Morning Headlines 1/10/19

Good things are happening’ as DoD continues to roll out MHS GENESIS

MHS EHR Functional Champion and Air Force Maj. Gen. Lee E. Payne, MD reassures federal health employees that the MHS Genesis project is on track, and that the next wave of implementations will benefit from improved training processes.

IBM isn’t retreating from using Watson in health care, CEO Rometty says

Despite a plethora of recent bad press about IBM Watson’s healthcare capabilities, IBM CEO Ginni Rometty says it is still a viable part of the company’s business, and that Watson for Oncology is doing well.

Tim Cook teases new Apple services: Our ‘most important contribution to mankind’ will be in health

Apple CEO Tim Cook reveals the company will soon roll out new healthcare services that have been years in the making, and that will ultimately revolutionize the way patients manage their health data.

DSS Acquires veEDIS Emergency Department Information System Software

Federal health IT vendor DSS adds newly acquired Emergency Department Information System software from VeEDIS Clinical Systems to its new Juno EHR.

Northwestern Medicine MyChart Users: Your Legal Options Just Shrank

Northwestern Medicine (IL) requires patients to agree not to pursue class-action lawsuits or jury trials over the mishandling of digital services like Epic’s MyChart.

Readers Write: Expanding the Horizon of Clinical Surveillance

Expanding the Horizon of Clinical Surveillance
By Janet Dillione

Janet Dillione is CEO of Bernoulli Health of Milford, CT.

Pay-for-performance programs, like the Centers for Medicare & Medicaid Services (CMS) Hospital-Acquired Condition Reduction Program (HACRP), determine provider reimbursement based on a hospital’s ability to meet key patient safety and performance measures. To reap the financial incentives—and avoid the penalties—of HACRP, more hospitals are “investing in clinical surveillance solutions that utilize real-time patient data to reveal deteriorating patient conditions at an early stage,” according to a report released in November by AGC Partners, a multi-vertical research and investment firm.

Continuous surveillance traditionally has been the near-exclusive domain of hospital departments that care for high-acuity patients with the greatest risks for deterioration, such as the ICU. However, the persistence of preventable catastrophic events, such as post-surgical opioid-induced respiratory depression (OIRD) — which accounts for more than half of medication-related deaths in care settings — suggests that the ability to monitor patients continuously and communicate insights to clinical teams in real-time must extend beyond the ICU.

According to a new KLAS report on the subject, “clinical surveillance tools hold the promise of giving caregivers clinically actionable insights that decrease mortality, reduce readmissions, and improve overall patient outcomes, and clinicians expect these alerts to be embedded directly within their workflow.”

However, successfully broadening the utilization of this technology can be complex and disruptive and can bring new uncertainties to the entire organization.

How Scalable is Continuous Surveillance?

For many health systems, continuous surveillance can be broadly used with existing technology infrastructure, especially organizations with critical care units or ICUs. Optimizing that infrastructure’s capabilities and incorporating it into existing clinical workflows is the real heavy lift, but advances in monitoring technology, use of real-time physiological data and smart alarms, and sophisticated analytics and the ability to route that information to remote clinicians show promise for scaling continuous surveillance to a number of patient care departments, including telemetry, maternity, med-surg, and even beyond the walls of the hospital.

Additionally, health systems exploring the viability of continuous surveillance are using their EHRs as a natural starting point. Multivariate, real-time data from medical devices aggregated with retrospective data from EHRs, provides a holistic and complete source of objective information on a patient that can be used for prediction and clinical decision making.

Does It Save Lives—and Costs?

Hospital investments in clinical surveillance and analytics solutions are driven by organizations that are migrating toward value-based care models and are trying to achieve the objectives of value-based care, including improving care quality and outcomes, reducing clinical variation, and reducing healthcare costs.

Similarly, patient safety in the era of value-based care is increasingly defined as preventing adverse events before emergency interventions or costly escalations are required. However, most common monitoring practices are reactive, not proactive –interventions are often applied only after a patient has deteriorated.

A number of hospital-acquired illnesses (HAI) could be prevented by continuous clinical surveillance. Sepsis and respiratory compromise are among the most costly in terms of resources and morbidity and mortality.

• Industry costs. Respiratory failure that requires emergency mechanical ventilation occurs in 44,000 patients per year in the United States. The cost to US hospitals for opioid-induced respiratory depression (OIRD) interventions is estimated at nearly $2 billion per year.
• Hospitalization costs. Respiratory compromise ($22,300), ranks in the top five of 20 conditions that have the highest aggregate costs per stay due to the high frequency of hospitalization.
• Length of stay. Ventilator-associated complications (VAC) can lead to longer stays in the ICU and greater rates of readmission. VAC complications add approximately $40,000 in costs to each case, $1.2 billion in total costs annually.

Will Clinicians Adopt It?

Technology implemented without proper consideration of impacts on workflow and user ability to fulfill their core responsibilities can have deleterious effects on its overall efficacy.

Involving direct-care staff is critical to the success of any new technology. How will this new technology impact how nurses deliver patient care? What adjustments in workflow and practice need to be made, at go-live and beyond? Starting with these questions fosters buy-in from the staff who will be utilizing this equipment. If end users are not involved in the selection, adoption, and implementation of a technology, then the likelihood that they will become owners of that product is significantly lower.

According to a clinical surveillance report released this year by Spyglass Consulting Group, “hospitals recognize the importance of real-time capabilities to enhance patient safety and improve care quality.”

Ultimately, the ability to safely manage patient populations across the enterprise, reduce the cost of care, and align with reimbursement and regulatory incentives are driving and accelerating adoption. Clinical surveillance has arrived in healthcare and the future looks bright.

Machine Learning Primer for Clinicians–Part 11

Alexander Scarlat, MD is a physician and data scientist, board-certified in anesthesiology with a degree in computer sciences. He has a keen interest in machine learning applications in healthcare. He welcomes feedback on this series at drscarlat@gmail.com.

Previous articles:

• An Introduction to Machine Learning
• Supervised Learning
• Unsupervised Learning
• How to Properly Feed Data to a ML Model
• How Does a Machine Actually Learn?
• Artificial Neural Networks Exposed
• Controlling the Machine Learning Process
• Predict Hospital Mortality
• Predict Length of Stay
• Unsupervised Anomaly Detection in Antibiograms

Basics of Computer Vision

The best ML models in computer vision, as measured by various image classification competitions, are the deep-learning, convolutional neural networks. A convolutional NN (convnet) for image analysis usually has an input layer, several hidden layers, and one output layer — like a regular, densely or fully connected NN, we’ve met already in previous articles:

Modified from https://de.wikipedia.org/wiki/Convolutional_Neural_Network

The input layer of a convnet will accept a tensor in the form of:

• Image height
• Image width
• Number of channels: one if grayscale and three if colored (red, green, blue)

What we see as the digit 8 in grayscale, the computer sees as a 28 x 28 x 1 tensor, representing the intensity of the black color (0 to 255) at a specific location of a pixel:

From https://medium.com/@ageitgey/machine-learning-is-fun-part-3-deep-learning-and-convolutional-neural-networks-f40359318721

A color image would have three channels (RGB) and the input tensor would be image height x width x 3.

A convnet has two main parts: one that extracts features from an image and another, usually made of several fully connected NN layers, that classifies the features extracted and predicts an output — the image class. What is different from a regular NN and what makes a convnet so efficient in tasks involving vision perception are the layers responsible for the features extraction:

• Convolutional layers that learn local patterns of increasingly complex shapes
• Subsampling layers that downsize the feature map created by the convolutional layers while maximizing the presence of various features

Convolutional Layer

A convolutional layer moves a filter over an input feature map and summarizes the results in an output feature map:

In the following example, the input feature map is a 5 x 5 x 1 tensor (which initially could have been the original image). The 3 x 3 convolutional filter is moved over the input feature map while creating the output feature map:

Subsampling Max Pool Layer

The input of the Max Pool subsampling layer is the output of the previous convolutional layer. Max pool layer output is a smaller tensor that maximizes the presence of certain learned features:

From https://developers.google.com/machine-learning/practica/image-classification/convolutional-neural-networks

Filters and Feature Maps

Original image:

A simple 2 x 2 filter such as 

[+1,+1]

[-1,-1] 

will detect horizontal lines in an image. The output feature map after applying the filter:

While a similar 2 x 2 filter 

[+1,-1]
[+1,-1] 

will detect vertical lines in the same image, as the following output feature map shows:

The filters of a convnet layer (like the simple filters used above for the horizontal and vertical line detection) are learned by the model during the training process. Here are the filters learned by a convnet first layer:

From https://www.amazon.com/Deep-Learning-Practitioners-Josh-Patterson/dp/1491914254

Local vs. Global Pattern Recognition

The main difference between a fully-connected NN we’ve met previously and a convnet, is that the fully connected NN learns global patterns, while a convnet learns local patterns in an image. This fact translates into the main advantages of a convnet over a regular NN with image analysis problems:

Spatial Hierarchy

The first, deepest convolutional layers detect basic shapes and colors: horizontal, vertical, oblique lines, green spots, etc. The next convolutional layers detect more complex shapes such as curved lines, rectangles, circles, ellipses while the next layers identify the shape, texture and color of ears, eyes, noses, etc. The last layers may learn to identify higher abstract features, such as cat vs. dog facial characteristics – that can help with the final image classification. 

