Browse by Topic

Podcast by Topic

In the News

How Meditech Plans to Integrate Google’s Generative AI Into Its EHR

September 24, 2023

Along with many other companies in the healthcare technology space, Meditech is exploring ways to enhance its software through the use of generative AI.

The EHR vendor has been partnered with Google Cloud for about five years. Over the past year, Meditech has deepened its relationship with the tech giant by exploring ways to embed Google’s generative AI into its EHR. But Meditech is not adding generative AI to its EHR capabilities just because that seems like the hot thing to do right now — the company is approaching its generative AI efforts in a steady, intentional way, COO Helen Waters said in an interview this week.

Given that the digital health field is in the midst of a generative AI hype cycle, it’s imperative that companies in this space don’t fall into the trap of implementing new technologies just for the sake of adopting something new and exciting, Waters noted. Meditech is avoiding this by focusing its generative AI efforts on specific use cases that the company thinks will have serious potential to alleviate clinicians’ burnout, she said.

For one of its generative AI projects, Meditech is using Google’s large language models (LLMs) to power the search and summarization experience within its EHR. Meditech is relying on Google’s LLMs for data harmonization so that clinicians can quickly access a longitudinal view of their patient. In other words, the effort is seeking to ensure clinicians have quick and easy access to all relevant information about a patient — including health data from the Meditech Expanse EHR, health data from legacy technology platforms, scanned handwritten notes and medical images.

Having swift access to a comprehensive view of their patient accelerates physicians’ ability to make sound, informed decisions about treatment, Waters pointed out.

Meditech is also exploring how to layer Google’s Med-PaLM 2 into its EHR’s search and summarization capabilities. Unveiled in April, Med-PaLM 2 is a medical AI system that harnesses the power of Google’s LLMS. The tool is currently being piloted at Mayo Clinic and other health systems — they are testing its ability to answer medical questions, summarize unstructured texts and organize health data.

Once multiple LLMs are layered together into the EHR, clinicians may soon be able to ask the EHR more intelligent questions about the patient data that’s being summarized, said Rachel Wilkes, Meditech’s director of marketing.

As it works to integrate Google’s generative AI into its technology, another use case that Meditech is focusing on is the auto-generation of clinical documentation. Specifically, the company is developing an EHR functionality that will generate generate a “hospital course narrative” — a summary of the patient’s stay, composed at the time of discharge.

When a patient has an acute inpatient hospital stay, the length of the stay and the complexity of the care provided can make the documentation process quite arduous at the time of discharge. Providers tell Meditech that this documentation process can take 30 minutes of a clinician’s time each time a patient leaves the hospital, Wilkes pointed out.

She said Meditech is currently working with Google to determine the best way to leverage its LLMs to generate hospital course narratives in the EHR. These summaries of a patient’s stay will be presented to clinicians in their Meditech Expanse workflow, and they will have the option to edit the summary or any drafted pieces of documentation included within it.

“We’re not doing this just to do it. We’re looking to see how we can use this technology to make sure we can help our organizations deliver safe, efficient, impactful care. We’re doing this in a thoughtful, deliberate way. We’re doing a very careful review of the use cases that we’re pursuing, how they impact existing workflows and how we can embed this into Expanse for the betterment of the experience for our users,” Wilkes declared.

Photo: invincible_bulldog, Getty Images

Epic's biggest moves in 2023

September 24, 2023

Naomi Diaz - Tuesday, September 5th, 2023

From deepening its integration of Microsoft's generative AI tools into its ecosystem to rolling out a new online app gallery, here are 14 of Epic Systems' biggest moves this year:

