Disappointing Move to Electronic Health Records

Disappointing Electronic Health Records

The idea of Electronic Health Records was to improve data flow within a hospital system for improved efficiency and care, and to share medical histories between doctors and hospitals anywhere in the country. Applied correctly, such technology would cut costs and errors and help docs make life-and-death decisions in the ER.

Unfortunately, the vision and federal incentives implemented under the Obama administration were poorly conceived and applied. Ten years and $36 billion later, Fortune Magazine and Kaiser Health News studied the issue and published a detailed article by Erika Fry and Fred Schulte. Today’s post is based on Death by a Thousand Clicks: Where Electronic Health Records Went Wrong with my notes from the embedded videos.

Broken Records: How We Got Here

With federal funding and a promising future, the American healthcare system began a rapid transition to Electronic Health Records, also called Electronic Medical Records. Obama’s High Tech Act provided incentives that encouraged quick investment and growth. But in hindsight, development and implementation was done too quickly. Rather than reach its potential, this digital revolution caused too many cases of patient pain and death.

Broken Records: Broken Promises

Years after their adoption, electronic health records still have not lived up to their potential. The video below (6:56 min) describes the vision and how badly it missed its mark.

To be clear, the EHR concept is not new and has been around for decades. As an IBM Systems Engineer serving large hospital accounts in the 1980’s, I installed a patient accounting system, patient care system, pharmacy system, and an electronic medical records system. They worked together to improve hospital efficiency and billing with primary focus on containing costs.

Hospitals back then didn’t share medical data with patients or other hospitals, at least not electronically. The vision of the Obama administration was to create an electronic ecosystem of patient data that would give patients anywhere access and speed medical research.

Broken Records: Patient Impact

The EHR potential seemed limitless, and today 96% of hospitals have adopted EHR systems, up from just 9% in 2008 before President Obama was elected. But the technology has fallen short. Medical practitioners complain about clumsy user interfaces that can lead to mistakes imputing data. A rush to market by more than 700 vendors with proprietary software and data formats leaves doctors resorting to fax and CD-ROM to transfer medical data. And patients struggle to gain access to their own records.

Instead of reducing costs, many EHR systems have increased them. Instead of streamlining medicine, the digital transition has increased safety risks and contributed to deaths and serious injuries, described in this next video (6:48 min).

Broken Records: Doctor Impact

Doctors complain that EHR systems cause them to spend more time in front of the computer, leaving less time for patients, and leading to an increase in physician burnout.

Technology Notes

The rush to install EHR systems and show “meaningful use” worsened the expected problems of any large implementation. Here are my quick notes from the videos.

UI design confused the medical staff and lead to GIGO (garbage in / garbage out). Modern voice recognition systems should be able to fix that and free docs to spend more quality time with patients.

Poor programming techniques and old “spaghetti code” made changing systems difficult and often created new problems. A focus on structured programming techniques with modern programming languages, and change control systems, should help.

Implementation errors often left new EHR systems without properly tested interfacing with other hospital systems.

The lack of Interoperability between EHR systems, often intentional, made the sharing of patient data difficult. Each system had its own proprietary file format, and data fields did not line up between one system and others.

Searching for relevancy was difficult because of the mountain of accumulated data, some of it, like images, in non-encoded formats. Expert AI systems like IBM Watson can be of great help when the data is formatted into computer-readable fields, and the company is extending its abilities to include images.

SYNTRAN (SYNtactic TRANslation) – Automated systems like one I studied in the 1970s should be able to help medical staff find relevancy in large medical records by automatically creating abstracts in response to natural language inquires. SYNTRAN, which was written decades ago to run on an IBM 1401 with just 8KB of memory. With its built-in understanding of the English language, the program was able to convert entire books and manuals into abstracts of varying size (10%, 5%, or 1%). It did such a good job that you’d swear a human wrote the abstract.

Given today’s smartphone compute power, I’m surprised doctors aren’t already using a system like that to find the most relevant parts of a medical record, presented with whatever detail they want (in just one paragraph, or as a page or two), with the ability to dive deeper if needed. It’s certainly possible and has been for over 60 years.