A convnet learns during the training phase the spatial hierarchy of local patterns in an image: 

From https://www.amazon.com/Deep-Learning-Python-Francois-Chollet/dp/1617294438

Translation and Position Invariant

A convnet will identify a circle in the left lower corner of an image, even if during training the model was exposed only to circles appearing in the right upper corner of the images. Object or shape location within an image, zoom, angle, shear, etc. have almost no effect on a convnet capability to extract features from an image. 

In contrast, a fully-connected, dense NN will need to be trained on a sample for each possible object location, position, zoom, angle, etc. as it learns only global patterns from an image. A regular NN will require an extremely large number of (only slightly different) images for training.A convnet is more data efficient than a NN, as it needs a smaller number of samples to learn local patterns and features that in turn have more generalization power. 

The two filters below and their output feature maps — identifying oblique lines in an image. The convnet is invariant to the actual line position within the image. It will identify a local pattern disregarding its global location:

From https://ujjwalkarn.me/2016/08/11/intuitive-explanation-convnets/

Transfer Learning

Training a convnet on millions of labeled images necessitates powerful computers to work in parallel for many days and weeks. That’s usually cost prohibitive for most of us. Instead, one can use a pre-trained computer vision model that is available as open source. Keras (an open source ML framework) offers 10 such image analysis, pre-trained models.  All these models have been trained and tested on standard ImageNet databases. Their top, last layer has the same 1,000 categories: dogs, cats, planes, cars, etc. as this was the standardized challenge for the model.

There are two main methods to perform a transfer learning and use this amazing wealth of image analysis experience accumulated by these pre-trained models:

Feature Extraction

1. Import a pre-trained model such as VGG16 without the top layer. The 1,000 categories of ImageNet standard challenge are most probably not well aligned with your goals.
2. Freeze the imported model so it will not be modified during training.
3. Add on top of the imported model, your own NN — usually a fully-connected, dense NN — that is trainable.

Fine Tuning

1. Import a pre-trained model without the top layer,
2. Freeze the model so it will not be modified during training, except …
3. Unfreeze the last block of layers of the imported model, so this block will be trainable.
4. Add on top of the imported model, your own NN, usually a dense NN.
5. Train the ML model with a slow learning rate. Large modifications to the original pre-trained model weights of its last block will practically destroy their “knowledge.”

Next Article

Identify Melanoma in Images

Morning Headlines 1/9/19

The NHS Long Term Plan

In England, NHS issues a long-term plan that calls for technology to, among other goals, improve patient access and the ability to self-manage health, and give clinicians access to patient records from any location.

Healthgrades Acquires Influence Health

Physician and hospital information publisher Healthgrades acquires Influence Health, which offers web services, listings, reputation management, and CRM.

Amazon’s next big thing? Prime, but for healthcare

Analysts predict that Amazon will create Prime for healthcare, which could focus on offering lower drug prices via its acquired PillPack mail order pharmacy, Alexa services, and using its recently announced medical records analytical service to scribe clinical encounters.

AT&T and Rush System for Health Ink Agreement to Use 5G in Hospital Setting

AT&T and Rush University Medical Center (IL) will create the country’s first 5G-enabled hospital and will explore ways that a faster cellular network can improve operations and patient experience.

News 1/9/19

Top News

In England, NHS issues a long-term plan that calls for technology to:

• Improve patient access and the ability to self-manage health
• Give clinicians access to patient records from any location
• Apply best practices using clinical decision support and AI
• Apply population health prediction techniques to assign resources accordingly
• Capture data automatically to reduce administrative burden
• Protect privacy and give patients control over their medical record
• Link clinical, genomic, and other data to improve treatments

Reader Comments

From Jaye P. Morgan Gong Show: “Re: JP Morgan Healthcare Conference. I really like your post about the moneyed investors dodging the homeless on San Francisco’s sidewalks, but let’s remember Jonathan Bush stopping to administer CPR while he was walking down the street. Of course, we know what happened to him.” That unscripted drama from 2016’s conference said a lot about the character of the former Army medic and EMT, who didn’t hesitate to hit the dirt in his expensive suit to perform CPR. He explained at the time, “It was a lot like the healthcare industry: a lot of people were standing around tweeting about it, but no one was trying to do anything about this guy lying in the street. So I was like, turn him the f*** over! It was really dramatic. It was intense. The crowd was rooting for us.” Among the beauty queen sashes I ordered for that year’s HIStalkapalooza was one for JB that said, “I CPR’ed some random guy.” Meanwhile, Elliott Management’s Paul Singer had best hope that JB isn’t the only bystander if he goes to ground.

From Info Blocker: “Re: health system not sharing information with an HIE. That’s not like Epic.” Epic isn’t the problem here, it’s that one of their customers that doesn’t want to share patient information. Suggesting that EHR vendors are the bad guys distracts from reality. You only need to find one user of any EHR system that is sharing data by any means (HIE, Carequality, CommonWell, API, internal app, etc.) to disprove the idea that it’s not possible for that system to share information. The only way the vendor is the villain is if they charge unreasonable fees to make it happen. 

From See Me, Feel Me: “Re: National Federation of the Blind. Is suing Epic for discrimination, saying that Epic’s failure to support screen readers prevents blind people from working in Massachusetts hospitals.” That’s actually old news from July of last year. The organization does of lot of suing for inaccessible websites, self checkouts that don’t work well for the blind, universities that don’t make every function and benefit accessible, and hospitals that don’t offer all materials in Braille or electronic form. Section 508 of the Rehabilitation Act of 1973 requires the federal government to make its own technologies usable by the disabled, but I don’t think the requirement extends further and I’ve only heard of it in the context of public web pages. I can’t imagine that Cerner – the federal government’s most expensive IT system in history – is natively accessible, so if it supports use by the blind, it’s probably through a third-party screen reader. Good intentions aside, I don’t know how someone who is blind could navigate information-packed displays that require clicking, choosing drop-downs, and displaying dynamic patient information. The lawsuit notes that Epic’s patient-facing applications have been made accessible and concludes that “Epic thinks that blind people are only fit to be patients, not healthcare workers.” I’m not sure the sarcastic tone and claims of discrimination will win friends and influence people.

HIStalk Announcements and Requests

We’re putting together our HIMSS19 guide that features HIStalk sponsors, so if your company is exhibiting or attending, contact Lorre to get our information collection form about your booth, giveaways, or activities. You’re probably spending a fortune to be there, so you might as well get some free exposure. She also convinced me to offer some sweeteners to cash-strapped startups who sign up as new sponsors, especially those who realize that their exhibit hall time leaves them out of the spotlight for the 362 days of the year afterward.

HIMSS will have some holes in its agenda if the federal government shutdown continues for 33 more days, which I assume would leave some attendees and presenters unable to attend.

Ellkay sent some fun stuff my way (via Lorre) for the holidays – a beautifully packaged sampler box of their honey (from their rooftop bee hives) and a really cool drawing of their Christmas party guests with a find-the-object game included. I’m generally indifferent to unimaginative corporate giveaways, but Ellkay does it perfectly in not only providing something novel and useful, but that also expresses who they are.

Webinars

January 17 (Thursday) 1:00 ET. “Panel Discussion: Improving Clinician Satisfaction & Driving Outcomes.” Sponsor: Netsmart. Presenters: Denny Morrison, PhD, chief clinical advisor, Netsmart; Mary Gannon, RN, chief nursing officer, Netsmart; Sharon Boesl, deputy director, Sauk County Human Services; and Allen Pendell, SVP of IS and analytics, Lexington Health Network. This panel discussion will cover the state of clinician satisfaction across post-acute and human services communities, turnover trends, strategies that drive clinical engagement and satisfaction, and the use of technology that supports those strategies. Real-world examples will be provided.

Previous webinars are on our YouTube channel. Contact Lorre for information.

Acquisitions, Funding, Business, and Stock

Physician and hospital information publisher Healthgrades acquires Influence Health, which offers web services, listings, reputation management, and CRM.

Change Healthcare and Experian Health will combine their healthcare network and identity management capabilities, respectively, to create an identity management solution.

Patient engagement technology vendor Vatica Health acquires the technology of the defunct CareSync, for which it made a $1 million stalking horse bid to the bankruptcy court in October 2018. CareSync, founded in 2011, burned through nearly $50 million in funding before abruptly shutting down in June 2018.

Analysts predict that Amazon will create Prime for healthcare, which could focus on offering lower drug prices via its acquired PillPack mail order pharmacy, Alexa services, using its recently announced medical records analytical service to scribe clinical encounters, and providing services such as telehealth and medical devices.

Patient matching technology vendor Verato raises $10 million in a Series C funding round, increasing its total to $35 million.

Sales

• St. Luke’s University Health Network (PA) signs a three-year Epic managed services agreement with HCTec.
• Boston Medical Center Health System chooses ZeOmega Jiva for advanced care management in its Medicaid ACO. 

People

CHIME names Stanford Children’s Health CIO Ed Kopetsky, MS as 2018’s John E. Gall Jr. CIO of the Year.

Isaac “Zak” Kohane, MD, PhD (Harvard Medical School) joins the board of Inovalon.

