My team recently had a brownbag on the types of healthcare data available and I took the opportunity to share a bit about electronic health records. Other types of data shared included the MIMIC dataset and imaging data (e.g., X-rays, CTs, MRIs). I received feedback that it was a useful “EHR 101” and thought to share it here too.
Healthcare has a problem
In most places around the world, primary care and hospitals maintain their own, distinct systems for electronic medical record (EMR) data. As a result, patient and medical data across different providers are incompatible with each other, leading to a lack of interoperability.
Providers want to control all digital records of their patients, ensuring patient retention. This leads to data being siloed at each institution. Patients’ prescriptions, lab tests, diagnosis, etc. are not visible across institutions, contributing to significant wastage.
The other problem is that of poor usability. Often, these systems don’t account for human computer interaction principles. Thus, clinicians often spend more time talking to their laptop than to the patient, contributing to clinician burnout. Furthermore, while the system works and data is dumped in, it is often in such a mess that it is impossible to use.
Enter the electronic health record (EHR).
What are EHRs?
Simply put, EHRs are a digital version of a patient’s record. This includes medical history, diagnoses, medications, treatments, lab results, etc.
Digitization of medical records should be a good thing right? Theoretically, it should help to automate and streamline a clinician’s workflow, as well as solve the issues of interoperability and usability.
Wait a minute, time out… what’s the difference between EMRs and EHRs?
- EMRs are a patient’s record by a single provider, which may or may not be compatible with systems from other providers.
- EHRs are a complete record of a patient’s health (i.e., a longitudinal record across all providers)
Unfortunately, EHRs have a bad rep too. While it may have solved the problem of interoperability, they are still notoriously difficult to use. Here’s a short 4 minute “documentary” on EHRs 🤣
While the usability of EHRs by clinicians is a big problem, this post will sidestep it and focus instead of the potential of EHR data in applying data science to improve healthcare.
An interface for EHRs: FHIR
Fast Healthcare Interoperability Resources is the current standard for healthcare data exchange. It is abbreviated as FHIR and pronounced “fire”. It provides modular components called “Resources” that developers can assemble into working systems for clinical and administrative problems. From its website, its strengths include:
- Focus on implementation, with multiple libraries to kick-start development
- Foundation for web standards with RESTful architectures
- Interoperability out-of-the-box
FHIR allows you to combine the component resources into an interconnected application that tracks and records the patient journey. Here’s an example of how FHIR would work (credits to fhirblog.com).
One possible FHIR framework for a primary care provider
Next, given the scenario of a 5-year-old boy visiting a primary care clinician for a consultation (ear pain), as well as a follow up visit a couple days later…
We can then overlay the FHIR resources on it like so:
The end result
We have beautiful data, with edges indicating the relationship between each component. Yes, it’s probably a lot of work getting the notes labelled into those components neatly. However, I’m optimistic that this could be automated in the future.
For example, advances in natural language understanding and prediction could help with labelling these notes. By adding speech-to-text capabilities (which are easily available as commodity APIs now), we could be seeing a future where doctors don’t have to take notes. (I’ll skip the cliche on doctor’s handwriting.)
Putting health in your hands: PHRs
Personal health records (PHRs) are an electronic application through which individuals can access, manage, and share their health information, possibly solving the issues of usability from the patient angle. They are becoming increasingly practical with the increased efficiency in compute and the widespread use of the internet.
Notwithstanding, there are some challenges to adoption.
Firstly, how can one easily get the data to patients? Many patients, especially the elderly, may not be technologically savvy and have trouble navigating PHRs.
Next, even if the data can be accessed via patient devices, how should we engage and educate patients to use the data to gain better insight and improve their health? Behavioural change based on PHR insights/nudges, or anything for that matter, is difficult.
If we manage to get patients enthusiastic about their own PHRs, how would we make it usable? Existing EHRs are known for being difficult for trained medical professionals to use—how do we design simple processes and interfaces for the lay person?
From the patient perspective, they want access to their data, such as medical history, medications, appointments, past lab tests, as well as the ability to chat with their doctors.
Across multiple providers.
From a single interface.
One possible solution is that of HealthVault, a web-based PHR. It allows users to gather, store, use, and share their health information online, helping them to create a more complete picture of their health.
Here’s a sample walkthrough of how to use HealthVault:
- Create an account on HealthVault and log in. You should see the screen like that above.
- Download the sample Continuity of Care Document (CCD) that is in xml format here. You’ll notice that it’s in XML format and difficult to read.
- Click on the “Add” button at the top menu bar, select “Continuity of Care Document (CCD)”, and upload the sample CCD.
- After adding the sample, you can browse the CCD uploaded, including patient information, allergies, medications, encounters, etc.
Seems rather simple right?
One small caveat: From a cursory search online, it doesn’t seem directly compatible with FHIR. Nonetheless, there are efforts to link HealthVault with FHIR, such as here.
By aggregating patients’ medical history, medications, treatments, and appointments in a single place, it should make it easier to access. With this, the intent is to encourage patient involvement in their own health, hopefully enabling them to manage their health ad (chronic) conditions better.
Research and applications on EHRs
There have been many papers published on using EHRs for data science, of which I’ll highlight just a few:
- Risk Prediction with EHRs: A Deep Learning Approach
- Deep Learning to Predict Patient Future Diseases from EHRs
- Scalable and accurate deep learning with EHRs
Hmm. Odd that the paper titles all have “deep learning” in them—must be some kind of fad or hype.
Anyhow, that’s my short primer on EHR data, with examples you can further explore (e..g., FHIR, HealthVault) and papers to read up on. Hope you found this useful!