UPMC is moving from nine EHRs to Epic Systems.
Epic Systems integrated two-way patient texting, which allows hospitals and health systems using its software to offer patients mobile messaging that can send them alerts about appointments, prescriptions and billing.
Epic said it will roll out an online app gallery to help users find resources and connections.
Epic Systems founder and CEO Judy Faulkner told audiences at the company's annual Users Group Meeting that the vendor plans to provide more training for workers struggling to learn its software systems, connect millions more patient records to data systems and use generative AI to reduce provider workload.
Epic Systems and Microsoft again expanded their partnership and will integrate conversational, ambient and generative AI technologies into Epic's EHR ecosystem.
KeyCare, the only Epic-based virtual care company in the U.S., ended its first year with 10 health system and health tech partnerships.
Epic created the "Partnership and Pals" collaboration program allowing companies to work with the vendor.
Epic Systems said it will partner with Nuance, a clinical documentation software company owned by Microsoft, to integrate GPT-4-powered clinical documentation technology into the company's EHR software. Under the partnership, Nuance's Dragon Ambient eXperience Express, an AI-powered clinical documentation application that uses GPT-4, will be integrated into Epic.
According to KLAS Research, the EHR vendor added 83 hospitals to its network in 2022, the most of any EHR vendor last year.
Leading health systems from across the country, including Palo Alto, Calif.-based Stanford Health Care and Rochester, Minn.-based Mayo Clinic, have pledged to use Epic's software to share health information with the Trusted Exchange Framework and Common Agreement.
Epic entered a partnership with Microsoft to develop and integrate generative AI into its EHR software. Under the partnership, the companies will combine Microsoft's Azure OpenAI Service with Epic's EHR software. UC San Diego Health, Madison Wis.-based UW Health, Chapel Hill, N.C.-based UNC Health and Palo Alto, Calif.-based Stanford Health Care have already started to use the new integration to automatically draft message responses.
The EHR vendor snagged New Hyde Park, N.Y.-based Northwell Health as a new customer. The health system will move from an Allscripts EHR system to an Epic one.
Epic Systems was approved for onboarding to join the Trusted Exchange Framework and Common Agreement, a new health information exchange framework.
Epic was named the top overall health IT software suite for the 13th year in a row in the 2023 Best in KLAS Awards.

Reproducibility of GPT4 in Healthcare - An Experiment

September 24, 2023

Bill Russell on a recent podcast covered HealthcareITNews article entitled "Generative AI 'not reliable yet,' says Mayo Clinic's John Halamka, M.D., M.S."

I've been enthralled by OpenAI's function calling where you can basically request a response in JSON to plug into your user interface (ie FHIR app, mPage, etc).

So I created a simple webapp at gptexperiments.patient.dev that would use function calling and let the user enter a triage nurse blurb and get the top five likely conditions according to GPT4, as well as probability and reasoning.

As I played around with it, I noticed that even with the same exact input, GPT would give me different answers. So I captured 50 sequential API requests using functional calling against the OpenAI API and put them in a google sheet ("overall" sheet) You can check out the exact object sent at the bottom (and copy and paste for yourself)

The nursing triage note (user prompt) was "55M hx of DM, HTN with 2 weeks of right upper quadrant, chest pain."

49/50 (98%) of the time, GPT returned the top, most probable diagnosis as having to doing with gallbladder disease (acute cholecystitis, chronic cholecystitis, gallstones, etc).

Oddly, the one non-gallbladder related top diagnosis was GERD, which is somewhat disturbing as I asked GPT to image they were an ER doc ... and I don't know any ER doc who would think the most probable diagnosis in a "55M hx of DM, HTN with 2 weeks of right upper quadrant, chest pain" would be GERD. I imagine most would think of more life threatening diagnoses as acute coronary syndrome, gallbladder pathology, pancreatitis, etc.

I asked GPT Advanced Data Analysis to help with how similar each run was to the other. It suggested using Jaccard similarity.

"On average, the runs have a Jaccard similarity of 3.41%±6.97%, indicating that there's a moderate amount of variation in the diagnoses generated across different runs."

GPT Advanced Data Analysis Jaccard Similarity for 50 Runs of Triage Nurse Narrative to five conditions.

My take home from the experiment is:

Function calling has insane potential for FHIR apps / mPages / anything in the EHR that can do a simple network request and then process JSON.
There is a considerable amount of variation which worries me, but it probably produces more relevant and useful NLP responses than anything we got. It brings to mind the saying, "Don't let perfect be the enemy of getting stuff done." Given the right use cases (such as a co-pilot for differentials based on minimal triage text), functions calling in GPT can be helpful in the ER, especially at 3am.

Addendum:

I used the default temperature (0.8) in creating the sheet overall. I ran again with temperature = 0 on overall2. There does seem to be less variation, however, there is still considerable variation beyond the top two diagnoses.

Appendix: This is what was sent to GPT 50 times in a row:

Digital Health – A Fast Track Guide for Healthcare Workers

September 24, 2023

In today’s rapidly evolving healthcare landscape, embracing digital health is no longer an option – it’s a necessity. The benefits of integrating technology into healthcare are undeniable: enhanced patient care, streamlined operations, improved outcomes, and, when done well, higher satisfaction for all stakeholders with lower burnout for employees and staff. Given the benefits, you cannot afford to delay. It’s essential to create a fast track to digital health.

Before listing 10 actions you can take now, I want to address one factor that deserves special attention: vendor loyalty. Too much or too little is unwise. If you’re fortunate, you will view and work with your primary vendor as a partner, a key member of your team. That said, no primary vendor provides all the IT solutions you need. Every shop is multivendor. Even if you buy all your products and services from one vendor, they have integrated products and services from many other sources (vendors). Think of your vendor as a platform provider.

