The increasing digitisation of healthcare and medical research, from telemedicine to electronic health records, is creating exciting opportunities to use data to drive new efficiencies.
Technologies such as artificial intelligence (AI), digital twin and graph data science are generating insights that enable better prevention, more accurate diagnoses, more effective patient treatments and solve complicated healthcare and life sciences problems.
The massive volume of data that is generated and used in today’s digital economy requires technology that can effectively solve modern data challenges. Graph technology, which works differently from traditional databases, can compare multiple datasets and contexts since the data is stored as nodes and links, which helps structure and identify the relationships between entities.
Improving the patient’s journey
The healthcare sector is inundated with large volumes of unstructured data which can be overwhelming for many healthcare providers and difficult to analyse and make sense of.
In one case, a large US health insurance company wanted to use patient data to improve health and outcomes. With 3.5 million members, it had amassed a huge amount of data, including claims, explanations of diagnoses and procedure codes, and it saw opportunities to generate insights from this. By looking at people who managed chronic conditions well, and how they did this, they could share these insights with other members.
For example, what should be the next best action for a particular member based on where they are in their clinical journey? If they do A, B and C, what should D be? To identify this, the company needed to connect all the elements of a patient’s journey. Using natural language processing (NLP) to gather health information from providers’ notes, test results and more, they created a graph with 1.4 billion nodes and nearly 3 billion relationships.
It’s now possible to explore patient journeys and extrapolate insights into what a more successful journey looks like.
Helping with chronic pain
With nearly one in five people suffering from the debilitating effects of chronic pain, e-health company Dooloo wanted to create a platform that would give recommendations on managing chronic pain based on a patient’s self-reported health and behaviour. Such a platform would need to synthesise a vast amount of data, including all of a patient’s medical records from different providers and their prescription history, as well as information on the latest treatments and their efficacy.
After trying to arrange data in a relational database, with two-dimensional tables, the company realised it was simply too complex. They switched to building a graph to easily store the relationships between different data points. The platform can now guide patients to the most effective educational modules and coping strategies given their unique history and set of circumstances.
Dooloo is also planning to add a layer of AI to enable predictive and prescriptive analytics, such as detecting similarities between patients and offering refined personalised recommendations that promise even more impact.
Analysing complex cancer research data
Graph technology is also being used for cancer research. One research institute, the Candiolo Cancer Institute, wanted a way to track the data of cancer samples, such as biological and molecular properties, and the procedures performed on them. The aim was to analyse this data and generate high-level biological hypotheses.
Trying to use a relational database with MySQL resulted in very sluggish queries and problems with data integration and coherence. Instead, by building a graph database, researchers could capture data more accurately and continue to import data from publicly available resources.
The graph database is much more flexible, allowing it to evolve and accommodate continually changing biological research and its findings. The team can also easily share data with other researchers across the world as they try to identify more effective cancer treatments.
Graph algorithms are specifically designed to query the topology of highly connected data. Through finding common ground, uncovering influential components, and inferring patterns, predictive elements can be converted into machine learning methods. This increases the model’s accuracy and allows for better predictions.
According to Gartner, smarter and ethically responsible AI and machine learning will deliver higher business impact. Gartner predicts that graph technologies will be used in 80% of data and analytics innovations by 2025, facilitating rapid decision-making across organisations.
When dealing with copious amounts of data in the healthcare industry, graph technology is an ideal option. Using knowledge graphs, data lineage, which makes it easy to see how data has changed, where it is used, and who is accessing it, ensures reliability in a sector such as healthcare that depends upon confidential and highly sensitive data.
Networked data also quickly identifies so-called data biases within existing data, which means AI models trained on this data will be fairer and less susceptible to discrimination. The comprehensive transparency resulting from context-driven AI can improve the overall trustworthiness of AI and robotics in healthcare.
For the healthcare sector, this means improved efficiencies, better patient care, more effective treatments and ultimately saving more lives.
Peter Philipp is Australia and NZ general manager at Neo4j.