Announcements and Implementations

Vocera launches a new hands-free, voice-powered Smartbadge that offers a larger color screen, improved audio, a dedicated panic button, and extended battery life.

Another 12 health systems representing 250 hospitals join Civica Rx, which will manufacture its own generic drugs – many of them in IV form — to save money and reduce shortages.

Withings will offer consumers a less-expensive EKG device than the Apple Watch with its $129 Move ECG, which hasn’t yet earned FDA’s marketing clearance. AliveCor’s $99 KardiaMobile came out two years ago as the first and arguably best of the lot.

Omrom launches a $499 blood pressure watch that uses an inflatable cuff built into the band rather than the usual questionably accurate optical sensors. It also announces Complete, which adds EKG capability to the blood pressure monitor.

AT&T and Rush University Medical Center will create the country’s first 5G-enabled hospital and will explore ways that a faster cellular network can improve operations and patient experience. 

California health data network Manifest MedEx goes live on NextGate’s EMPI.

Other

Wired magazine notes that China’s healthcare AI efforts, such as imaging analysis, benefit from that country’s less-rigorous privacy regulations that allow vendors to train their systems using millions of readily available patient images. An example is InferVision, which is being tested at Wake Radiology (NC) and Stanford Children’s Hospital.

Vox finds that the ED of taxpayer-funded Zuckerberg San Francisco General Hospital intentionally remains out of network for all private insurers, which the hospital explains is necessary to generate the money it needs to offset charity and Medicaid care. The hospital billed a 24-year-old woman whose broken arm was treated in its ED $24,000 (12 times the Medicare price), of which Blue Cross paid $3,800, leaving her on the hook for over $20,000.

A study published in Health Affairs finds that the 20 top-funded digital health companies have had minimal documented impact on disease burden or cost, with few published studies and an avoidance of measuring outcomes in sicker patients.

A large-scale consumer survey by NRC Health finds that 80 percent would change providers based on convenience alone; long waits and lack of respect are big dissatisfiers; people want to provide feedback quickly after their encounter and preferably by email; and patients don’t care much about provider brand identity and instead focus on their experience with individual clinicians.

Observer polls a panel of experts to name its 20 best “flyover tech” digital health companies that aren’t on either coast:

• Bind (MN) – insurance management for consumers
• Solera Health (AZ) – connecting patients to community organizations and apps
• NightWare (MN) – intervention for PTSD-caused nightmares
• ClearData (TAX) – cloud computing and information security
• Healthe (MN) – eye protection from computer device blue light
• MyMeds (MN) – medication adherence
• Visibly (IL) – online vision testing
• Higi (IL) – health kiosks
• HistoSonics (MI) – non-invasive treatment robotics
• Lumea (UT) – digital pathology
• Springbuk (IN) – actionable health insights
• Sansoro Health (MN) – healthcare data exchange
• LearnToLive (MN) – online mental health treatment
• Smile Direct Club (TN) – teeth straightening aligners
• SteadyMD (MO) – remote primary care that matches the lifestyles of patients and doctors
• Collective Medical (UT) – ED patient data sharing
• Limb Lab (MN) – prosthetics
• Upfront Health (IL) – care journey “next best action”
• AbiliTech Medical (MN) – robotic assistance for those upper-limb with neuromuscular conditions
• Vivify Health (TX) – remote care mobile devices

Tennessee pays contracted doctors a piecework rate for reviewing disability applications, with one of them finishing cases – of which 80 percent were denied — in an average of 12 minutes, allowing him to make $420,000 in the past year and $2.2 million since 2013. At least two of the contracted 50 physicians are felons, while others have had their medical licenses revoked.

In England, experts take a hard-eyed view of sloppily handwritten prescriptions after female patient irritates her eyes with what was supposed to be a soothing ophthalmic lubricant, which the pharmacist mistook as an order for a cream for erectile dysfunction. One might assume that the pharmacist ignored a series of computer warnings for issuing a drug for an inappropriate route of administration and patient sex.

Sponsor Updates

• Divurgent and Gevity will offer their combined healthcare information systems consulting expertise.
• Access and Dimensional Insight will exhibit at the MUSE Executive Institute January 13-15 in Newport Beach, CA.
• AdvancedMD announces the winners of its annual Healthcare Innovator of the Year Awards.
• Tampa Bay Tech awards AssessURHealth with its Emerging Tech Company of the Year award.
• The Best and Brightest names Burwood Group a Wellness Winner.
• The Chartis Group publishes a new report, “Why Your Provider Workforce Plan Isn’t Working.”
• The local news highlights UCSF Medical Center’s use of Collective Medical technology to help “frequent flier” ER patients.
• Divurgent and Gevity announce a strategic business alliance to expand their services across the US and Canada.
• DocuTap announces its 2018 student scholarship essay winners.

Blog Posts

• Are You Ready For Your Next Epic Upgrade? (Optimum Healthcare IT)
• 5 Tips for Adopting Telemedicine into Your Practice (AdvancedMD)
• Bluetree’s take on the new release cycles (Bluetree)
• Hybrid Cloud, not Multi-Cloud, is the Future (Datica)
• 6 Top Cybersecurity Trends to Embrace in 2019 (Burwood Group)

Contacts

Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates. Send news or rumors.
Contact us.

Morning Headlines 1/8/19

Vatica Health Acquires CareSync’s Technology to Expand Value-Based Care Capabilities

Vatica Health acquires CareSync’s care coordination and chronic care management technology following a $1 million bid made last October.

Stanford Children’s Health CIO Ed Kopetsky Receives Healthcare CIO of the Year Award

CHIME and HIMSS honor Stanford Children’s Health CIO Ed Kopetsky with their CIO of the Year award.

Statement from FDA Commissioner Scott Gottlieb, M.D., on the agency’s new actions under the Pre-Cert Pilot Program to promote a more efficient framework for the review of safe and effective digital health innovations

The FDA releases a regulatory framework, test plan, and working model as part of the next phase of its Pre-Cert program for the review of digital health product applications.

EarlySense Completes $39 Million Financing Round to Accelerate Global Expansion of Contact-Free Sensing and Analytics Solution

Patient monitoring company EarlySense secures $39 million in a financing round led by Hill-Rom and Wells Fargo Strategic Capital.

Curbside Consult with Dr. Jayne 1/7/19

The government shutdown has impacted some electronic media, including the National Zoo’s Giant Panda Cam and the National Park Service website. It hasn’t slowed CMS, which continues to send regular emails about the recent final rule redesigning the Medicare Accountable Care Organization (ACO) program. Referred to as “Pathways to Success,” it is designed to advance five goals: accountability, competition, engagement, integrity, and quality. The program modifies the participation options available to push ACOs toward taking on real financial risk faster than they had been under the previous programs.

I have been under the weather and tried to use my illness-imposed downtime to read my way through various fact sheets and documents around the program, but have had trouble making sense of some of it. The CMS press releases reference different announcements and rules that have been put out, including the Calendar Year 2019 Physician Fee Schedule (aka the November 2018 final rule). According to the release, the Pathways to Success final rule also takes a step back in time, finalizing policies for extreme or uncontrollable circumstances for performance year 2017, which were initially established via an interim final rule in December 2017. I had to read that part of the announcement several times since I’m not exactly sure how that works to modify a program year that ended 370 days ago. I thought maybe it was confusing because I was reading it while I was hopped up on cold medicine, but I eventually decided that it’s just confusing.

In trying to distill the communications, my assessment is this. Medicare has figured out that the majority of ACOs are participating in Track 1 for the maximum time allowable and some of them are generating losses. Track 1 is a one-sided model with sharing of savings without the ACO having to take on risk, therefore Medicare absorbs any losses. The original idea was for organizations to use Track 1 as a way to get their feet wet with shared savings in hopes that they’d quickly move to more risk-bearing agreements. That hasn’t happened, so now the proverbial stick has to come out.

The other existing ACO varieties (Track 2 and Track 2) are two-sided. Eligible ACOs share a larger portion of any savings, but in exchange they’re required to share losses if spending exceeds benchmarks. These programs have been shown to generate savings for Medicare and are improving quality, so Medicare wants to further those types of arrangements.

Medicare has also figured out that so-called low-revenue ACOs (mainly made up of physician practices or rural hospitals) are outperforming high-revenue ACOs, which typically include hospitals. There are challenges for the low-revenue ACOs to move to a more risk-bearing arrangement because those organizations may have less control over how their assigned beneficiaries use services and therefore spend money. Medicare piloted the “Track 1+ ACO Model” during 2018, with the goal of proving that a two-sided model with lower risk would be attractive. Its success influenced the construction of the new redesigned program, according to CMS.

The redesigned program offers two tracks, named BASIC and ENHANCED, which are open for five-year agreement periods starting July 1, 2019. The BASIC track lets ACOs start under a one-sided model and gradually accept higher risks as they move through five levels A, B, C, D, and E. Once they reach the highest level, they’d be recognized as an Advanced Alternative Payment Model (APM) under the Medicare Quality Payment Program. The ENHANCED track is based on the existing Track 3 and allows flexibility for ACOs willing to take on the highest levels of risk. The existing Track 1 and Track 2 programs will be discontinued, as will new application cycles for Track 1+. CMS feels those options would be redundant to the new program.

