Death by a Thousand Clicks Part 2 - Errors in the EHR

This is our second part of a four part series looking at the Fortune Magazine article “Death by a Thousand Clicks.”

Part 1

Part 3

Part 4

Many hoped that EHRs were going to completely remove medical errors from patient charts. While it is safe to assume that errors have significantly decreased from the days of reading physician handwriting in charts, errors are still very prevalent in EHRs today.

These errors can come about from a few different sources, some from human-error and others from the EHR software.

Nurses and providers will always be susceptible to entering incorrect information, whether this is simply transposing numbers or charting on the wrong patient. It is the EHR’s role to catch these mistakes and prevent them from becoming a bigger issue.

Even when the correct information is placed in the correct patient’s record, the configuration of the EHR may prevent the correct action from being taken. If a provider does not take the correct sequence of actions, it could result in orders not flowing downstream. This can significantly delay patient care and cause irreparable harm.

These configuration issues can cause lengthy legal battles to determine who is at fault, the hospital or the EHR vendor. It is the responsibility of the EHR vendor to not design a system that allows harm to come to patients. All hospitals are unique, however, so the design also needs to be flexible enough to fit everyone’s needs.

Extract wants to minimize errors in the patient’s chart as much as possible. Extract’s HealthyData platform minimizes the number of clicks and amount of manual information that needs to be entered.

After reading the document with OCR technology, Extract’s data capture engine will automatically classify the document and pull out all of the pertinent information. The document will be mapped to a specific patient’s order or encounter so all of the data will get into the correct place in the correct chart. Based off of the document’s classification, only the fields that are needed for that document type will be available to the user.

All of that will require the user to only verify that the information was extracted correctly instead of manually entering in each field, which can be a very error-prone process.

While the automatic extraction of data is not perfect, there are many safeguards in place to prevent incorrect information from being submitted. The text on the document will be highlighted a different color depending upon how confident the software is in the OCR data. There are guardrails on results to not allow users to submit them if the data does not line up. This can include requiring all necessary tests to be present on an order, result values to be within a reference range, and result flags to be present when needed.

Learning as users are verifying documents is also a vital piece for Extract to minimize errors. Machine learning can be implemented to automatically improve the data capture engine. Extract will also make manual changes to the automation process to accommodate changes to documents. To continuously monitor how well the automation is doing, Extract offers an analytics dashboard that can track how accurate the data capture engine is along with how accurate the users are through a QA process.

While EHRs are not perfect at preventing errors, Extract’s HealthyData platform can help to minimize the amount of incorrect information going into patient charts.


About the Author: Sean McGee

Sean is a Customer Support Specialist & Project Manager at Extract. Prior to Extract, Sean spent 7 years at Epic learning about Healthcare IT and everything that goes into setting up and supporting EMRs. He has a Bachelor of Science in Mathematics and Economics from the University of Notre Dame.