Structuring the Unstructured: How Rush University Medical Center is Harnessing the Cloud

While Chicago’s Rush University Medical Center was chartered in 1837, it’s looking to the future to move the medical care forward. Rush is in the midst of a project to move its medical records to the Google Cloud Platform (GCP) and also to transform their unstructured clinician notes into searchable and analyzable data. With more than 8 million charted notes per year, it’s not an easy or quick task. But once complete, and with new notes added to the cloud in real time, Rush will be in a position to better serve its patients, and the local and research community, with access to analyzable and pertinent data they previously couldn’t see. 

Dredging the Data Lakes of Unstructured Data

Healthcare systems like Rush have a wealth of information, perhaps too much to weed through, if it’s not in a usable form. The problem is that the data is logged and kept in multiple places. Unstructured data and structured information in modules (e.g. pathology reports and lab reports) are siloed in the system, making it difficult to integrate. About 80%1 of data in electronic medical records (EMR) is unstructured, such as clinical notes. It’s typically difficult to draw insights from unstructured data, however with new machine learning capabilities, this data will soon be accessible in certain use cases. 

Maven Wave partnered with Rush to move 10 years of unstructured data into GCP and draw out insights using machine learning and the Google Healthcare API. Without the data uploaded to the cloud, and the unstructured data lacking uniform phrasing, Rush has a limited ability to create analytical insights. There is no reliable way to track procedures or link them with related pathology results. EMR notes are disjointed, with clinically relevant information distributed in various notes, between physician notes, MyChart notes, lab reports, and telephone encounters.

Rush’s goal is to increase collaboration, both internal and external. They want an intuitive user experience for note searching and analysis. They’re also looking to create a searchable repository for all of their processed structured and unstructured data, with one source of truth for the records. And they’ve already begun moving their 8+ million notes per year to the GCP.

By the time 10 years of data is in the cloud, 1.5 years will have passed, and their terabyte of data will equal 10% of what’s in the Library of Congress. The analysis then begins. Rush plans to start by applying analytics in these three use cases:

Adenoma detection rate (ADR): ADR is the proportion of patients age 50+ who undergo a screening colonoscopy for the first time, with at least one adenomatous polyp detected. If small polyps are found, presumably the physician would identify large ones if they exist. To meet the CMS quality metric, a clinician must meet the threshold percentage, as lower percentages can mean that the clinician is potentially missing some incidences of cancer2. Manual analysis of these notes will take an estimated 1,200 hours for each year, while computer analysis is understandably much faster.

History and physical completeness: 10 years of unstructured H&P clinician notes will be not only uploaded, but validated by a staff member for completeness, with clinicians signing off on them. All notes will be searchable, and available for Joint Commission audits. 

Clinical documentation improvement: By using machine learning, Rush will be able to see trends and patterns and notice when doctor notes may be incomplete. This could prompt doctors to add important detail or point out places where information may have been missed.

This is just the beginning, though. Data and clinical uses are endless, once patient data is searchable. Access to data unlocks the potential to address multiple other business use cases, such as:

Population health: With better analytics tools, clinicians can see trends and patterns that can improve healthcare overall for a group of people.

Personalized health: Choosing the best treatment for a patient is increasingly dependent on genetic factors. Understanding both a patient’s individual genetics and how the treatment impacts specific cell mutations or disease characteristics increases the likelihood of a better treatment choice and outcome.

Patient care: Even for treatment that doesn’t rely on personal genetics, analytics can greatly aid in better care. Analyzing aggregated data can help clinicians identify conditions more quickly, leading to faster and more accurate care.

Improved coding: Coding errors cost healthcare systems money, in overpayments or underpayments. A report showed that over a three year period, CMS overpaid Medicare contractors in 13 jurisdictions by $35.8 million3, and went back after that money, recovering at least 63% of it. Analytics can help improve coding practices and catch over and under billing.

Improved operations: Analyzing big data can provide better insights into expenses and resource utilization. The results can identify patients who may not show up for visits or who may have payment problems, allowing the healthcare system to proactively follow up with them to address issues before they happen. Analyzing wait times and ER or imaging facility usage can lead to better scheduling, improving the patient experience, while also maximizing the healthcare system’s revenue-producing opportunities.

The Digital Data Warehouse

As data capturing and analysis techniques mature, the data is becoming easier to analyze and apply. Rush has an ambitious goal: develop a robust analytics platform that can scale and grow with all newly acquired data. They want to do this with a serverless platform to incorporate storage and computing. Data can be collected from multiple places, including wearable devices, bedside devices in or out of the healthcare facilities, genetic testing organizations, Extract, Transform and Load (ETL) tools, databases, and data warehouses. These will be on a platform as a service (PaaS) model, which has the power to provide data as a service, enabling its use in a timely manner, while maintaining security and accuracy. Maven Wave is helping Rush implement this solution using GCP and various Google tools for data ingestion, data processing, data analytics, and machine learning. 

Doctors often rely on clinical narratives to understand patient care, but these narratives are difficult to process and understand in a larger context if they can’t be standardized, searched and analyzed. Rush’s goal is to continue providing patients with the best care, and unlocking more information to do that. They will have a single source of truth with their records, giving them the potential to better help individual patients and patient populations.

Rush is leading the way with this type of solution, but by no means does it have to be an unattainable dream for other medical organizations. Contact Maven Wave to find out how we can enable you use your data to best help your patients and medical center.


About the Author

Kylie McKee
Kylie McKee is a Content Marketing Strategist at Maven Wave with more than eight years of tech industry experience and five years of content marketing experience. Prior to joining the Maven Wave team, Kylie worked as a Content Marketing Specialist for WebPT, Inc. and earned an Associate in Applied Science in Motion Picture, Television, and New Media Production with a CCL in Screenwriting from Scottsdale Community College.
August 1st, 2019

Get the latest industry news and insights delivered straight to your inbox.

Sign up for our Newsletter