CMS aims to move BASIC organizations through the alphabetical levels (which they refer to as the “glide path”) by automatically advancing them at the start of a new performance year. Organizations would also be able to jump to a higher level faster if desired. The ultimate goal is to move all ACOs to the ENHANCED track, with high revenue ACOs being required to transition more quickly. There are also stratifications based on whether ACOs are identified as experienced or inexperienced with performance-based risk but to be honest I skimmed over those particulars in my pharmaceutical-induced fog.

The final rule updates the mechanisms for repayment when ACOs have shared losses. Both new tracks may start with lower repayment amounts based on a percentage of Medicare Part A and Part B revenues, with the amounts recalculated annually based on changes in the ACO participant list. Benchmarks will also be recalibrated, incorporating data from ACO experience and regional performance measures. The rule also aims to reduce “opportunities for gaming” by holding terminated ACOs accountable for pro-rated shared losses. ACOs are also able to choose between different beneficiary assignment methodologies and to change their selections for subsequent performance years. Starting in January 2020, eligible ACO providers will be able to receive payment for telehealth services for certain beneficiaries in certain situations. There are also changes to expand the Skilled Nursing Facility (SNF) 3-day rule waiver.

The redesign also allows ACOs under certain two-sided models to operate a beneficiary incentive program, which may allow for incentive payments of up to $20 to assigned beneficiaries who receive certain qualifying primary care services from ACO members. It also clarifies that under existing program regulations, vouchers and gift cards can be provided to beneficiaries assuming they meet other program requirements such as being connected to the beneficiaries’ medical care. There are a few other tidbits in the rule including updates to beneficiary notification requirements. Beneficiaries have to be notified of the opportunity to opt-out of claims data sharing along with how to change their assigned primary clinician. CMS is developing templates for these notices in an effort to reduce the burden to participating practices.

I’m only marginally involved in the ACO realm, so I’m sure those who are deeper in the process might have additional insights. I’ll be looking to read digests and summaries in the coming days until I’m on the mend. Until then, my next reading list involves chicken soup.

Email Dr. Jayne.

Morning Headlines 1/7/19

Senate Confirms James Gfrerer as VA’s Information, Technology Leader

The VA finally has its first permanent CIO in two years after the Senate confirms James Gfrerer as assistant secretary of information and technology.

SVB Financial Group Completes Acquisition of Healthcare Investment Bank Leerink Partners

SVB Financial Group completes its acquisition of healthcare and life sciences investor Leerink Partners for $280 million in cash.

Livongo Names Lee Shapiro Chief Financial Officer

Livongo hires its board member and investor Lee Shapiro as CFO.

Monday Morning Update 1/7/19

Top News

The VA finally has its first permanent CIO in two years after the Senate confirms James Gfrerer as assistant secretary of information and technology.

The 20-year Marine Corps veteran and US Naval Academy computer science graduate most recently worked for Ernst & Young. He has spent most of his career working on IT business risk and cybersecurity.

Gfrerer replaces interim CIO Camilo Sandoval, a military veteran and former Trump campaign executive.

Reader Comments

From Casual Commitment: “Re: sharing Epic screen shots. Some customers would release shots of every single screen. It’s hard to create a competitive advantage with software and harder to maintain it over time. Customers could release every screen shot of a new release to the world even before going live and competitors could simply copy it, taking away the incentive to create innovation and usability. Some would argue that it’s in the best interest of science and/or safety, but I think most vendors are OK with using images for those purposes and would not ask a client to take them down. Vendor contracts nearly always require clients to get permission before sharing confidential information and academic medical centers often require the same assurances that their confidential information not be shared.” Sometimes I question whether just getting a look at the user interface exposes the intellectual property underneath, but I admit that I’ve written some programs that were inspired by seeing a screenshot or demo, then figuring out how to make it work under the covers, so I can buy that.

From Epic Customer: “Re: sharing Epic screen shots. Epic publishes its straightforward approach, but it isn’t well publicized at customer organizations and folks don’t check with the IT department when preparing their publications. Then they are notified that they have used a screenshot inappropriately and are upset that their publication needs to be take down or changed. Blaming Epic doesn’t seem appropriate, but clients are challenged to make it clear to staff the restrictions that ALL vendors impose on to prevent inappropriate use of proprietary information. Most health systems have committees, which would seem to be needed to guide compliance.” Epic Customer provided Epic’s predictably thorough and clear screenshot guidelines, which say:

• Any public sharing of Epic content (to websites, other vendors, presentations other than to Epic clients, research papers, publications, and books) requires Epic’s review via its Content Approval process, which takes about two weeks.
• The submission requires stating why using the screen shots is necessary, who will see it, how it will be distributed, and the goal or conclusion of any research (that last aspect troubles me a tiny bit, such as the case where the argument is that Epic’s design endangers patients or burdens clinicians – would that impact the likelihood of approval?)
• Any vendor that will receive Epic screen shots or functionality description must ask Epic before using it.
• An Epic copyright notice must be included on every screen shot.
• Screen shots should crop or blur information not needed for the specific purpose, such as removing menus and toolbars.
• Content can’t be posted on private video or quiz sites (including YouTube) because the terms and conditions of those sites say that anything posted there becomes the property of that site. 

From Merger Frenzy: “Re: CommonSpirit Health (the soon-to-be merged Dignity Health and Catholic Health Initiatives). Deanna Wise was announced as the consolidated CIO a few weeks ago, then was gone a few days later with two interims in place. The organizations have very different cultures and IT systems (Cerner at Dignity and Epic, Cerner, Allscripts, Meditech, and others at CHI, I think) but they’ve been working on this for over two years. In addition, the merger has been pushed back again to February 1.” I’ll first say that I detest that embarrassing married name, as I do any time the marketing geniuses decide it will be amazing to simply remove the space between two words while leaving them capitalized (“common spirit” sounds like a bar’s cheap well drink). Deanna Wise was named CIO of the 140-hospital, $30 billion, Chicago-based mega-system in a December 4 announcement. Her LinkedIn hasn’t changed and she’s still listed as EVP/CIO on Dignity’s executive page. That’s all I know, other than that big-ego organizations that are used to calling their own shots often can’t stop arm-wrestling for control before, during, and after a superficially friendly merger.

From Telebicycle Coach: “Re: information blocking. My HIE employer has ONE large non-profit hospital in the entire state that refuses to contribute lab result data, which I suspect is because it wants to sell Epic to small practices that don’t need it and can’t afford it, so it tells them it’s the only way they can get lab data. I would love to hear thoughts.” Intentional health system blocking is rampant, as you might expect when trying to convince competitors in any industry to share internal information. I don’t really have any new thoughts except to say that it should either be made a strictly enforced law (driven by complaints like yours) or a condition of being paid taxpayer money in the form of Medicare. No amount of shaming or dangling the patient benefit carrot has worked, just like it hasn’t for getting hospitals to give patients copies of their own records quickly and inexpensively. Shame is an effective weapon only to the extent that an organization fears being shamed.

From Barnard Rubble: “Re: big data. Is it still a contender for the HISsies ‘most overhyped’ category?” The perpetual frontrunner is actually not on this year’s ballot due to a plethora of fizzy competitors, such as blockchain and IBM Watson Health. Maybe big data has finally summited Gartner’s Peak of Inflated Expectations, although I expect the Trough of Disillusionment to be in the form of the lawyer’s warning to “don’t ask a question for which you don’t want to know the answer.” Big data will tell us what we already know and can’t solve – that our system healthcare system is unfair, unaffordable, inefficient, reflective of primitive social policies, incapable of delivering consistently high outcomes, and rife with profiteers and political influence. It’s nice for society’s financial winners to foresee a world in which their every malady is machine-diagnosed and optimally treated with the best, most personalized therapies available, but people are suffering and dying due to problems that have nothing to do with analytics. We should just declare ourselves a third-world country and then take the Bill Gates public health funding approach – use analytics to identify the most health-impacting problems that can be fixed creatively and inexpensively at scale to benefit the most people in hoping to move up from the dregs of developed nation health rankings.

HIStalk Announcements and Requests

Poll respondents aren’t interested in reading executive interviews that focus on the company’s products or the individual’s personal history – they would rather learn more about the executive’s views on healthcare trends and what they think about life in general. Me, too – I’ve interviewed executives who couldn’t stop blabbing about the amazing potential of their companies under their skilled watch, only to have the company or their careers take a startling stumble soon after. Show us your character and let the rest of us decide whether that piques our interest enough to want to learn more about your business. I’m proudest of my interviews that led readers to feel that they got to know a CEO.

New poll to your right or here: For longstanding HIMSS members: how do you feel about the organization now vs. five years ago? Vote and then click the poll’s comments link to describe what has changed for the better or worse.

Thanks to these companies for recently supporting HIStalk. Click a link for more information.

Two warnings as HIMSS19 approaches (anybody have other Orlando scam alerts?):

• Don’t book your hotels through email solicitations, even when they list seemingly legit hotel names and prices. HIMSS blocks all the hotel rooms, so you can only reserve through them for hotels on their official list.
• Don’t call pizza places whose flyers are shoved under your hotel room door. Those are scams, too, as criminals make up restaurant names, create flyers with phone numbers, and then steal your credit card information when you order that pizza that never arrives (they could get even smarter using a national brand name like Domino’s but with a phony phone number).