Embracing your vendor for as many solutions as possible is great ONLY IF they can deliver the products and services you need quickly and cost-effectively. If you wait for that vendor to develop a product or provide a service that you could get from another vendor, you decelerate the benefits realization that could potentially save lives, alleviate suffering, reduce expenses, increase revenues, and/or improve satisfaction.

In our mobile-centric world, it’s beneficial to deploy as much functionality as possible on a device carried by patients and their families and providers. If your vendor doesn’t provide the broadest range of functionality on a mobile platform, find a partner who can. Often, mobile platform providers can integrate broad functionality from a range of applications that uses your company’s brand and masks the family of technologies you’re using, enhances usability, improves care, and increases provider brand loyalty.

ACTIONS

Adopt a Consumer-Oriented Attitude. Healthcare delivery is no longer limited to traditional healthcare providers. To compete with non-traditional participants and because it makes sense, you must become more consumer oriented. Make care convenient and affordable (think pricing transparency), efficient, personalized, mobile-enabled, and easily customized. Ensure your care teams mirror the makeup of your community and develop cultural competency. Do whatever you can to help patients and their families establish, preserve, or restore a sense of self-control.
Enable Mobile Health (mHealth) Solutions. The ubiquity of smartphones opens doors to what should be your sponsored or provided mHealth applications that empower your patients to take control of their health. From medication reminders to fitness tracking, mHealth apps can enhance patient engagement and adherence to treatment plans. Let your technology help manage their care.
Invest in AI and Machine Learning. You must use Artificial Intelligence (AI) and Machine Learning (ML) to assist in healthcare diagnostics, treatment recommendations, and predictive modeling. Integrate AI-powered tools to aid clinicians in producing documentation and making accurate and timely decisions. For AI in particular, discuss all the potential consequences of relying on AI for decision making. Use the same discipline, processes, and governance as you would for clinical trials.
Harness the Power of Data Analytics. Data-driven insights are a game-changer, particularly in healthcare. Leverage advanced analytics to identify trends, predict outbreaks, and optimize resource allocation.
Stay Agile and Open to Innovation. The digital health landscape is dynamic, with innovations emerging frequently. Maintain an agile mindset and be open to reviewing current investments and optimizing existing processes, often with little investment. Cautiously experiment with new technologies that can enhance patient care and operational efficiency.
Embrace Telemedicine with Open Arms. Telemedicine has revolutionized patient-doctor interactions, offers convenience and accessibility like never before. If you have not already done so, invest in user-friendly telemedicine platforms that facilitate virtual consultations, remote monitoring, and secure data sharing. Prioritize patient privacy and ensure compliance with relevant regulations (such as HIPAA) to build trust in the digital healthcare ecosystem.
Prioritize Data Security and Privacy. The digital health realm comes with cybersecurity challenges. Invest in robust data security measures to safeguard patient information from breaches and cyberattacks. Encryption, regular security audits, and staff training are crucial components of a comprehensive cybersecurity strategy.
Foster Collaboration and Interconnectivity. Seamless communication among healthcare professionals is pivotal. Embrace collaboration tools and platforms that enable secure messaging, video conferencing, and real-time information sharing.
Engage in Continuous Training. Digital health technologies evolve rapidly. Keep your staff well-trained to adapt to new tools and techniques. Organize workshops, webinars, and training sessions to ensure that everyone is up to speed with the latest developments.
Promote Patient Education. Empower patients with digital resources that provide accurate medical information, lifestyle tips, and self-care guidance. Work with patients to instill a sense of personal accountability to improve their overall situation and reduce unnecessary visits to your facilities.

By embracing any or all of these strategies, you can help your organization get on the fast track to digital health. The journey will be challenging, but the rewards in terms of improved patient outcomes and operational excellence make every step worthwhile.

Below is a list of 10 digital health blogs I wrote that might be useful. Each takes about 3 minutes to read.

Link: Digital Health – Is Healthcare Ready? Are You and Your Organization Ready? (Jan 2020)
Link: Becoming a Digital Health System (Mar 2020)
Link: Digital Health – Governance in a Digital Health System (Jul 2020)
Link: Digital Health – The Role of Empathy and Understanding (Oct 2020)
Link: Digital Health – Language and Comprehension (Oct 2020)
Link: In a Digital World, The Human Elements are Essential (Feb 2021)
Link: Digital Health – Planning for the Virtual Campus (May 2021)
Link: Digital Health – A Practical Model for Change Management (Jul 2021)
Link: Digital Health – Literacy Matters (Oct 2021)
Link: Data Transformation Model – Deliberate Approach to Achieve Widespread Data Utility (May 2023)