Data Analytics covers the collection and analysis of structured and unstructured data from a variety of different sources. The fields of application for data analytics in healthcare are many, covering areas within diagnostics, prevention and treatment. The Danish healthcare sector is one of the most profoundly digitised in the World and therefore generates a lot of data. This makes a good starting point for large scale data analytics. There are, however, ethical, technical and legal challenges related to using and sharing data in healthcare. Companies should develop data analytics solutions and services in strong collaboration with users and other experts, and they should build an architecture that supports transparency, a high degree of security and interoperability.
Data Analytics in the future
In this video the experts explore the future potential, challenges and barriers for Data Analytics. Watch and learn about both the values and the risks of Data Analytics.
We define Data Analytics as the process of examining raw data, using analytical and logical reasoning to draw conclusions about the information they contain. Data analytics can be used on all sorts of data both structured and unstructured, for example on numbers, texts and images. It can be used to handle vast amounts of data as well as data from a number of different sources that could be very time-consuming or even impossible for human beings to analyse.
Analysis of healthcare data has been used to improve care delivery and to prevent hospitalisation. Data analytics holds the potential to take previous advancements to a whole new level. We have seen the first glimpses of it when it comes to image recognition and decision support but we are just starting to scratch the surface. We are starting to understand on the one hand the future potential within the field of advanced predictive measures, personalised medicine and behavioural pattern analysis, and on the other hand the negative consequences such as automated inaccurate decisions carried out based on data analytics, which forces a potential distrust and public resistance towards adopting the technology.
Data Analytics is taking various bits of data, information from different sources and different types of information, and breaking it into separate components. These components can then be linked and combined and thereby provide new information and useful insights that would otherwise not have been available. Hospitals are already generating a lot of different types of data for citizen care, medicine admirations, operations etc. and all these data can be used to improve the quality in the healthcare sector through big data analytics.
Data Analytics applies data mining tools to generate qualified datasets based on many different data sources containing semi structured and unstructured data. Niall McDonagh from Microsoft explains:
The development of healthcare technologies has increased significantly over the last couple of years, which means that the healthcare sector is going to face a major transformation in the years to come. As a result of digitalisation the amount of data generated in the healthcare sector is expected to double every 8 months (Mandag Morgen & Danske Regioner, 2017)Mandag Morgen. & Danske Regioner. (2017). Sundhed i Skyen – Et kig ind i den digitale fremtid på sundhedsområdet. Retrieved from https://www.mm.dk/artikel/sundhed-i-skyen-et-kig-ind-i-den-digitale-fremtid-paa-sundhedsomraadet. The potential for data analytics in the future is therefore even more promising than it is today.
Human facial image/video can convey information regarding to expression, mental condition, physiological parameters such as heart-rate and respiratory rate as well as symptoms of disease. This thesis aims to contribute to the development of facial image-based diagnostic and monitoring systems for in-home patient or elderly by employing computer-vision methods.
Development of facial image-based diagnostic and monitoring systems using computer-vision methods imply a number of challenges to be addressed. Major challenges include setting up an appropriate network of sensors for facial data acquisition, pre-processing of facial image data, selecting the area of interest as a subset of captured data, extraction of appropriate features from the data, and employing effective machine learning methods to automatically detect clinically important factors from the captured data.
In this project, emphasis will be put on face extraction from video using a quality assessment, facial skin colour analysis, face alignment and facial feature tracking in video frames, expression and emotion recognition, and finally, developing clinically relevant systems using relevant facial information. The thesis proposes an improved method by using a new heartbeat footprint tracking approaches in the face. The thesis also introduces a novel way of analysing heartbeat traces in facial video to provide visible heartbeat peaks in the signal. The way to extract and utilize this biometric trait in person recognition and face spoofing detection is described in the project.
Data Analytics is influenced by the megatrends of democratisation, increased health focus, acceleration, digitalisation and the paradigm shift in patient groups.
At the same time Data Analytics is an instrument in addressing the demands and pressure that the demographic change presents.
According to Karen Andersen-Ranberg:
Her expertise is epidemiology and geriatrics and she has been the head of several ground-breaking projects investigating the use of Data Analytics in healthcare. Her concern for the future of healthcare is based on knowledge about the demographic composition of the future:
We are facing a burning platform and something must change in order to preserve the level of healthcare that the Danish population has come to expect. Data analytics could be a part of that change and a component of the answer to how we will provide healthcare in the future.