Webinars

January 17 (Thursday) 1:00 ET. “Panel Discussion: Improving Clinician Satisfaction & Driving Outcomes.” Sponsor: Netsmart. Presenters: Denny Morrison, PhD, chief clinical advisor, Netsmart; Mary Gannon, RN, chief nursing officer, Netsmart; Sharon Boesl, deputy director, Sauk County Human Services; and Allen Pendell, SVP of IS and analytics, Lexington Health Network. This panel discussion will cover the state of clinician satisfaction across post-acute and human services communities, turnover trends, strategies that drive clinical engagement and satisfaction, and the use of technology that supports those strategies. Real-world examples will be provided.

Previous webinars are on our YouTube channel. Contact Lorre for information.

Acquisitions, Funding, Business, and Stock

SVB Financial Group completes its acquisition of healthcare and life sciences investor Leerink Partners for $280 million in cash and $60 million as a five-year retention pool for Leerink employees (of which D&B Hoovers says there are just 42).

For those who might otherwise forget that US healthcare revolves around business rather than patients, a gallon of meeting room coffee at this week’s JP Morgan Healthcare Conference costs $170 ($21 per cup), hotel rooms for the besuited moneychangers run thousands of dollars, and borrowing 14 power strips in a hotel conference room costs $1,000. Meanwhile, attendees complain about having to dodge San Francisco’s unwashed to get to their all-important meetings about profiting from healthcare services delivery, as hotel security guards and side job police officers shoo away people who are homeless, addicted to drugs, or suffering from mental illness and are thus offensive to the dealmakers whose influence over healthcare policy and delivery keeps increasing. Someone should snap a photo of a money mover in a $5,000 suit (“come on!”) who invests in tech companies pitching population health management or social determinants of health who snootily sidesteps the people the company claims to serve.

Patrick Soon-Shiong is once again a leading candidate in the “pie in the face” HISsies category, so I took a look at NantHealth share price – it’s at $0.62, valuing the once-touted enterprise at a paltry $68 million. NH shares rose to as high as $21 on IPO day in June 2016, having since lost a startling 97 percent of their value. I’m pretty sure I can predict the HISsies pie vote of  those early shareholders. The logo reminds me of an old, possibly appropriate, not-safe-for-work joke involving a feather and the distal colon.

People

Echo Health Ventures hires Jessica Zeaske, MHS, PhD, MBA (GE Ventures) as partner.

Government and Politics

Livongo hires its board member and investor Lee Shapiro as CFO. Shapiro was president of Allscripts when Livongo board chair Glen Tullman was CEO of that company.

Other

I finally got around to looking over last month’s telehealth-focused Health Affairs, with these snips from various articles catching my eye:

• Not many doctors or patients are using telehealth, although the numbers are increasing and the available information is dated.
• Lessons learned by four large health systems that have implemented a telehealth program include making sure executives agree on its goals and strategic contribution; coordinating telehealth efforts among multiple departments to set priorities and ensure the availability of support resources; identify champions who can help overcome the resistance of clinicians and employees; develop a patient education and marketing strategy; and evaluate outcomes to support improvement.
• Major health system adoption barriers are cost, payment, and technical issues, with a key factor being how well state Medicaid pays for the service.
• A literature review finds that while telehealth interventions appear equivalent to in-person care, its effect on the the usage of other services is not clear.
• Use of remote experts to support neonatal resuscitation at small hospitals reduces transfers and thus cost, but it is rarely used because nobody pays for it.
• Kaiser Permanente doctors who take chest pain triage telephone calls spent less time per call compared to nurses and sent fewer patients to the ED, but patients accepted the recommendations of doctors at a higher rate. Mortality rates for calls taken by doctors and nurses were similarly low, but direct-to-physician protocols worked best to reduce ED visits and costs.
• Appropriate antibiotic use for acute respiratory infections was about the same in telemedicine and in-person visits, but strep tests were used in only 1 percent of direct-to-consumer visits vs. 78 percent in urgent care centers, leading to more repeat visits following telemedicine sessions.
• CMS’s 2013 decision that a telemedicine doctor can serve as the physician backup for advance practice providers in critical access hospital EDs has led some hospitals to replace local doctor coverage to reduce costs.

I took a quick look at the websites of a few big health systems to see if they had posted their price lists on January 1 as CMS requires, with these results from checking news releases and then searching for “price list” (of course, ignoring the fact that the lists themselves are unhelpful gibberish to consumers):

• New York Presbyterian – has information for one campus only that I could find
• Florida Hospital (now AdventHealth Orlando as of January 1) – yes, but as an XML document that gives an immediate browser error
• Jackson Memorial Miami – yes, but buried deep in the site’s structure
• UPMC Presbyterian – yes, in Excel (the best job of all those I checked)
• Methodist Indianapolis – no
• Montefiore – yes
• Methodist San Antonio – yes
• Orlando Regional Medical Center – no
• Methodist Memphis – yes
• UCSF – yes
• Ohio State – yes

Pro tip from downloading hospital price lists: if admitted to UPMC Shadyside and you take Ativan, the 1 mg dose costs 99 percent less than the 0.25 mg dose (I guess it’s expensive to have someone cut the 0.5 mg tablet in half). Their CDM is full of oddities like this that I assume are the result of shortened descriptions that don’t tell the full story. I started to compare prices across health systems for a few common items, then realized how pointless that would be for consumers or anyone else.

Former New York City Health + Hospitals SVP/CIO Bert Robles is fined $9,000 by the city’s Conflicts of Interest Board for convincing an Epic EVP to let his girlfriend take an Epic certification course with him and for asking his employees to get the girlfriend an H+H ID card so she could use office space and computers to study.

California plastic surgeons derail the attempt of cosmetic surgeons to advertise themselves as “board certified,” with the former questioning the American Board of Cosmetic Surgery’s training program as a “bogus marketing tool” and claiming that 24 percent of its members have been the subject of disciplinary actions after they left other specialties to perform elective and cash-paid breast augmentation, hair transplants, and tummy tucks without extensive training.

Sponsor Updates

• The Journal of Clinical Pathways interviews Richard Loomis, MD chief informatics officer for clinical solutions at Elsevier.

Blog Posts

• Why External Audits are the Best Way to Ensure Revenue Integrity (Hayes Management Consulting)
• Creative Solutions to EHR Training Challenges (Part One) (HCI Group)
• Optimum’s Top 5 Blog Posts of 2018 (Optimum Healthcare IT)
• Seeking Solutions: The Opioid Crisis (Huntzinger Management Group)
• IMAT Solutions Helps Health Payers to Become Data-Driven Organizations (Imat Solutions)
• Similarities between HIPAA and GDPR (Impact Advisors)
• The top five takeaways from the 2018 Gartner Identity and Access Management Summit (Imprivata)
• How Allina Health used EHR data to increase provider efficiency and satisfaction (Nordic)
• A Balanced Digital Marketing Strategy for Your Medical Practice (Influence Health)
• A Simple Wish: Free Up Time for Patient Care (InterSystems)
• Do You Have a “Doctor Famous” Problem? (Kyruus)
• The New Year’s Resolutions You Need to Make if Your Industry Relies on Data (Liaison Technologies)
• “This call is being recorded for Quality Assurance …” (Vyne Medical)
• Rethinking Your Routine in 2019 (Netsmart)
• Checklist for a Consumer-Friendly Patient Financial Experience (Patientco)
• Home health agency thinks like a hospital (PreparedHealth)
• What Do the Top 10 Most Wired Hospitals Have in Common? (Spok)
• ONC Technical Expert Panel: An Insider View (Sansoro Health)
• The Ins and Outs of Test Automation (Santa Rosa Consulting)
• Consumerism Will Spark Change Across the Industry: My Top Seven Health IT Predictions for 2019 (Surescripts)
• The Secret to Documenting for Medical Necessity (WebPT)

Contacts

Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates. Send news or rumors.
Contact us.

Weekender 1/4/19

Weekly News Recap

• Alphabet’s Verily raises $1 billion in a new funding round
• Health Level Seven International publishes the FHIR Release 4 standard
• Epic says that it will will move “beyond the walls” to store patient records from dentists, on-site clinics, drugstores, and potentially home and hospice care providers
• The New York Times looks at Facebook’s suicide risk screening algorithms, which try to walk the line between user privacy and public health but have not been validated
• A study finds that free text comments entered by clinicians when overriding clinical decision support recommendations can be mined to identify system errors or shortcomings

Watercooler Talk Tidbits

Readers funded the DonorsChoose teacher grant request of Ms. B from Virginia, who asked for five sets of virtual reality headsets for her high school classroom and for her projects as an instructional specialist (she based her program on STEM integration classes she’s taking through Columbia University and NASA). She provides this update:

I truly apologize for the delay in writing my thank you note, but, there is a reason, a very good reason! Google Expeditions came to our school to launch their newest VR experience and I and my students were a part of the LAUNCH! The pictures included do NOT do justice the the excitement of the day. Imagine third grade students walking around the Coliseum and remarking at all of the archwork, or flying around a Roman ship and realizing how large the galleys had to be! And then being surrounded by bees as they became a flower or as they moved from room to room with the astronauts in the International Space Station. Thank you for helping make this possible!

