By Janae Sharp, May 23, 2019
Electronic Health records do not always give the data to physicians at the point of care that would help them give better patient care and save money.
Translation Errors Impede Understanding
The problems in interoperability are akin to code on the backend of GoogleTranslate: The system has incomplete logic, due to the fact that languages have idiomatic expressions and thus words sometimes don’t translate directly. For instance, in the German language, nouns have genders. This creates a challenge for translation systems, particularly in ones that go back and forth. The data will get translated to the appropriate gender into German–but then if the data (or in this case, the phrase you want translated) are passed back to English, the gendered nouns can create some strange translations.
With healthcare data, we don’t just need a system that is able to read the words–we need a system that knows when one of the languages has articles, or gender, and the other does not. We also need a system that knows context. In English we say, “great minds think alike.” In German they say “zwei dumme eine Gedanke,” or “two dummies, one thought.”
What do two idiots have to do with your electronic health records? Well, if you don’t know that the idiomatic expression in German is that “two idiots have one thought,” you don’t realize that saying “two minds” has a negative connotation.
Another instance: Wissenschaftler, in German, means scientist. Wissenschafttlerin means female scientist. Use Google translate to translate “female scientist” from English to German, and you input the English gender nonspecific word. On the German end, suddenly the female scientist is male. Context matters, and the right kind of system must decode these disparities to smooth over the translation. Interoperability is translation, and it has the same vulnerabilities that translating between two languages does.
One example of this is the difference between weight measurement. If someone has weight in lbs or kg the system will not always translate those straight across. This becomes an issue with pediatric medication. If the system does not have specific units of measure for weight, the dosage for a 15 kg infant is different than the dosage for a 15lb infant. This can be further confused when mL and L are not specified.
Like google translate, the template behind the meaning is assumed to be the same. When those meanings aren’t the same, such as pediatric dosage, interoperability without proper context can be dangerous.
Bridging the Digital Divide
I spoke with Dirk Stanley, MD, CMIO of UCONN, about the clinical and digital divide. “You can have the most unbelievable solutions in the world that are not supported,” he says. “There is a really big digital divide between the IT people and the clinical people. The best laid plans can often fail because, at the end of the day, if it’s not helping to treat pneumonia, or to make a better diagnosis, it’s far more difficult to argue for adoption.”
Holon Solutions was recently awarded a patent for their context-sensing technology, which surfaces actionable data to providers within an EHR, from a third party source–but without the need for external interfaces. More than just translation, Holon is targeting data relevant to the patient so that physicians can see it in “real time,” when the patient is sitting right in front of them. This is a significant step forward in giving physicians access to patient data at the point of care.
When meanings don’t match and physicians are given data that does not apply to patient care, they stop paying attention to that data. We also learn to ignore the things that don’t have meaning. Within an electronic health record, this can be learning to delete notifications on the least obtrusive end and lead to medical errors in some instances.
Sometimes the record isn’t actually speaking the same language as the doctors or patients. Saurabh Mathur of Holon Solutions spoke with me about how technology in healthcare is adding a burden on the physicians’ end. More notifications and more technology cause physicians to get “alert fatigue,” particularly when none of the notifications are relevant to their work. As a result, they stop looking at the notifications–at which point they might miss the few crucial insights to come across their screens.
Mathur mentioned that technology in general is adding abstractions and workarounds rather than solving the root problem of disparate meanings between technology-provided information and actionable data for physicians. Earlier in his career, Mathur worked on creating and managing electronic health records, and that work led to his mission to improve what data physicians could see when patients were with them.
“We were managing more than 80% of their orders. We dealt with a lot of different registries.
Over time as technologists we always took the approach of solving problems by throwing more technology at things with the hope that technology will simplify it. the layers and layers of abstraction in the middle are not helping”
-Saurabh Mathur
Workarounds in Data Interoperability
The problem of providing so many irrelevant “alerts” is that when you have too many alerts, everything gets ignored. You might not see the data that you need to see, or you get information about a patient after they have left the office. Population health programs aren’t always designed to deliver better patient data at the point of care and this delay creates worse health outcomes and duplicated care. For instance, a diabetic patient might appear at an appointment without their most recent glucose test levels, or they might self-report the information incorrectly. The time to realize that a patient with diabetes needs new testing, or is having difficulty maintaining their sugar levels, is when the patient is with the physician. In some instances, a physician will find out important information only after that patient has already left the office.
Moreover, not every health system has blood glucose tables on the back end that match types of tests or patients or relevant health conditions. For example when blood glucose levels aren’t appropriately monitored, patients have negative side effects such as hypoglycemia, hyperglycemia, diabetic ketoacidosis, or even organ damage over the long term. If information about potential risks doesn’t get back to physicians at the point of care, it is far more difficult for them to identify patients that should have more help and prevent long-term complications.
There are many examples of workarounds. Aaron Miri, MD, Chief Information Officer of Dell Medical School, came up with some online, such as purchasing commercial wifi so physicians wouldn’t be cut off from the network. Regulations and security don’t always align with the work physicians are doing.
But beyond providing applicable digital solutions, we need to codify the language so it has the same meaning for all of the people involved in healthcare. The information that the billing department needs is different than the information a physician needs at the point of care. It’s not just about the conversation–it’s not just the language; you need a common topic to talk about, but at the moment, there’s different jargon, depending on the target audience. In healthcare, physicians and electronic health records don’t have the same contextual language, and this causes problems.
We talk about interoperability as a data problem. The systems have to be able to speak the same language. We have mediator companies. Yet, we still encounter the same problems. Sometimes we’re missing relevant data, but sometimes we just interpret it incorrectly.
Holon Solutions’s now patented technology has solved for some of healthcare’s biggest data challenges: clinical workflow and contextual insights. They enable patient-specific data points to be presented to the physician at the point and time of care Healthcare Technology as an industry has a long way to go. And the issue couldn’t be more pressing; inability to access useful, actionable, and comprehensible data at the point of care diminishes both physician satisfaction and patient care, and patient health outcomes in the long run.
Original Article: Healthcare IT Today