The role of the healthcare professional has always been to provide the best possible care, using all available information. Currently, this may include a patient’s medical record, the symptoms they present with, the information they tell the clinician, and the clinician’s medical knowledge.

As well as being expected to keep up with new research papers and clinical studies, clinicians are often required to absorb and update vast amounts of patient information outside of their working hours.  If we take the scenario of a primary care physician consulting a patient in a 15-minute appointment, or a physician working in the Emergency Room treating several patients an hour, we can start to see that this is an unrealistic expectation for doctors to make the best decisions with such limited information, and time. 

Electronic health records are becoming more comprehensive as providers become more connected, incorporating data from across the healthcare system. However, patients’ interactions with the healthcare system are not the only data points that we can collect. The volume and variety of data is rapidly increasing, which provides a wealth of opportunity, as well as some major challenges. New data types have the potential to make electronic health records far more robust and detailed. Clinicians are amazing data processors, but with the breadth of data available we need to understand and the scale of records we need to study to find these patterns, it would take a human years to do what a machine could do in seconds.

More information about an individual patient is helpful, but imagine if clinicians had the tools to process all available data on a population, and present them with possible diagnoses based on similar patients. The question for the data science community is how can machine learning and data analytics be applied, and how can we present this data to the people who need it the most, in a way that is meaningful to them?

Machines are efficient at processing data and finding patterns, even the ones people can’t see. Firstly, speeding up the processing of data would free up time for doctors to spend with individual patients, to ensure they do what’s best for that particular patient, not just the average patient. Secondly, machines could find patterns in the data, presenting a clinician with a view of possible diagnoses, based on other patients’ data.

As transformational as this sounds, there are a few challenges we need to solve before we are able to use data science to bring real benefit to clinicians and patients. Data scientists are building models that provide the ability to identify patterns and predict outcomes – both of which will provide significant support for clinicians in their decision making. However, safely accessing healthcare data poses complex data problems. Orion Health data scientists are tackling these problems, in particular the de-identification of patient health information to ensure patient privacy. We’re also making progress towards solving how to deal with incomplete or missing data, and how to overcome bias in data. These are difficult, but crucial building blocks towards making precision health a reality.

Dr. Kevin Ross is presenting at the HIMSS conference in Orlando:
Machine Learning Over Our Growing Electronic Health Records

  • 2:30-3:30pm
  • Wednesday, Feb. 13 
  • Orange County Convention Centre W308A

If you’d like to know more about how we can help your organisation in the journey from population health to precision medicine, schedule a meeting with one of our subject matter experts, or visit our booth #4758.