Anonymous Epic Developer’s donation, when matched with funds from my anonymous vendor executive and other sources, fully funded these teacher grant requests:

• Math manipulatives for Ms. D’s elementary school class in Aransas Pass, TX
• Document camera and headsets for Ms. M’s elementary school class in Myrtle Beach, SC
• Two Chromebooks for Mr. T’s high school math class in Cleveland, TX

An orthopedist raises concerns about “selfie wrist,” a form of carpal tunnel syndrome caused by the wrist rotation of people who take endless mugging phone photos of themselves. I expect the next selfie-induced health crisis to be: (a) women whose lips go permanently numb after too many self-adoring kiss poses for photos; (b) Asians who throw their elbows out making that horizontal V-for-victory thing while preening in front of a scenic backdrop; or (c) tendonitis precipitated by women taking point-of-view photos of their feet from beach chairs pointed at the ocean.

California’s Medi-Cal Medicaid program is investigating private equity-owned Agilon Health, which was subcontracted to coordinate care but reportedly intentionally delayed or denied services purely because of cost. SynerMed, another contractor, was found to have placed patients in danger by denying services and falsifying documents to hide it. Agilon’s own reports show that nurses had to review up to 200 care requests per day and their decisions were reviewed only by a private practice family doctor who checked in during breaks from his own medical practice.

Everybody’s likely to be confused at or angered by hospital-posted list prices, so here’s an early example – the father of a seven-year-old whose broken arm was treated at the outpatient clinic of Cincinnati Children’s Hospital Medical Center – sent there by the ED to save him money – finds that based on the just-posted price list, his $2,000 bill included being charged at the highest ED rate. He also notes that he was charged more for an X-ray than the price list says and was billed for surgery for just having a cast applied by a technician. The hospital responded that nobody pays list prices and bills include facility fees that aren’t part of the item charge. A Louisville TV station notes that two hospitals that are next door to each other have CDM prices of $162,000 and $55,000, respectively, for the same prostate cancer drug.

Cleveland Clinic fires first-year medical resident Lara Kollab, DO, who used her social media accounts to call for violence against Jews and to threaten that she would intentionally give Jewish patients the wrong medications. She graduated from the obviously appalled Touro College of Osteopathic Medicine, which is ironically part of a system that was created to serve Jewish students.

Humans – not particularly bright ones – rise up against the machine (and presumably the potential loss of unskilled jobs) by attacking self-driving test cars in Arizona. Yokels are slashing test car tires, throwing rocks at them, trying to run them off the road or stopping sharply hoping to get rear-ended, screaming at them to get out of their neighborhoods, and waving guns at them. Experts say people are reacting to driverless cars as “robotic incarnations of scabs” and expressing hostility toward big corporations like Waymo owner Alphabet that are furthering their own interests by running beta tests without approval. The companies generally always decline to file charges even when the offender is identified or confesses.

In Case You Missed It

• News 1/4/19
• EPtalk by Dr. Jayne 1/3/19
• Machine Learning Primer for Clinicians–Part 10
• Monday Morning Update 12/31/18

Get Involved

• Join my Rolodex to provide occasional news reaction or ideas
• Contribute regularly as a provider CIO, IT director, or informatics nurse (anonymous or not)
• Be interviewed (providers)
• Deliver an educational webinar (providers)
• Sponsor
• Report a news item or rumor (anonymous or not)
• Sign up for email updates
• Connect on LinkedIn
• Contact Mr. H

Morning Headlines 1/4/19

Alphabet’s Verily gets $1 billion in funding round led by Silver Lake

Alphabet’s life sciences company Verily —  the former Google Life Sciences — raises $1 billion in a new capital funding round.

Medsphere Acquires HealthLine Solutions, Expands CareVue Platform

Medsphere acquires supply chain vendor HealthLine System a few weeks after raising $32 million.

Best Buy bones up on health care biz

Best Buy promotes Asheesh Saksena to president of its new health division, which will focus on helping seniors age in place.

Medici starts 2019 by buying another Austin health-tech startup

Telemedicine vendor Medici acquires competitor Chiron Health for an undisclosed amount.

News 1/4/19

Top News

Alphabet’s life sciences company Verily —  the former Google Life Sciences — raises $1 billion in a new capital funding round.

CEO Andrew Conrad says that the money will prepare the company to “execute as healthcare continues the shift toward evidence generation and value-based reimbursement models.”

Some of Verily’s projects include smart contact lens, continuous glucose monitors, development of bioelectric medicines, hand tremor reduction software, retinal imaging, surgical robotics, healthcare performance measurement, risk prediction models for chronic disease management, and precision medicine.

Reader Comments

From Dr. Demento: “Re: Annals of Internal Medicine article on machine learning and inequitable treatment. I’m wondering how you’ll respond to these, especially the editorial.” The article says that because machine learning analyzes historic data, it will perpetuate longstanding health disparities that are based on racism and classism, an issue that has arisen in other areas such as trying to predict which citizens are likely to commit crimes. The article provides a healthcare example of using ML to predict patient deterioration, which might underrepresent African American patients who were treated differently, or discharge care planning that might place too much emphasis on patients from high-income ZIP codes since they have more control over discharge conditions such as transportation, home meds, etc. I see the potential problem, but all aspects of life contain certain assumptions and biases and it’s asking a lot of a machine to somehow iron them out even though we ourselves are usually unaware of them. There’s also the issue of whether a machine is being “unfair” when it makes observations that may or may not need to consider social determinants of health instead of simply making a recommendation that works for most people. I’m all for being aware of these issues and fine-tuning the algorithms accordingly, but ML has to reflect an inherently unfair reality (see: Flint’s water crisis). I worry more about it drawing incorrect inferences because, like the humans that oversee it, the technology can’t always distinguish between correlation and causation. The bottom line is that humans should always be managing the machine, second-guessing it, and demanding transparency into how it arrives at its conclusions because we’re the adults in the room. Expecting a machine to be less biased than the human history it learns from is a nice idea and worth keeping in mind, but we’re probably a long way from making that the most important issue in machine learning.

HIStalk Announcements and Requests

HISsies ballots went out via email Thursday morning to readers who subscribe to HIStalk updates and have recently clicked the link to read new posts.

I’ve lamented lately that healthcare providers rarely use electronic signature platforms such as DocuSign to process emailed forms, clinging instead to mailed paper copies and faxes. I received an emailed contract for a family event this week sent via ESign Genie – it’s just $8 per sender per month. It’s probably naive to think that just because providers could improve their own efficiency, delight their customers, and improve their records retention for just $8 per month that they will actually do so, but at least we customers know that if they don’t, they simply don’t care what makes life easier or better for us.

Webinars

January 17 (Thursday) 1:00 ET. “Panel Discussion: Improving Clinician Satisfaction & Driving Outcomes.” Sponsor: Netsmart. Presenters: Denny Morrison, PhD, chief clinical advisor, Netsmart; Mary Gannon, RN, chief nursing officer, Netsmart; Sharon Boesl, deputy director, Sauk County Human Services; and Allen Pendell, SVP of IS and analytics, Lexington Health Network. This panel discussion will cover the state of clinician satisfaction across post-acute and human services communities, turnover trends, strategies that drive clinical engagement and satisfaction, and the use of technology that supports those strategies. Real-world examples will be provided.

Previous webinars are on our YouTube channel. Contact Lorre for information.

Acquisitions, Funding, Business, and Stock

Medsphere acquires supply chain vendor HealthLine System a few weeks after raising $32 million that it said would be used for acquisitions.

Epic expands sales to include dentists and life insurance and diagnostics companies as VP of Population Health Alan Hutchison says the company is “moving beyond the walls” as it seeks to create – and dominate – the single repository space for patient health data. Hutchinson adds that the company is also in talks with assisted-living and nursing home facilities.

Apple drags the stock market down after lowering revenue guidance, which the company blames on poor sales in China due to that country’s economic problems and its trade war with the US. Analysts noted that despite Apple’s blaming the Chinese economy alone, the market for expensive phones is falling apart everywhere and consumers are pushing back against the “Apple tax” in which they faithfully line up every year to pay a premium for a commodity product. AAPL shares dropped 10 percent Thursday and the company is now a long way from its recent trillion-dollar valuation, as share price has dropped nearly 40 percent since early October.

Jvion raises a growth equity investment of an unspecified amount, led by JMI Equity. The Atlanta-based company, which last raised $9 million in 2016, seems to have evolved from a coding optimization and analytics vendor to “the market leader in healthcare AI.”

Good news for patients already scrambling to pay for outrageously priced meds: Pending regulatory approval, Bristol-Myers Squibb will acquire Celgene in a $74 billion deal that will combine two of the nation’s biggest pharma companies.

HIMSS forms a media partnership with FindBiometrics, a publisher of biometric ID solution news.