Hal Wolf agrees:
By analysing health data researchers and doctors can use the patterns of data to gain a whole new level of insight into the citizen’s health and thereby markedly improve predictions on how a particular type of treatment will affect the citizen. Data Analytics can also be used to target preventive efforts and reduce the effects of illness and hospital admissions. It can help prevent malpractice because the doctors have more information available and it can help reduce side effects from medicine. In addition, Data Analytics is relevant to the development of personalised medicine as well as being the obvious solution for carrying out routine tasks (such as text and image analysis, literature search etc.) faster and more efficient than humans, which may free up resources.
Niall McDonagh argues that the acceleration, including the increasing technological possibilities, availability of information etc., enables the further exploitation of Data Analytics:
Some of the prominent uses of Data Analytics at the moment lie within data mining, artificial intelligence, deep learning and data visualisation.
Kevin Dean explains some of the ways that Data Analytics are used today:
One of the methods for Data Analytics is called data mining. The basic concept of data mining is to sort through large sets of data in order to identify trends, patterns and relationships. An example of how data mining can be used within healthcare is predictive analysis, which seeks to predict future events such as hospital readmissions by identifying and acting upon trends, patterns and relationships that may be linked to readmissions.
AI (artificial intelligence) is the Data Analytics method that is discussed most and holds the largest potential at the moment. AI uses algorithms to decode large amounts of unstructured data (big data) such as images and texts in a short period of time, much faster than a human being. The operative word is unstructured since big data analysis enables studies to be carried out on data from a number of different and very varied sources such as social media, images, medical records etc. This is data, which was extremely difficult to combine in the past.
Deep learning is a form of AI attempting to simulate the neocortex’s large array of neurons in an artificial “neural network”, i.e. creating an artificial brain, so to speak. In order to emulate the human brain an artificial network consisting of layers of interconnected “neurons” is set up. In connection with e.g. image recognition, each layer analyses the image and communicates its conclusions to the layer above until finally reaching the top layer, which forms the final conclusion. Particularly within image analysis and recognition, deep learning is expected to have a huge impact on healthcare.
Data visualisation plays an important role in predictive analytics and machine learning as the outputs of advanced predictive analytics or machine learning algorithms need to be visualised in order to monitor and show results and ensure that the models are performing as intended.
Identifying patients who unexpectedly deteriorate during their admission to the emergency department challenges hospitals worldwide. In Danish emergency departments, 10 patients unexpectedly deteriorate every day – resulting in heart/respiratory failures, ICU transfer, or in-hospital death.
In this project, the aim is to develop and evaluate a new monitoring platform that utilize existing streams of clinical data from patient monitors to strengthen the situational awareness of clinicians – helping them spot risk patients in due time.
Building such a platform successfully requires close collaboration with subject matter experts and clinicians who would utilize the system on a daily basis. Fortunately, a strong alliance with the management and staff at Fælles Akutmodtagelsen (Emergency Room), Odense University Hospital, has enabled the project to design a prototype of a novel monitoring system. In the work, the system aggregates vital sign readings from the department’s Philips monitors, and couples these streams of small data with clinical logistics systems information to provide clinicians with an overview that build upon their existing scoring systems and mental models.
The prototype has been validated in a feasibility study with very positive results, and is currently being elaborated upon to prepare it for a large scale effect evaluation. This will be conducted as a cluster randomized trial in 2018 at the emergency departments of Odense University Hospital and Hospital of South Western Jutland in Esbjerg. The study will be evaluated in terms of clinical, technical, and economical outcomes.
The Danish Healthcare Sector today is characterised by a significant availablity of data and a strong IT infrastructure for healthcare data. The healthcare sector is actively employing and gaining experience with the use of Data Analytics, and learning from both the positive and more negative experiences.
In Denmark every citizen can be identified by a unique number (social security number also called CPR in Danish). There are also a number of national electronic databases containing healthcare data on for example cancer, use of medication, side-effects, deaths and births, just to name a few. The unique number allows for cross references between the many national databases. This gives Denmark a unique position when it comes to carrying out data analytics in healthcare. According to Peder Jest, Danish hospitals are also at the forefront of implementing and using IT (Jest, 2018)Jest, P. (2018, January 2). Peder Jest. Interview performed by Health Innovation Centre of Southern Denmark.
