How Does AI Improve EHRsDec 05, 2019 | Jonathan Maisel
Artificial intelligence (AI) is quickly becoming one of the fastest-growing facets of today’s technology. One survey of experts in machine learning and AI collected the predictions of Yale and Oxford University researchers on how this new technology would affect every industry including healthcare. These experts concluded that in less than 120 years, machines might be able to automate all jobs performed by humans.
For now, physicians are completely essential in caring for our population, yet they are increasingly required to spend more of their time on documentation. Measures like the Merit-based Incentive Payment System (MIPS) require physicians to use Electronic Health Records (EHRs) and populate them with detailed notes. These requirements result in more complete and accurate patient notes, but they can be a drain on doctors who want to spend their time actively treating patients.
Is AI the answer to making marked improvements in patient care while reducing the burden of documentation on physicians? Understanding the impact of AI on healthcare and EHRs and Electronic Medical Records (EMRs) can help you make a more informed decision when considering new technologies for your practice.
AI Is Rapidly Facilitating Healthcare Industry Changes
Artificial intelligence has already begun touching our lives in multiple ways. For instance, when you speak to the smart assistant on your smartphone or home hub, you’re engaging with AI. However, in non-consumer industries like healthcare, AI is being developed to do much of what humans currently do but with better efficiency and a lower cost to everyone involved. Artificial intelligence and healthcare go hand in hand, and the face and function of healthcare is now being rapidly transformed in the following five ways:
1. Staying Healthy
The growing popularity of technology like smartwatches has opened up a whole new range of possibilities of using AI in healthcare. Many smartwatches already have impressive health-relevant capabilities such as heart rate monitoring. If your smart device is watching your heart rate all the time, it can also be equipped to let you know if you have a serious condition like an arrhythmia.
The connected Internet of Things (IOT) encourages people to take their health into their own hands. When people can check their watch or smartphone to look at some of their health indicators, they often feel more motivated to manage their lifestyle themselves. For example, AI is beginning to have capabilities like estimating and predicting blood sugar to help diabetics make better choices from moment to moment.
2. Early Disease Detection
The earlier a disease is identified, the more effective treatment can be. AI has become significantly more accurate in diagnosing diseases in the past decade, to the point that its accuracy is sometimes nearly equivalent to the accuracy of human healthcare professionals.
AI can also be of particular benefit for diseases with high rates of misdiagnosis. For example, mammograms frequently produce false-positives for breast cancer. As such, multiple research teams have begun inquiries into the efficacy of AI for improving the accuracy of mammogram readings, and the studies that have been conducted so far have largely found that AI is surprisingly at least equally effective at diagnosing breast cancer as physicians are. AI has also been shown to be useful in predicting the risk of cancer, allowing patients to effectively monitor their health more closely.
3. End-of-Life Care
As healthcare technology improves, people around the world are expected to live longer than ever before. That means we will struggle more with long-term conditions like osteoporosis, heart failure and dementia. AI is helping to make end-of-life care more beneficial to recipients by identifying pain and uncomfortable symptoms that someone at the end of their life may not be able to communicate effectively.
4. Medical and Pharmacological Research
Currently, a new drug takes an average of 10 years to get to market from research in the lab to reaching the patient. Only very few of the thousands of drugs that start pre-clinical trials actually move on to human tests, and after that, just a handful are ever approved for sale to the public. The cost of developing a prescription drug is more than $2.5 billion, but the implementation of AI research may be able to reduce that by streamlining testing.
5. Physician Training
One of the most tangible results of advances in AI is in medical training of all varieties. Standard physician training currently consists of book and classroom learning combined with observation and possibly practice facilitated with a software program. AI implementation in these software programs can be much more responsive to the minute movements a surgeon makes, for example, providing learners with more life-like and real-world experiences before they even touch an actual person.
How AI Is Improving Medical Documentation
One sub-field of medicine AI has great promise in improving is the medical documentation industry. Today’s EHRs have driven great improvements in the overall quality of patients’ medical records, but physicians often feel the manual data entry tasks involved in using them is too burdensome given their busy schedules and multitude of responsibilities. As EHR AI capabilities are becoming more robust, an increasing amount of EHR vendors are using it to improve their services for clinicians. Here are just four ways that using AI can improve electronic health records and care.
1. Data Extraction
Faxes are still alive and well in the healthcare industry, and gleaning information from them remains largely a human task. However, many EHRs offer the ability to extract data from faxes and other text-based forms of documentation with the help of AI systems. These AI technologies still rely on things like key words and phrases as the foundation of the extraction, but developments in deep learning are helping machines process the data similarly to physicians.
2. Diagnostic Algorithms
We’ve mentioned how AI can help make diagnoses more accurate, especially in cases of frequently misdiagnosed diseases. But in combination with reviewing the documentation found in an EHR, AI can now also uncover disease or significant risk factors in patients. This can be especially invaluable when patients are moving from one doctor to another, as an AI system can monitor a patient’s whole history for a pattern, whereas a doctor simply cannot do this for every single patient since it would be too overwhelming to do so.
3. Decision Support
The combination of EHR data and AI analysis creates the foundation for more personalized care. Every doctor knows that multiple individual factors can impact care outcomes for patients, and AI assistance can provide actionable insights from patient data to help support clinicians in making the best decisions for each patient.