A fascinating ProPublica article looks at the forced “retirement”of older employees, noting that more than half of American workers over age 50 leave jobs because of their employer’s decision, not their own, imperiling their financial planning for retirement. Notes:

• Only 10 percent of those older employees who are pushed out find a job paying the same or more
• Both employers and employees use the term “retirement” to save face
• Companies pitch long-term benefits such as 401Ks and promotions even though they know full well most employees will never benefit from them
• Employers can force older workers out the door by changing their job responsibilities, pay, hours, work locations, or annual review expectations
• Employers use “stealth layoffs” of early retirement and eliminated positions to replace older workers with younger, cheaper employees or to offshore their jobs
• Federal protections have been cut way back as companies – many of them publicly traded and desperately trying to prop up earnings – have made pleas to “remake their workforces”
• Employees who refuse to relocate for job opportunities are often cast aside for more eager co-workers
• IBM is an example of layoffs, forced retirements, and mandatory relocations that push older workers to leave – not just of those who make more money, but to skew the mix younger by ditching older workers
• Companies intentionally skirt age discrimination laws by forcing job applicants to list their education dates (as a proxy for birthdate), hoping to avoid the high insurance premiums and perceived lower productivity of older workers
• The article concludes that older workers need to keep up their learning (especially technology); network instead of wasting time chasing online ads; save money under the assumption that their entire earning career could be limited to the 30 or so years after college and before being forced out of the job market; and to avoid waiting too long to work for themselves instead of for someone else

People

Zac Jiwa (MI7) joins The Karis Group as president/CEO.

Best Buy promotes Asheesh Saksena to president of its new health division, which will focus on technology and services to help seniors age in place. The company acquired senior-focused mobile device and emergency call service company GreatCall last October for $800 million.

NLP software vendor Talix names Bob Hetchler SVP of sales and Eileen Rivera VP of marketing. Both come from Ciox Health.

Announcements and Implementations

Mercy Health Network selects PatientPing’s real-time patient notification service for its ACO members in Iowa.

Health Level Seven International publishes the FHIR Release 4 standard.

KLAS publishes Part 1 of its overview of population health management technology, noting the “good product, good relationship” vendors above and listing the weak partners as Allscripts, Athenahealth, and Philips Wellcentive. HealthEC is the only vendor KLAS calls out as excelling at providing strategic guidance. HealthEC and Health Catalyst finished tops for helping healthcare organizations negotiate value-based reimbursement contracts, while Forward Health Group finished best in ongoing optimization and training. Philips Wellcentive and NextGen Healthcare slipped year over year after acquiring their respective products, while Allscripts and Athenahealth haven’t met customer functionality expectations.

Privacy and Security

HHS publishes cybersecurity guidelines for managing threats and protecting patients, plus technical resources and templates for healthcare organizations of varying sizes.

Other

An NHS provider’s tweet lamenting a string of no-show patients goes viral in England, prompting many to suggest that patients be fined for missing appointments. NHS data show that missed appointments cost the national service $273 million each year and result in over 1 million wasted clinical hours.

Nurses at Intermountain Medical Center in Utah comfort grieving families with printouts of their loved one’s last EKG enclosed in a tube with a note that reads, “May my heart be a gentle reminder of the love I have for you.”

Eric Topol, MD pokes fun at the hype surrounding wearables, a timely observation given the many people who will likely spend the next few weeks addicted to their devices as they struggle to maintain what will soon become short-lived New Year’s resolutions.

This question never seems to go away – is it OK that Epic (reading between the lines here) has a reputation of bringing its legal muscle to bear on anyone who uses a screenshot of its software in an research article, especially if the screen in question was built or customized by a customer? Epic, despite being full of nice people most of the time, is somewhere between rigorous and paranoid in sending the lawyers after anyone (even a customer) who says or writes basically anything about Epic’s contracts, products, project management methods, training materials, or documentation outside of an Epic-controlled environment. If you have experience with this as a provider either way, let’s hear your story.

A couple sues OB/GYN John Boyd Coates, MD after discovering that he, rather than an anonymous sperm donor, is the father of their 41 year-old daughter. The discovery came to light after their daughter received results from a consumer genetics test. Coates delivered the baby girl himself in 1977.

Sponsor Updates

• ROI Healthcare Solutions publishes its 2018 highlights.
• Cerner adds prescription pricing and benefit information from CoverMyMeds to e-prescribing workflows within its EHR.
• Healthcare Growth Partners publishes its December Health IT Insights.
• Nordic posts a podcast titled “Preparing for changes to value-based care reimbursement in 2019.”
• AdvancedMD publishes a case study featuring Surgical Specialists of Jackson (MS).
• Atlantic.Net announces GA of its Windows Server 2019 Datacenter Cloud Serve OS for use in its Public Cloud.
• Datica releases its new book, “Complete Cloud Compliance: How regulated companies de-risk the cloud and kickstart transformation.”

Blog Posts

• Primary care’s next high-return opportunity: Substance use disorder treatment (Advisory Board)
• Happy and Health 2019! (Agfa Healthcare)
• Five Ways You’re Already Using Machine Learning: A Day with AI (Apixio)
• Best Practices for Switching EHRs (Aprima)
• Why Is It Important to Test Backups? (Atlantic.Net)
• Get the most out of your decision support tools (Bluetree)
• ‘Twas the Night Before Clinical Surveillance (Bernoulli Health)
• 6 Top Cybersecurity Trends to Embrace in 2019 (Burwood Group)
• Informed Decisions at the Point-of-Prescribing – A Possible Solution for Medication Adherence (CoverMyMeds)
• Update from the 2nd NCQA Digital Quality Summit (Diameter Health)

Contacts

Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates. Send news or rumors.
Contact us.

EPtalk by Dr. Jayne 1/3/19

I mentioned being tired from seeing patients the day after Christmas. Now that we’ve done our month-end close, I know why. We broke our own record and saw nearly 1,300 patients that day, 70-something of which were on my schedule. It truly takes a village to be able to care for that many patients and I’m grateful that my practice’s leadership believes in systems-based care and building teams so that they can run like well-oiled machines. The Centers for Disease Control lists nine states as having high flu activity and another 11 with widespread activity, so it might be a long winter.

This article in the Journal of the American Medical Informatics Association caught my eye with its title: “Cranky comments: detecting clinical decision support malfunctions through free-text override reasons.” I’ve done my share of chart reviews where providers have entered interesting comments when faced with an alert, so I was eager to read more. The authors looked at their database to identify those clinical decision support rules that had at least 10 override comments. The comments were classified into three categories based on whether the user felt the rule was “broken,” “not broken, but could be improved,” or “not broken.” They also looked at the comment frequency and a “cranky word list heuristic” to rank the rules based on the override comments.

Parsing the comments uncovered malfunctions in more than a quarter of all the active rules in the system, which was higher than expected. The authors recommend that “even for low-resource organizations, reviewing comments identified by the cranky word list heuristic may be an effective and feasible way of finding broken alerts.”

For those who are curious, representative override comments included items such as “you are stupid,” “stupid EPIC (sic) reminder,” and comments with many exclamation points. Other cranky words include: dumb; idiot; please stop; why; misfire error, epic, and wrong. The authors note that “swear words were originally included but are omitted from this list because they did not yield comments. Other organizations may wish to include them.”

I’ve heard a fair amount of chatter among physicians regarding the new rule that starting this month in which all US hospitals must publish their price lists online in a downloadable format. Nearly everyone I’ve heard from agrees that having the charges isn’t terribly helpful. What one would really need to know is the negotiated rate between the insurer and the hospital. Alternatively, knowing the hospital’s self-pay discount might be helpful.

Hospital lobbying organizations are concerned that patients might forego care if they see prices that are too high, not understanding that most services are discounted. Of course, there’s nothing stopping hospitals from publishing that data, but we’re not likely to see it soon. I poked around on the websites of the local health systems and within their patient portals but haven’t been able to find anything. I’d be interested to hear from readers who may have actually downloaded one.

On the flip side, CMS has put forward a proposal to require pharmaceutical companies to publish Medicare and Medicaid prices for drugs featured in TV ads. Physicians I’ve talked to are supportive of that idea, and many would like to see TV advertising of drugs eliminated completely. I don’t usually watch broadcast TV, but was exposed to it over the holidays at relatives’ houses. There certainly are a lot of drug ads out there and some of them aren’t even clear on what condition the drug is designed to treat or which patient population is being targeted. Print ads aren’t much better, and even some of the ads in physician trade publications are confusing.

At the beginning of every year, I do some general office housekeeping tasks, including updating my filing and accounting systems for the year and making sure all my recurring calendar appointments are in order. As I was paging through the weeks, I realized that HIMSS is just around the corner and I’m feeling like the next month is really going to be a crunch. I have some serious shoe shopping to do before then, along with breaking in any new purchases. My favorite pair of “trade show shoes” met its demise last year and finding just the right kicks to get me through hours and hours of the exhibit hall will be a challenge. Advance registration discounts end January 14, so if you’re planning to go but haven’t registered yet, you can still save a few bucks.

Working in healthcare IT has been a great way to meet people who work non-traditional hours. Some people make their employment choices based on their enjoyment of overnight shifts because it suits their personalities as “night owls.” Telemedicine and teleradiology have opened up additional options, which has led to the need for support personnel covering the physicians who are staffing those shifts.

I spent some time working at an observatory and know that I do well on the night shift, but only when I can completely tweak my schedule to support it long-term. The occasional night shift kills me. The New York Times recently ran a piece about businesses that are encouraging their employees to work at periods of maximal wakefulness as a way to boost productivity and to help avoid safety incidents due to fatigue.

The US Navy recently migrated the traditional 18-hour submarine schedule to a 24-hour one, and other companies are allowing employees to select day or night shifts based on their preferences. Working remotely can help as well. I know I’m more productive when I work in focused blocks of time outside of the office and having even a 20-minute nap in the afternoon makes a huge difference for me. That’s supported by the “window of circadian low,” which forecasts a midafternoon dip in wakefulness for most people, as mentioned in the article.