Denmark has a comprehensive public healthcare system divided into several different sectors; primary care provided by the GPs, secondary care provided by the hospitals, social care provided by the municipalities as well as a number of privately practising specialist doctors, dental care, physiotherapists etc. All of these healthcare providers collect data on the citizens they come into contact with and they are all obligated to report to national databases.
The Danish Health Data Agency provides access to health data in the national databases for citizen, healthcare personnel and others with an interest in healthcare data. The data are anonymised and therefore available to the public. The agency is currently working on a health data programme with the aim of providing even better healthcare through better use of data (Sundhedsdatastyrelsen)Sundhedsdatastyrelsen. Den nationale sundhedsplatform. Retrieved from https://sundhedsdatastyrelsen.dk/da/rammer-og-retningslinjer/om-sundhedsdataprogrammet. The health data agency is also responsible for a national service platform for healthcare data, which enables the use of national registries and services to support the citizen’s individual treatment and care (Sundhedsdatastyrelsen)Sundhedsdatastyrelsen. Den nationale sundhedsplatform. Retrieved from https://sundhedsdatastyrelsen.dk/da/registre-og-services/om-nsp. This national platform also supports the use and integration of data from home monitoring and patient reported outcome (PRO) data. Alongside the national registries for storing healthcare data, this is an important part of the Danish IT-infrastructure for healthcare data at national level.
Data analytics has been attempted in a number of different projects in Denmark already, with varying degrees of success. Some of the positive stories can be seen in case examples and some of the negative are publicly known through the Press, such as the attempt to use the IBM Watson technology in oncology at the largest Danish hospital, Rigshospitalet. Some of the critiques from the doctors are that AI has trouble reading the patient journals, it does not generate new knowledge, has errors in a third of the conclusions and there is no third party or peer reviews to qualify the work that is done (MedWatch, 2017)MedWatch. (2017). Massiv kritik af supercomputeren Dr. Watson. Retrieved from https://medwatch.dk/secure/medicinal___biotek/article9952902.ece.
However, we also see positive examples of the use of different forms of Data Analytics in the healthcare sector today. Karen Andersen-Ranberg argues that we see it used both administratively and in connection with research:
John Christiansen argues that the use of data analytics supports the paradigm shift in which the role of the patient changes:
The expectations for the use of Data Analytics is that it will increase significantly towards 2025 and even further after 2025. There is a good basis for Data Analytics in Denmark, however further testing and documentation will be necessary, even in 2025. Within research the use of Data Analytics will be well-advanced in 2025. Denmark has also decided to focus on Data Analytics when it comes to education and study programmes.
Without a doubt Big Data, AI and machine learning hold a vast potential for providing significant improvements for the benefit of society and citizens. Due to the amount of data already collected and the advancement of IT in hospitals today, Denmark has the potential to gain a real advantage when it comes to data analytics. However, Data Analytics is an unproven technology and its degree of complexity requires thorough testing and investigation. Although we can expect to see some real advantages and results from the use of Data Analytics in 2025 we cannot expect to have realised its full potential until sometime further in the future.
When it comes to applying Data Analytics to research, desk studies and logistics, however, the potential may be expected to be realised in the near future since the data is already available and ready to be tested. In these cases it is merely a question of programming the right algorithms and starting to train them.
In January of 2018 the Danish Minister for Education and Research, Søren Pind, announced that six new study programmes are starting within the field of Big Data in the Danish Universities,and an extra economic boost of 30.4 mio DKK is given to the country’s eight universities. He states that:
And with the health data programme initiated by the health data agency the path is set for Denmark to fully unleash the potential of Data Analytics in 2025.
The use of personalised medicine is expected to take off over the next couple of years and Data Analytics is expected to play an important part in this, as does the availabilty of data. Jonas Moll argues that:
And he continues:
Kevin Dean argues the relevance of long-term analysis:
Studies of long term analysis as mentioned by Kevin Dean have already been carried out and shown great potential, for example within the field of predictive analysis. Set-ups for applying Data Analytics on existing data and using it for predictive analysis are being tested. Others are soon to follow. The key to success is identifying which data to analyse and how to gain access to them. The next step will be combing these data with other data generated by the citizens themselves for example via their use of smart home solutions, wearables or social media.
