Big data in healthcare: From theory to practice in 4 steps

Big data in healthcare: From theory to practice in 4 steps

Big data is hot! But for healthcare, big data is still mostly in its hype phase. People talk about it a lot, but don’t use it very often in practice. That mostly has to do with ignorance.

Dan Ariely from Duke University put it nicely when he said: ‘Big data is like teenage sex: everyone talks about it, nobody really knows how to do it, everyone thinks everyone else is doing it, so everyone claims they are doing it.’ This needs to change, because big data has a lot of opportunities to offer the healthcare sector. Using it the right way it can make healthcare better, more customized and with less mistakes, which all makes it cheaper and above all better. In other words: bringing value-based healthcare to a higher level. But what is the optimal way to use it?

What is big data?

The Dictionary calls it ‘large unstructured amounts of digital data, especially as a source of information through further analysis.’ We wouldn’t want to contradict this definition, but big data is definitely more than just large amounts of data. A common description by Mark Beyer and Douglas Laney (2012) points this out. They say that big data, in addition to volume, the speed at which the data is generated, changes and spreads is defining (‘velocity’). In addition, there are different types of data that are important (‘variety’). And lastly, reliability (‘veracity’) is a requirement to be able to carry out a useful analysis, see Image 1.

Image 1: What is big data

The added value of (big) data analysis

And if you already have that data, then what can you do with it? A lot. In his research ‘The ‘big data’ revolution in healthcare’ (2013), McKinsey describes the following four types of big data analyses, ascending in impact and complexity:

1. Reporting: what happened?

This is related to simple analyses to gain insights into how many operations have taken place in a hospital over a certain period of time. These results are suitable to present to for example directors, healthcare professionals or consumers.

2. Monitoring: what happens now?

For this type of data analysis, one uses recent as well as real-time data. This makes it possible to compare the current situation to a benchmark or a desired situation. This type of data analysis can be suitable for monitoring the OR scheduling and staffing in a hospital. This data also makes it possible to track contraindications when medication is prescribed or provided.

3. Data mining and evaluation: why did a certain event take place?

Data mining and evaluating take things a step further. These forms of analyses show the mutual relationship between factors, which gives us an insight into the causes and consequences. For example, you could discover an (unexpected) relationship between patients who got an infection and the hospital room these patients were in.

4. Predicting and simulating: what is going to happen?

Predicting the future based on data analysis is the most complex, but also the most interesting part. For these analyses, you don’t just need a big data set, but you also need the necessary technological and statistical knowledge. This is not just about processing patient-related data, but also about the comparison of for example the most recent scientific professional literature, comparable situations in other patients and the effects of treatment. This means we can use predictive algorithms to give a diagnosis and treatment advice.

Image 2: Big Data Analytics

Big data mid-19thcentury In August 1854 there was a cholera outbreak in the London district of Soho. Hundreds of inhabitants died within a month. Because the cause of this sudden outbreak was unclear, doctor and scientist John Snow tried to plot victims on a map of the district based on their address information. It soon became clear to him that most of the victims lived near a water pump on Broad Street. By collecting information in a smart way, Snow found out the root cause of the outbreak and the council could act accordingly. The water pump was cut off and the cholera epidemic ended quickly. They also took measures to prevent the same problem from happening in other areas and cities.

Big data-analysis in practice

Big data offers lots of opportunities for healthcare. But before we can reap the benefits, we have to go through a number of steps.

Step 1: Report outcomes instead of processes

If you do report something, report the useful things. So, no process indicators – which is what usually happens now – but outcome indicators. Examples of outcome indicators are quality, health and cost data. Mainly focus on quality and health. It allows you to serve your patients and attract the right healthcare providers. They would rather work in an organization where the focus is on improving healthcare and increasing patient satisfaction, than on lower cost. It kills two birds with one stone. To be sure you report the correct outcome measures, it is important to determine clear goals, outcomes and definitions beforehand. Look at which data sources are available. An evaluation matrix can help you gain and keep a clear overview.

Step 2: Monitor outcomes in result dashboards

As an organization, you want the best healthcare with the best outcomes at the lowest possible cost. To achieve that, it is important to put the right accents on your care provision. It helps to regularly keep everyone informed of the main outcome indicators. Healthcare providers can see they are performing and where there are possibilities for improvements. Result dashboards for patient experiences, quality goals and cost goals can offer these insights. Discuss these dashboards during regular meetings. And make sure that all dashboards are accessible to everyone at all times, so that people can check their progress at any moment.

Step 3: Detect variations in practice on the basis of outcomes

If you encounter variations in practice on the basis of your outcomes, this asks for a deeper analysis. Because what is the explanation for that variation in practice? By comparing processes and methods, you can improve. This allows your outcome indicators to function as a base point for process innovation and quality improvements.

Moreover, it is important to not take the outcomes of a specific group of patients, but of your entire patient population. By isolating a certain group from the start, you distort the image, because processes are intertwined. The multimorbid patients for example, cannot be labeled under one disease profile. By taking a holistic approach, you can get a real image of how your entire practice functions.

Step 4: Predict disease and adverse outcomes

The last step, predicting the disease and adverse outcomes, is possibly the most beneficial part for a healthcare organization. International research has pointed out that it is most reliable to use diagnostic analysis models and algorithms. Not everyone agrees with that last bit. Which doesn’t make sense, since algorithms are a lot better at tracking patterns than the average physician which can be a big advantage for your organizations. You can use it to detect high risk patients, assess economical and clinical risks and make your practice more efficient which allows you to create more capacity. This leads to better quality of care and a better negotiating power with financers like governments and health insurers.

Where do we stand and what needs to be done?

This is the theory, but where do we stand in practice? It could be much better. We will try to clarify this through figure three. Vision and culture, structure and governance and interprofessional cooperations form the basis for better quality and lower costs – value-based healthcare. But in addition, you need to take care of the financing and have your data sets in order. You can’t see those two as separately from one another. Once your data is correct and you apply the right analyses, you can work with value-based payment models. Because it is only then that you know exactly which risks are present in your patient population. At the moment data collection, especially in primary care, leaves a lot to be desired. Without a good dataset, useful analyses are impossible, and it is difficult to reach good financial agreements with health governments and health insurers.

Image 3: Value-based healthcare organization

So, it is important to start working with data. Make sure your organization does it itself, without the intervention of third parties. That prevents miscommunications, and forces you to develop or hire the right skills. Register the outcomes measures that are important to your organization and make sure everything is measured correctly. It takes times but will pay for itself eventually. And don’t be afraid that you need to reinvent the wheel. There are tested international standards to take measurements that for some inexplicable reason just have not made it to practice yet. Use those!

For the analysis, we advise you use diagnostic models. That isn’t easy. You need some smart people to do that. People that are good at math and understand healthcare. There are not a lot of those. As a hospital, it is still smart to have these people in-house. For primary care organizations this is a lot more difficult, but that can be overcome by organizing it at a higher level, for example within a care group. That way the primary care providers can also join a step that the healthcare system will really have to take to get their healthcare costs under control and keep them there.

Read more about value based care strategies in our whitepaper.