health figures paint another picture, one drastically at odds with Silicon Valley executives using high-tech devices to optimise their health. The global problem of chronic diseases continues to grow, outstripping cancer and communicable diseases as the primary cause of mortality. A recent publication in The Lancet, estimated the global cost of diabetes to be $1.31 trillion (~€1.1tn) in 2015¹. Ischaemic heart disease and strokes remain the world’s leading causes of mortality, accounting for a combined 15 million deaths in 2015². Chronic diseases are a macroeconomic problem too, even though they are largely preventable.

It seems that we’re constantly measuring a multitude of things in health, but these measurements fail to provide us with an effective management system for chronic diseases. Peter Drucker famously said, “you can’t manage what you don’t measure.” If we realistically hope to improve global health, we need to start measuring the right variables and quickly.

To understand this issue better, let’s take a look at what currently is measured. Today’s devices typically produce readings for BMI, physical activity, calories, heart rate and quality of sleep. However, all of these variables are indirect measures of health. One aspect they share is that they’re mainly reflective of an individual’s behaviour, not the internal physiology that determines a person’s true health state.

So, what would be a more representative measure of health? The entire concept of modern healthcare has been built around the opposite of being healthy; a patient is diagnosed with a disease due to their symptoms and a corresponding treatment is provided. Preventative medicine is a new approach that turns this idea on its head. When we aim to predict the future onset of a disease, a diagnosis is never definitive. Instead the focus is on monitoring the baseline, identifying disease indicators and reducing an individual’s disease risk by following up the effects of minor interventions.

Changing our perspective from the status quo to a new system of preventative care, provides us with novel tools with which we can accurately assess health.

Quality, not quantity

In order to measure the likelihood of a person developing a chronic disease, we need better quality data.

There are three key tasks we need to achieve if we are to successfully implement disease prevention: first we need to be able to identify those who have the greatest risk of developing a disease, then we need to define medical and lifestyle interventions to help those individuals, and finally, we need to measure outcomes to find out if interventions have worked. Monitoring the efficacy of treatments will also help us to identify the best practices for different individuals.

All these tasks have their own challenges and identifying those most in need of interventions is far from straightforward with current routine tests. For example, in a nationwide study conducted in the University College of Los Angeles (UCLA), looking at the medical records of over 130,000 individuals, almost 75% of patients that were hospitalised with a heart attack had normal cholesterol levels.³

Likewise, at the individual level, there are currently few ways of accurately telling how well preventive interventions have worked. With current data, efficiency can only be assessed after a decade, when a certain percentage of the treated population has already developed a disease. Sometimes an intervention requires prescribed medication. Many clinicians argue that this should be considered as the final line of defence. The sooner the risk for a chronic disease is detected, the more likely it is that simple changes in lifestyle are all that’s required to keep an individual on the right track. Preventive actions at an early stage are also the best option to turn back the tide of rising costs for treating chronic diseases.

We need to be able to generate high quality data in healthcare that is focused on keeping people healthy, rather than solely treating patients that are already sick. Switching focus in healthcare will be a fundamental paradigm change and requires a great deal of leadership and collaboration to achieve. It is by far though, the best strategy towards addressing the global burden of chronic diseases.

Measuring on a molecular level

Molecular level data provides us with an opportunity to drive innovation in healthcare systems. With routinely measured and comprehensively detailed molecular data, it’s possible to predict the risk of developing a disease, suggest appropriate medical or lifestyle interventions, and keep people motivated through constant feedback. As the positive internal changes that occur inside the body as a result of exercise or improved nutrition don’t always manifest themselves clearly, it can be difficult for people to recognise the benefits of interventions. Instead of using gadgets to give us ambiguous health data that basically tells us more about we’ve doing physically, molecular data has the power to prove the direct effects of interventions by accurately telling us about our current health state.

Measuring health directly on a molecular level can also put ubiquitous activity data into better context. It offers us the chance to be able to learn what kinds of choices have the biggest beneficial effect on our health in the long run, not to mention discovering what aspects of our lifestyle are having a negative impact.

Imagine if there were a scientifically accurate number that told you how healthy you are. You’ve just had your blood analysed and the concentrations of the molecules circulating in your blood have indicated that, although you are still doing well (no need to worry about having a heart attack in the next five years!), your number has actually gone down compared to the last time. Wouldn’t it be nice to get those health points back up?

The powerful potential of molecular level data is immense. A trusted health score would not only create a nice positive feedback loop but also act as a reliable indicator for the risk of developing a serious health condition. Such a risk would also be quantitative and individual – providing you with a timeline of your health progress.

Feasible implementation

A measure for health might sound like an unrealistic concept to implement. However, with appropriate data, it’s perfectly feasible to achieve. If we can change focus from the “big” data that records activity and focus more on smaller variables, such as molecules produced by chemical reactions (metabolites), we can develop a comprehensive picture of our health. For chronic diseases, this would require shifting our current emphasis on collecting data from mainly sick people, to people in various different stages of health.

With an approach that incorporates molecular data, we can start implementing data-driven healthcare for chronic diseases and uncover the most efficient strategies for treatment. The tools and technologies are available today, the biggest challenge is how to start implementing them to bring about a preventive healthcare system that’s effective for all.

Antti Kangas

Antti Kangas also debated this topic as part of Upgraded Life Festival’s panel discussion on “Health Data is everywhere – then what?” We will also be at Slush this year, so if you want to learn more about Nightingale and our blood analysis service, please get in contact with us to arrange a meeting ( and let’s have a chat! Otherwise, you can also contact us via our website:


  1. Bommer, C. et al. The Lancet. Vol 5, 6, 423-430 (2017)
  2. World Health Organization. The Top 10 causes of death – Fact Sheet (2017).
  3. University College of Los Angeles. Most heart attack patients’ cholesterol levels did not indicate cardiac risk – 2009.
Back to top