The overall challenges for Data Analytics consist of challenges regarding sharing data, ensuring data security, anonymising data fully and data privacy.
In order to ensure that data is used in the most profitable way across sectors and across countries, collaboration is necessary. Erik Jylling explains:
Karen Andersen-Ranberg supports him and elaborates on why cross-sector collaboration is valuable:
In May 2018 the new person directive from EU (GDPR: General Data Protection Regulation) will enter into force in Denmark. The purpose of the directive is to strengthen citizens’ fundamental rights when it comes to data, privacy and digitalisation – but also to simplify rules for companies and thereby facilitate growth. Some of the more noteworthy changes enforced by the directive are the possibilities of issuing fines amounting to up to 4% of a company’s annual turnover. Particularly the new obligations on such matters as data subject consent, data anonymisation, data mining/automated processing, is likely to have a significant impact.
When it comes to consent it must be explicit and the citizen must be clearly informed of the precise and defined purpose of data collection. However, many solutions based on Data Analytics are not precise and defined. On the contrary, they are explorative as you look for patterns in data and use the results to ask new questions. The regulation on data security and privacy will therefore be a huge challenge in terms of achieving the full potential of Data Analytics. In fact, according to the EU directive, the citizen has the right not to be subjected to automated decision-making without consent. The potential and risks of AI, machine learning and particularly automated profiling is being discussed quite heavily as the bill is being debated by members of parliament. Since it has not yet been passed we cannot say exactly how the wording of the GDPR will turn out in Denmark.
Niall McDonagh argues that concerns about data security and privacy are relevant barriers:
If data has been fully anonymised then there is no requirement for consent when it comes to Data Analytics. But data must be fully anonymised. ‘Pseudonymisation’ of data is not sufficient. The difference is that pseudonymised data is depersonalised, which means that it cannot be directly linked to an individual. However, by providing additional information, data can be linked to a specific person again. When data is fully anonymised, linkage to a specific person is impossible.
Although the obligations related to e.g. consent are more lenient when it comes to scientific research the daily operations, Karen Andersen-Ranberg sees a definite barrier in the general fear in the public about the use of their data:
Jonas Moll agrees with her on this barrier and argues that we should work towards an acknowledgement of the fact that data will be shared and instead focus on and invest in how we ensure privacy:
One possible answer could be to carry out more scientific studies in order to identify and test relevant sources of data and possibilities for anonymising the data used for analysis when implementing the solution. Further collaborations with hospitals, municipalities, universities and private companies could very well turn out to produce new and innovative solutions within the field of Data Analytics and privacy legislation.
Data Analytics in the Danish Healthcare Sector may offer the following opportunities for companies developing solutions for prehabilitation:
There are, however also som barriers for Data Analytics solutions to overcome:
When developing solutions for healthcare, particularly solutions that handle personal data, the following aspects will be relevant to consider.
In May 2018 the General Data Protection Directive from EU (GDPR) will enter into force in the EU (European Council, 2016)European Council. (2016). The General Data Protection Regulation. Retrieved from http://www.consilium.europa.eu/da/policies/data-protection-reform/data-protection-regulation/. The purpose of the directive is to strengthen citizens’ fundamental rights when it comes to data, privacy and digitalisation – but also to simplify rules for companies and thereby facilitate growth. Some of the more noteworthy changes enforced by the directive are the possibilities of issuing fines amounting to up to 4% of a company’s annual turnover.
In order to adhere to the GDPR, companies may look at the Guidelines for Cybersecurity (ISO 27032).
The regulation regarding data subject consent has been further strengthened and clarified. Consent must be explicit and the citizen must be clearly informed of the precise and defined purpose of data collection. Furthermore the citizen has the right to revoke consent. If consent is revoked the data must be deleted and proof that it has taken place presented to the citizen. This will affect all companies handling data pertaining to the citizen’s health.
Data portability is a new topic introduced by the GDPR. With GDPR the citizen will have the right to data portability. This means that if you collect personal data the citizen has the right to receive the personal data concerning him or her in a structured, commonly used and machine-readable format. They also have the right to transmit those data to another organisation that collects data about the citizen. The purpose of this obligation is to limit the number of times citizens have to answer questions about the same subject matter, e.g. age, height, gender etc.
