EHRs… ahhh, love them, hate them. If you do any work in the medical space, you will need to interact with them.
In this post, I talk about what they are, what they do, their strengths and weaknesses.
What are they?
EHRs (electronic health records) are digital records of patient’s clinical records.
EHR platforms are software platforms that capture, store, and transmit EHR data.
It’s good to know the distinction as people can use these terms interchangeably sometimes…
EHR data is a major source of medical information. See this post on other sources of medical data.
Vendors of EHR platforms
There are many EHR software vendors (software companies that build EHR platforms), the biggest 2 are Epic and Cerner.
|Top 5 vendors for Hospital||Top 5 vendors for Doctor offices|
|Cerner 24%||Epic 27%|
|Epic 22%||Allscripts 10%|
|Meditech 21%||eClinicalWorks 7%|
|McKesson 10%||NextGen 5%|
|MedHost 8%||GE Healthcare 5%|
Source: 2015 data from ONC
If you wish to learn how EHRs work, Epic is a good starting point, because it has the largest footprint and longest history (that does bring legacy issues and lots of area of improvement though).
Note that EHR platforms are mind numbingly complex. You will likely be provided training to enable you to perform your job, so follow that lead and focus on what you need to know. After the basic introductions to the Epic GUI, I’ve found generic training of other functionalities to mean little, as I never had to use those functions daily.
What EHR platforms do
EHR platforms are widely adopted in hospitals, doctor offices in the US. Clinical/nonclinical staff depend critically on these platforms to do their jobs.
- Clinical functions enabled include recording patient’s problems, diagnoses, allergies, clinical notes etc.
- Non-clinical operations include appointment making, ordering diagnostic tests/scans, sending medical prescription orders etc.
They facilitate the operations at complex institutions like large academic hospitals, connecting different departments across various functions, e.g. emergency department, cardiology, surgery, laboratory, accounting and procurement etc.
EHRs do a LOT and are very robust in a lot of way.
- Enable complex workflow: EHRs are critical to the operations of medical facilities. These places are enormously complex and the EHR platforms do incredible jobs in enabling the operations.
- Rich(er) data: EHR data captures more information than do the usual insurance claims data. Typically, insurance claims data is only a small subset of the information recorded during a patient visit, e.g. primary + secondary diagnoses, procedures along with dates and fees, ignoring all the other important bits of information such as whether the patient had family history, allergies, vital measures etc.
- The clinical notes and imaging data are particularly fun for machine learning. Natural language processing and deep learning are prime tools to apply to these unstructured data types.
- Safe(r): HIPAA Privacy Rules have put in place various requirements that medical institutions (covered entities) have to abide by. These ensure a fairly high degree of security. All medical institution staff that interact with patient data must undergo annual security training to ensure continued awareness and prudent practice. EHR platforms are not water tight, but they multiple layers of safe measures usually work well in ensuring system security.
- Stability: EHR by design have been tried and tested over many years of use. Thus they are generally fairly stable software platforms.
- Customization possible: to meet the specific needs of different clients, EHR vendors typically allow some degree of customization.
All staff at medical facilities will have opinions on their EHR platform. Whether they love it, or hate it, you will see very strong opinions. For those not needing to use it daily, making sense of the backend data can also be very complex and confusing.
Below are a few typical issues you will encounter using EHRs:
- Ageing systems: A lot of these platforms were built decades ago, so they come with legacy clunkiness that will rub a daily user the wrong way for sure. A lot of the user interface have not kept up with modern GUI standards. Think Windows95 vs Windows10.
- Complexity: EHR platforms have to facilitate the complexities of operations at hospitals, so the EHR platforms themselves can be complex too. This means sometimes, to perform the simplest task, clinical staff have to click multiple times. To obtain simple information, one has to go to multiple locations. This frustrates users.
- Inconsistent accuracy: Frustrated users are more likely to just want to click enough to get the patient record completed. This degrades the quality and accuracy of the information being recorded.
- Silos: Silos of data exist within the same institution. It can be very difficult to obtain access such data, making research/analytics at scale, across time difficult or impossible.
- Interoperability (lack thereof): EHR platforms are competitors and try to protect their own market share and information architecture. Even though industry standards such as HL7 aim to bridge the gap between different EHR platforms, there remains significant difficulties when trying to compare EHR data across platforms. There are a few new companies that are attempting to bridge various health data sources, whether they can provide meaningful access to EHR data remains to be seen.
Other specific issues related to analytics
See this post on general issues with medical data. Below are some EHR specific ones.
- Gaps in data – this one is worth exploring a bit more: as EHR platforms are not usually connected (whether same vendor or different), doctors working in one institutions are unlikely to see clinical activities performed in other institutions. This partial view of the patient’s care has potentially catastrophic impact, e.g. not seeing contraindications to a drug detected in other institutions.
- Duplicating records: whether through error or reflection of the multiple doctors visited and each recording the same information.
- Outdated info: staff may default to older values during subsequent visits or not be updating information timely.
- Non-standard mappings: some EHRs do not use standards medical coding schema, such as LOINC or NDC. This makes analytics difficult as you would have to map clinical concepts each time, requiring tedious manual mapping.
Thanks for reading! Let me know if you would like to learn more about EHRs.
Please SUBSCRIBE to ensure you don’t miss future posts.