4. Documentation and Data Entry
When EHRs were first introduced, clinicians had few options. They could type their notes directly into an EHR during the course of an appointment or opt to take handwritten notes. If they typed the notes into the EHR themselves, they’d often spend more time typing than they would on interacting with the patient. If they took handwritten notes the traditional way, either a physician or a staff member would then have to insert the notes into the EHR system.
New AI capabilities in speech recognition and language processing have resulted in new technology that allows doctors to record sessions and create documentation without the need to create and edit the notes on their own. Speech recognition technology was the first iteration of this convenient note-taking format by allowing physicians to speak their notes and have them translated into text by software.
More advanced documentation methods like the one employed by ZyDoc use speech recognition in combination with the editing and formatting power of a human transcriptionist for the quickest and most accurate documentation possible.
What Are the Limitations of AI and EHRs?
While artificial intelligence has the potential to revolutionize how physicians interact with their EHR systems, there are still obstacles that unfortunately put a finite limit on what AI can do for clinicians.
1. Interoperability is Limited
The ongoing regulatory push for the adoption of EHRs has always included interoperability between systems as a central focus. The Office of the National Coordinator for Health IT (ONC) has even created a 10-year interoperability roadmap detailing the expectations for EHR systems in the coming years.
One of AI’s limits is the need for huge amounts of data that can’t necessarily be fulfilled by the developer of the individual EHR. Because interoperability hasn’t been fully implemented by every healthcare provider and every EHR vendor, no proprietary AI technology has access to the full breadth of patient information it could use to make predictions more accurate when it comes to diagnostics.
2. Physician Input is Still Necessary
We’re still a long way from having technology completely displace physicians, and AI-based tech still must rely on the work done by them. For instance, a physician still needs to feed correct health information to an AI system. If a physician puts bad data into an EHR, the associated AI will be learning bad or even potentially biased information which it will then use to make more, potentially incorrect predictions. AI requires close supervision to do its job properly. Without the right input, an AI’s output simply cannot be fully trusted to be accurate and effective for the patient.
3. Limited Accuracy
AI used for technologies like speech recognition is still in developmental phases. A study published in the Journal of the American Medical Association examined the error rates of documentation produced via speech recognition software, and the results were disturbing.
The researchers found an error rate of 7.4% across EHR documentation produced with speech recognition technology, indicating that there is a long way to go in perfecting the use of AI in this application. Of course, the range of consequences that accompany errors in medical documentation is wide. A small mistake in a patient’s chart might never be caught, and it could result in a treatment choice leading to a patient’s death in some instances. As such, an error rate this high should make physicians wary of using speech recognition software exclusively.
Interestingly, when a transcriptionist reviewed documentation generated by speech recognition software, the error rate dropped all the way down to 0.4%. This makes it even clearer that a combination of AI and human oversight is still the most effective way to take the burden of documentation off of physicians.
How Can AI Eventually Make EHR Integration More Accurate?
Artificial intelligence ultimately has the potential to streamline medicine at almost every step in the diagnostic and treatment process. Advances in technology have given AI the power to integrate with EHRs and promote interoperability between them more effectively than ever before. These frontiers represent challenges that AI can help physicians address when it comes to EHR integration.
- Permissions across systems:It is currently difficult to facilitate approval for moving patient data from one EHR system to another, and the process often requires multiple approvals that drain people’s time. AI may one day be able to streamline how information is released and shared with the appropriate parties who use different EHRs.
- Interoperability across systems:Even when all the appropriate permissions are obtained, not all EHRs use the same formats for data and this can lead to interoperability issues. AI applications may soon be able to translate between different formats or work on a universal template to ensure all documents are readable between systems.
- Provider competition:Because it’s so difficult to share information between providers, some try to keep a lid on their patients’ data and prevent it from being shared due to security concerns. More sophisticated AI could potentially remove this barrier by automating permissions based on information gathered from the patient’s EHR.
The consumer side of AI may have benefits for EHR integration as well. With patients being able to take and record information like their vitals, blood sugar or other medical indicators, there is a new wellspring of information that can be mined for use in achieving better patient outcomes. A patient who can monitor their own heart rate can also send that information directly to their EHR, where AI technology may be able to draw critical conclusions about current or potential health conditions.
EHRs are a technology that has been highly beneficial in enhancing the quality of patient records, but clinicians have always struggled with the time-consuming documentation needs required. With AI bolstering the accuracy of EHR documentation integration, physicians can do more of what they’re trained to do and less of the tedious administrative work associated with taking notes.
ZyDoc: A Cutting-Edge Medical Documentation Company
If you’re looking for a service to reduce the documentation burden on physicians in your practice, ZyDoc is here to help. We know you want to provide the best care for your patients, and it’s likely that manual note-taking or speech recognition alone is not conducive to providing the highest care efficiency. ZyDoc’s medical transcription services use the proven combination of AI-powered speech recognition technology and human expertise to provide documentation that’s 38% more accurate than speech recognition alone. Additionally, we’re 61% faster than manual typing and clicking, and complete 90% of dictation jobs within a two-hour time frame.
Our transcription services with EHR integration take even more of the work off of physicians’ plates by automatically inserting clinical data into the right places in your EHR, no matter who your vendor is. Your data will be ready for review by other treating physicians and any other approved healthcare providers quickly and with the utmost accuracy.
If you’re interested in not only saving time but improving the accuracy of your practice’s clinical documentation to get better reimbursements, call ZyDoc at 1-800-546-5633 for a 14-day free trial. You can also browse our plans and pricing at your own pace to learn more about how transcription with EHR integration can streamline your workflow.