SYNTRAN’s linguistic intelligence was programmed in, rather than learned. That makes me wonder about the potential of similar systems developed using recent advances in neural networks, artificial intelligence, and generative AI, and other forms of machine learning.

Imagine a simple app that records audio of each patient-doctor visit and automatically transcribes the spoken words into text for the EHR. Too much content, you say? That’s where abstracting software like SYNTRAN could help.

Business Model Notes

The incentive for tech developers was to be quick to market, resulting in poor designs and a Tower of Babel of sorts, with too many competing systems that didn’t talk with each other. The incentive for hospitals was similar, and they also saw each other as competitors. Hospital A, for example, had no incentive to share patient data with Hospital B, and they worried about losing the patient to a competitor. Likewise, Hospital B had no real incentive to receive patient data from Hospital A if they didn’t trust its accuracy or could profit from retesting themselves.

Best Practices Notes

The US Veterans Administration has long been a shiny example of how electronic health records are supposed to work. Although I’m a Vet, I don’t qualify for VA services, but Yvonne and I get our healthcare from the Baylor Scott & White System. They use the EPIC EHR system across all of their clinics and labs. That greatly simplifies things for us, from making it easier to setup appointments, to speeding the check-in process, to getting results and having online access to our patient records. We’re quite happy with EPIC but might have trouble accessing our data if we have to go to another system for care.

Data Ownership Notes

As e-Patient Dave says, “Give me my damn data.” Because hospitals see patient data as proprietary, gaining access to it can be difficult, especially if you need it in a different format. That brings up questions like…

  • Where’s the data stored (in the cloud, on a local server, on a mobile device)?
  • Who can change it, and is there an audit trail of such changes?
  • Who’s liable for data errors?

Legal Notes

The incentive among doctors and hospital systems is to CYA (cover your ass), even if that means modifying records to cover-up mistakes. The incentive among product developers is to protect proprietary technologies and company reputations, often with gag orders. The general lack of trust seems to have caused everyone to take a defensive posture.

Related mHealthTalk articles on EMR & PHR systems

Related 3rd party articles & apps

  • Few U.S. hospitals can fully share electronic medical records (Reuters)
  • Get your electronic health record: It’s your right (LA Times, 9/11/2015. I commented)
  • Medcorder is an interesting and free app for iOS and Android that makes it easy to record and transcribe your doctor visit. Remember the conversation and detailed instructions is now easier — a win-win for you and your doctor. That itself is interesting enough, but I see bigger opportunities for this company and will reach out to them. I imagine connecting the transcription to the Electronic Medical Records and using something like IBM SYNTRAN to create abstracts or search as needed.

ABOUT THE AUTHOR

Wayne Caswell is a retired IBM technologist, futurist, market strategist, consumer advocate, sleep economist, and founding editor of Modern Health Talk. With international leadership experience developing wireless networks, sensors, and smart home technologies, he’s been an advocate for Big Broadband and fiber-to-the-home while also enjoying success lobbying for consumers. Wayne leans left to support progressive policies but considers himself politically independent. (contact & BIO)

Similar Posts

4 Comments

  1. Vikram Singh says:

    *Some points that could make an EHR into a chaos have been outlined. EHR Service providers must find ways to resolve these issues. Interoperability is other point to be taken care. Thanks for posting.

  2. Result Laser Clinic says:

    Yes, the article I was looking for. Your article gives me another approach on the subject. I hope to read more articles from you.

  3. MORE ABOUT SYNTRAN…

    I mention IBM’s 1965-era SYNTRAN (SYNtactic TRANSlation) in this article, but with little detail. While researching it and other automatic abstracting systems online, I came across an article by Horacio Saggion, a PhD professor at University Pompeu Fabra. And to capture what I know about SYNTRAN and seek his insight, I sent the following note.


    Dr. Saggion,

    I came across your 1997 article, “Automatic Abstracting; towards a Text Based Generation,” while researching an earlier automatic abstracting system, SYNTRAN. And I hope you might weigh in on the vision I have, based on that old technology.