This is particularly interesting from a healthcare perspective because data might be required to be shared across different organisations in the healthcare sector to a much greater extent than they are today. This might also prove a new business opportunity for companies, since there may be a whole new market emerging for solutions to support data portability, e.g. by providing system integration or sharing information between different IT systems.
In addition to the more general GDPR directive, an updated directive on Medical Devices will enter into force in the spring of 2020 and 2022. The two directives (EU) 2017/745 “MDR” & EU 2017/746 “IVDR” – (European Parliament & European Council, 2017a)European Parliament. & European Council. (2017a). (EU) 2017/745. Retrieved from http://data.europa.eu/eli/reg/2017/745/oj/eng and (European Parliament & European Council, 2017b)European Parliament. & European Council. (2017b). (EU) 2017/746. Retrieved from http://data.europa.eu/eli/reg/2017/746/oj/eng heavily regulate what is defined as medical devices, and how such devices can be tested and used within the boundaries of the EU. This is central for especially Data Analytics and Smart Health Technologies. ‘Medical purpose’ is defined as any type of diagnosis, prevention, monitoring or treatment or alleviation of disease or disability. The vast majority of devices which collect health information are likely to be considered medical devices, even if they do not process or analyse the data. Companies operating within the domain of health should proactively investigate compliance with these regulations and adjust development processes accordingly.
Bringing technology into the sphere of healthcare services brings with it relevant ethical considerations. The Health Innovation Centre of Southern Denmark has developed two videos that illustrate the expectations and challenges that may arise when new technology meets the healthcare sector. The videos focus on the perspectives of the patients at home and the clinicians working across sectors, respectively. Companies may consider these ethical aspects in their development process.
Companies developing Data Analytics solutions for the Danish healthcare sector of 2025 should particularly consider the following:
In order to fully utilise the vast potential held by using Data Analytics in healthcare, the companies developing such solutions need to collaborate with experts from disciplines beyond computer science and engineering as well. Uffe Koch Wiil argues that it is important to work on Data Analytics in interdisciplinary teams with both healthcare personnel and technical experts present – and perhaps also including patients’ representatives’ organisations. The right combination of people is a prerequisite for identifying the right innovative combinations of data on which data analytics will add value (Kock-Wiil, 2017)Kock-Wiil, U. (2017). Oplæg ved Uffe Kock-Wiil. .
Due to the nature of the GDPR it is important for companies working with Data Analytics solutions that they employ experts on anonymisation and perhaps also collaborate with legal experts who can assess whether the collected data may be used for the intended purpose. Being able to anonymise data is key to being able to work with it and train algorithms on it as required in order to success with Data Analytics. Having a clear and transparent set-up for their collection of data and implementing privacy by design in their solutions is also an extremely wise move for any company desiring to be successful with data analytics.
Trust is a key issue in successful implementation and utilisation of Data Analytics for healthcare. Healthcare data is extremely sensitive and the development and use of data is essential for both the development and the use of Data Analytics. Even the slightest hint of any lapses in the security of data could be fatal. It is therefore vital to ensure that proper control systems are implemented and constantly updated. Sound governance processes regarding data use should be in place. Uffe Kock Wiil underlines that “we need to avoid black-box approaches” (Kock-Wiil, 2017)Kock-Wiil, U. (2017). Oplæg ved Uffe Kock-Wiil. .
The legislation that covers this area is complex and varies a great deal between Europe and the US. The implications of not adhering to the legislation can be significant, not only financially but also in terms of loss of trust and thereby loss of future access to data. It is therefore important to work closely with legal experts who can help to ensure the full legality of the solutions developed, as well as with communication experts who can ensure that transparency is achieved. Involving the healthcare personnel and the citizens/public health organisations is important to ensure trust.
Facilitation of data sharing between owners and users of data is paramount for the success of Data Analytics. It is vital to ensure that exchanges are secure and mutually beneficial. As Niall McDonagh explains:
Data Analytics may add value for the citizens, healthcare personnel, healthcare sector and society as a whole in relation to:
However there are also potential risks to consider:
Data reliability and – qualification