Better care through the use of medical data
Better care through the use of medical data
‘Healthcare', 'hospitals' and 'patient records' are probably not terms you immediately associate with KPN. Yet the technology company plays an all-important role in the world of medical data with KPN Health Exchange. What technologies are used in this platform? And what does the future hold?
A lot has happened in healthcare in recent years. In particular, the 'compartmentalization', whereby various agencies and care providers have access to medical information, but do not always exchange it efficiently, causes unnecessary delays. KPN's Health Exchange allows healthcare organizations to exchange medical information more quickly, easily, and securely. This is a SaaS platform that is connected to various agreement systems in the healthcare sector.
A system like Health Exchange has several advantages. Information is easier to share between healthcare professionals, medical information is more transparent for patients and healthcare professionals experience less administrative burden. To make the last point concrete: instead of constantly having to fill in forms, data is already recorded digitally and in a structured manner.
As a former nurse, Teun - now Information Manager at KPN Health - tries to link digital technology to the human dimension. He has only been with KPN for six months but was seduced by its vision of using medical data. One thing is certain: interpreting patient data is becoming increasingly important.
Discovering and analyzing correlations in health data
Anyone who thinks that healthcare institutions do not have their data in order is wrong. Hospitals, GPs, care institutions, etc. often have their patient data in order. Health Exchange, however, ensures that the data can 'talk' to each other, as it were. Teun: 'We don't do anything with the data ourselves, but we do ensure that it can be better accessed so that the various healthcare professionals can exchange it with each other.’ Ultimately, professionals must not only use all the data from all the different healthcare institution source systems to see how patients are doing, but also what the data tells them. What can they learn from the data? Can certain connections be made? And can certain things be predicted?
An example makes this more concrete. In health care, modern sensors are increasingly used to remotely measure certain values and behaviors in patients. Suppose a certain patient is a frequent sleepwalker and starts wandering the corridors of a healthcare facility at night. Healthcare providers receive a signal about this: a purely factual observation that prompts the nurse to go to the patient and escort him safely back to his room. ‘A noble thing, but so much more is possible,' says Teun. ‘It is probably much more interesting to link wandering to the patient's medication details. Has anything changed recently? Or did the patient have an argument with someone earlier in the day, for example? And can we link this incident to similar cases? Were there more patients who showed this behavior recently? And what about their medication? There might be a correlation.’
Teun therefore advocates doing much more with health data. In addition to medical assistance, this data can be used for education, research, personnel management and so on. In other words, data-driven knowledge can improve healthcare, and KPN Health Exchange plays an important role in this.
The technology behind Health Exchange: building blocks and Kubernetes
Health Exchange makes use of cloud-native technologies and containers; the techniques Docker and Kubernetes immediately present themselves.
Docker is a suite of software development tools for creating, sharing, and running individual containers; Kubernetes is a system for controlling containerized applications at scale. Containers are like standardized containers for microservices with all the necessary application code and dependencies inside. The creation of these containers is the domain of Docker. A container can run anywhere, on a laptop, in the cloud, on local servers, and even on edge devices.
A modern application consists of many containers. Managing these in production is the task of Kubernetes. With Kubernetes it is possible to distribute software to virtual machines in an automated way, according to the principle of orchestration. The system is scalable so that a system manager can easily place more or fewer objects in a group. Many cloud platforms offer a Kubernetes platform or infrastructure as a service (PaaS or IaaS).
Since containers are easy to replicate, applications can work auto-scale: expanding or contracting processing capabilities according to user demand. Docker and Kubernetes are mostly complementary technologies.
At the front end, the Health Exchange system uses the recognition, classification, transformation, and transport of various FHIR resources defined via HL7 FHIR. The abbreviation HL7 stands for Health Level Seven: the global standard for secure, electronic information exchange in healthcare. The HL7 standard defines all types of data in all care domains and care sectors, is developed and managed by the international HL7 organization, and is active in more than 30 countries. The FHIR resources are related to the healthcare information building blocks managed by Nictiz. Examples of health information building blocks are blood pressure, history, patient, general practitioner, etc.
When medical data can be used quickly and securely via the cloud in accordance with the HL7 standard, a host of attractive applications will arise. Not least for the patient himself. Besides more insight into their own data, patients can always call up their medical data via an app, for example during a holiday where they unexpectedly must visit the local GP. ‘But soon you will also be able to adjust data in case of errors that have crept into your medical data,' says Teun. ‘Suppose it says that you have COPD, while you suffer from asthma. In our future scenario, we want to move towards a situation where patients can also tune things themselves, or at least mark any desired adjustments. This is a future function of DVZA but will always be with the intervention of the healthcare provider.'
The near future of health: streaming data and machine learning
Teun thinks the future looks very promising, in which the medical field will make use of so-called streaming data. ‘We are increasingly moving towards the use of models that can learn from a lot of data that is constantly available. This is what we call machine learning.’
Machine learning is a part of artificial intelligence in which a model discovers patterns in data to determine the relationship between data input and output. Based on this relationship, the model examines the most meaningful data input - known as features - and generates a score that predicts, classifies, or recommends future outcomes. Machine learning scores can also be derived from anomalies; breakthroughs in patterns that affect future outcomes.
In America, for example, real-time sepsis alerts are already being used as a key to save lives. Early identification of sepsis significantly increases the ability to administer antibiotics within the first critical 'golden hour'.
Streaming machine learning is the application of an ML model to a streaming data pipeline, or a workflow that pulls in and transforms data in real-time between a source and a target. Real-time can mean milliseconds, seconds, or minutes, depending on the use case. The machine learning model provides logic that helps the streaming data pipeline to uncover features within the stream and potentially within historical data. The model then generates real-time scores based on the features the pipeline has uncovered. The data pipeline delivers these real-time scores to business monitoring systems, business intelligence tools, medical applications, or workflows, so that these scores can make predictions, classifications, or recommendations.
Teun: 'Why would you measure someone's heart rate every five seconds and store that measurement? Whitin no time you'll end up with a gigantic number of terabytes of data that won't necessarily be of any use. If a heart rate drops below a certain critical value, you want a signal and you need to store that data. Thanks to machine learning and real-time streaming, you only store data when it is relevant.’