    MY PERSPECTIVE AND INTEREST: I retired from IBM in 1999 after over 30 years, where I evolved from punch card and computer operator through programming, systems engineering, marketing, and strategy. Yes, I had several encore careers afterwards but am completely retired now, although I still maintain a website and blog about healthcare policy and technology. I see potential for using voice dictation and automatic abstracting in the area of Electronic Medical Records, at least for the text and data portions.

    IBM SYNTRAN was developed in 1965 as a natural language processing system for automatic Indexing, Abstracting and Retrieval. I studied this system while taking a graduate course in Thesaurus and Linguistic Systems at American University in Washington, D.C. as part of an Information Science program. Unfortunately, I no longer have the term paper I wrote about it, and I’m unable to find any details in my online research.

    What I remember of SYNTRAN suggests that the concept has great promise today, given the tremendous improvements in compute capacity, memory, storage, program languages, user interfaces, and network access over the last 50+ years. You see, SYNTRAN was written in Autocoder, a derivative of Assembly Language, for IBM 1401 with 8K of memory and 4 tape drives. Today’s smartphones are a million times faster.

    The basic premise of SYNTRAN, which had a built-in understanding of natural language English prose, was that the most important part of a sentence was like the first or last sentence, and the most important part of a chapter or book was the first or last sentence. With that understanding, the program would craft an abstract with sufficient accuracy to provide virtually the same information as would be abstracted manually. Moreover, SYNTRAN could be instructed to provide abstracts of varying size: say 1%, 3%, or 5% the size of the original manuscript. It could also maintain a bias to serve a particular purpose, possibly medical records. I remember reading abstracts of different sizes with wonder, because it seemed that a human wrote each.

    MEDICAL APPLICATION: For physicians, getting relevant information in and out of the medical record is a huge problem, and it’s partly responsible for the high cost of American healthcare, which already exceeds $3.65 trillion/year and continues to rise. Some have resorted to having paid transcriptionists in the exam room to feed dictated comments into the medical record, but speech recognition software can do that too, especially if it’s medically specialized. Next comes the problem of searching through medical records for relevant and useful information – not just the patient’s record, but records from a larger but still relevant population. That’s where natural language indexing, abstracting, and retrieval come in.

    IBM’s Watson supercomputer is now using its linguistic understanding to make sense of huge amounts of complex information in split seconds, rank answers based on probability, and learn from its mistakes. In health care, Watson is helping doctors tailor medical treatment to every patient’s situation, even as the amount of medical information is doubling every five years. While the supercomputer power of Watson can be made available as a cloud service, it seems that something like SYNTRAN could run on a smartphone.

    Imagine a doctor sifting through a large medical record in real time, with the ability to vary the size of the abstract, giving higher priority to today’s topic of interest (diabetes related weight gain, or trends in cholesterol levels and blood work).

    The next challenge would be the capture and meaningful analysis or understanding of non-textual information, such as images, videos, and spoken conversations. Speech recognition, combined with an indexing and abstracting system, should give doctors summarized access to conversations from previous visits, even searching through conversations and replaying recordings from a specific point when a given topic is discussed. This all seems possible based on what we see today in Watson, and possibly even on a smartphone device based on the foundation set 50 years ago with SYNTRAN. ‘Thoughts?

  4. RELATED REFERENCES:

    Data Standardization And APIs Can Help Pave The Way Toward Data-Driven Healthcare (9/28/2022, Forbes) Fast Healthcare Interoperability Resources (FHIR) is a new standard for APIs in healthcare that is truly forward-thinking and could solve some of the problems described in my article, especially when applied with public policy incentives.

    BRINER, L. L. and CARNEY, G. J. SYNTRAN/360, a natural language
    processing system for preparing text references and retrieving
    text information. IBM Corp. , Gaithersburg, Md., 1968.

    Patent filings hint at Apple’s potential move into managing healthcare records (FierceHealthcare, 4/15/19) — Apple has already made notable inroads into healthcare with new health features on its Apple Watch and its continued expansion of its Health Records on iPhone feature.

Leave a Reply

Your email address will not be published. Required fields are marked *