The piece mentions strategies employed by pharmaceutical manufacturer AbbVie, where employees go through a nine-hour program to identify their best times for creativity vs. low-energy tasks. Employees are encouraged to mesh their work and professional lives to harness those periods as well as to work around family commitments, leading to a dramatic increase in employee satisfaction with work-life balance. I’ve worked in several clock-punching organizations where even salaried employees were expected to put on a show of being at their desk at certain times regardless of whether they were actually doing anything productive.

Has your employer done any investigation into harnessing people’s most productive periods during the day? Leave a comment or email me.

Email Dr. Jayne.

Morning Headlines 1/3/19

HL7 Publishes FHIR Release 4

Health Level Seven International publishes the FHIR Release 4 standard.

Bristol-Myers Squibb to buy Celgene in $74 billion deal

The merger, if approved by shareholders and regulators, will create the US’s fourth-largest drug company.

Jvion Secures Significant Strategic Growth Investment to Expand Application of Its Healthcare AI and Prescriptive Analytics Platform

Healthcare AI vendor Jvion raises a growth equity investment of an unstated amount, led by JMI Equity.

Machine Learning Primer for Clinicians–Part 10

Alexander Scarlat, MD is a physician and data scientist, board-certified in anesthesiology with a degree in computer sciences. He has a keen interest in machine learning applications in healthcare. He welcomes feedback on this series at drscarlat@gmail.com.

Previous articles:

• An Introduction to Machine Learning
• Supervised Learning
• Unsupervised Learning
• How to Properly Feed Data to a ML Model
• How Does a Machine Actually Learn?
• Artificial Neural Networks Exposed
• Controlling the Machine Learning Process
• Predict Hospital Mortality
• Predict Length of Stay

While previous articles have described supervised ML models of regression and classification, in this article, we’re going to detect anomalies in antibiotic resistance patterns using an unsupervised ML model. 

By definition, anomalies are rare, unpredictable events, so we usually don’t have labeled samples of anomalies to train a supervised ML model. Even if we had labeled samples of anomalies, a supervised model will not be able to identify a new anomaly, one it has never seen during training. The true magic of unsupervised learning is the ML model capability to identify an anomaly never seen.

The Python code and the dataset used for this article are available here. 

Data

The data for this article is based on a subset of MIMIC3 (Multiparameter Intelligent Monitoring in Intensive Care), a de-identified, freely available ICU database, Using SQL as detailed in my book , a dataset of 25,448 antibiograms was extracted.  The initial dataset includes 140 unique microorganisms and their resistance / sensitivity to 29 antibiotics arranged in 25,400 rows:

• +1 = organism is sensitive to antibiotic
• 0 = information is not available
• -1 = organism is resistant to antibiotic

Summarizing the above table by grouping on the organisms produces a general antibiogram:

A view perpendicular to the organisms axis on the above chart becomes the projection of all the organisms and their relative sensitivity vs. resistance to all antibiotics:

The view perpendicular to the antibiotics axis becomes the projection of all the antibiotics and the relevant susceptibility of all the organisms:

The challenge is to detect an anomaly that may manifest itself as a slight change of an organism susceptibility to one antibiotic, for example, along the black dashed line in the diagram below. During testing of the ML model, we’ll gradually modify one organism susceptibility to one antibiotic and test the model on F1 score, repeatedly, at different levels of susceptibility:

Let’s focus on the most frequent organism in the data set – Staph Aureus Coag Positive (Staph). 

Staph has 6,925 samples out of 25,400 (27.2 percent of the whole dataset) and its average antibiogram in a projected 3D view:

The above chart projected on 2D and rotated 90 degrees clockwise depicts the average antibiogram of Staph:

Note that in the above diagram, zero has an additional meaning besides “information not available.” Zero may be also the result of averaging a number of sensitive with the same number of resistant samples and thus and average of zero susceptibility.

Model

• Task: unsupervised anomaly detection in antibiograms, a binary decision: normal vs. anomaly
• Experience: 25,400 antibiograms defined as normal for model training purposes
• Performance: F1 score at various degrees of anomalies applied to the average antibiogram of one organism (Staph) and one antibiotic at a time

Local Outlier Factor (LOF) is an anomaly detection algorithm introduced in 2000, which finds outliers by comparing their location with respect to a given number of neighbors (k). LOF takes a local approach to detect outliers about their neighbors, whereas other global strategies might not be the best detection for datasets that fluctuate in density.

If k = 3, then the point A in the above diagram will be considered an outlier by LOF, as it is too far from its nearest three neighbors.

A comparison of four outlier detection algorithms from scikit-learn on various anomaly detection challenges: yellow dots are inliers and blue ones are outliers, with LOF in the rightmost column below:

The LOF model is initially trained with the original antibiograms dataset, which the model will memorize as normal.

After the model was trained, we gradually modify the Staph susceptibility to: 

• Vancomycin only: from the original +0.5 to 0.4, 0.2, 0, -0.2,… -1.0
• Gentamicin only: from the original +0.8 to 0.4, 0.2, 0, -0.2,… -1.0
• Both antibiotics at the same time: 0.4, 0.2, 0, -0.2,… -1.0 sensitivity / resistance

Below is a Staph antibiogram with only 13 percent sensitive to gentamicin compared to the normal Staph antibiotic susceptibility at 83 percent for the same antibiotic:

And the same comparison as above, but with a Staph population that is 13 percent resistant to gentamicin:

Performance

The last example, Staph susceptibility to gentamicin significantly shifting from (+0.83) to (-0.13) creates a confusion matrix with the following performance metrics:

• Accuracy: 87.3 percent
• Recall: 82.6 percent
• Precision: 91.3 percent
• F1 score: 0.867

At each antibiotic sensitivity / resistance level applied as above, the model performance is measured with F1 score (the harmonic mean of precision and recall detailed in previous articles). The model performance charted over a range of Staph susceptibilities to vancomycin, gentamicin, and both antibiotics:

The more significant an anomaly of an antibiogram, the higher the F1 score of the model. An anomaly may manifest itself as a a single large change in the sensitivity of one organism to one antibiotic or as several small changes in the resistance to multiple antibiotics happening at the same time. A chart of LOF model performance over a range of anomalies can provide insights into the model capabilities at a specific F1 score. For example, at F1 = 0.75, Staph sensitivity to gentamicin declining from (+0.83) to (+0.2) will be flagged as an anomaly, but the same organism changing its vancomycin sensitivity from (+0.53) to (-0.2) will not be flagged as an anomaly. 

There are no hard coded rules in the form of “if…then…” when using an unsupervised ML model. As there are no anomalies to use as labeled samples, there is a need to synthetically create outliers for testing the model performance by modifying samples features in what (we believe) may simulate an anomaly.

In all the testing scenarios performed with the LOF model, these synthetic anomalies have always been in one direction: from an existing level of sensitivity towards a more resistant organism, as this is the direction the bacteria are developing under the evolutionary pressures of antibiotics. A microorganism developing a new sensitivity towards an antibiotic is practically unheard of, as it dooms the bacteria to commit suicide when exposed to an antibiotic to which it was previously resistant.

Unsupervised anomaly detection is a promising area of development in AI, as these ML models have shown their uncanny, magic capabilities to sift through large datasets and decide, under their own volition, what’s normal and what should be considered an anomaly.

Next Article

Basics of Computer Vision

Morning Headlines 1/2/19

Epic Systems adds dentists, life insurers to its software network

Epic’s VP of population health says the company is “moving beyond the walls” in creating a single repository for all of a patient’s electronic health records – dental, on-site clinics, drugstores, specialty clinics, and potentially home health and hospice providers.

When Doctors Serve on Company Boards

Cancer centers are responding to the perceived conflicts of interest that arise when their researchers and doctors serve on the boards of drug, lab, and medical device companies.

In Screening for Suicide Risk, Facebook Takes On Tricky Public Health Role

Facebook has deployed suicidal thoughts detection algorithms without tracking their outcomes even as the company faces global scrutiny for its privacy practices.

Malware attack disrupts delivery of L.A. Times and Tribune papers across the U.S.

Several newspapers are left unable to print editions after Tribune Publishing is taken down by an attack of Ryuk ransomware, a potent strain spread via malicious emails that was the subject of an HHS Cybersecurity Program warning in August 2018.

Morning Headlines 12/31/18

CenturyLink outage knocks out 911 calls, hospital’s patient records

A CenturyLink outage left several hospitals without Internet service Friday, also taking down phones, 911 access, and ATMs all over the country.

Cranky comments: detecting clinical decision support malfunctions through free-text override reasons

A JAMIA-published study of comments that clinicians enter when overriding clinical decision support warnings finds that the text can be mined to identify system shortcomings about 26 percent of the time.

‘Bleed Out’ Shows How Medical Errors Can Have Life-Changing Consequences

A filmmaker’s new HBO documentary covers what he says was a medical error that left his mother with permanent brain damage, for which he primarily blames Aurora West Allis Medical Center’s use of E-ICU coverage.

Stanley Black & Decker Debuts Healthcare And Security Innovations At CES 2019

The company will launch a voice-controlled, smartphone-integrated medication management and caregiver communication tool that supports independent living.