Home » News » Currently Reading:

News 8/21/19

August 20, 2019 News 6 Comments

Top News


Nature magazine has run some good healthcare related articles lately (they are always called to my attention via tweets from Eric Topol). A new perspective piece covers the responsible use of machine learning in healthcare, containing many points that are likely new to the healthcare-inexperienced technologists who might be searching for any protruding nail for their proudly-created hammer:

  • Choose the right problem, not just one for which a convenient ML training database exists. You can predict in-hospital mortality from a wealth of data, but does it tell clinicians something they don’t already know? Will the right people be involved in considering the actions that will be taken in response?
  • Make sure the data elements are appropriate. ICD-10 codes entered after the patient’s encounter won’t be available when they are needed. They may also be driven by billing requirements rather than clinical ones.
  • Account for inconsistent data collection practices across departments and health systems.
  • Make sure that training data represents all populations.
  • Watch for potential bias, such as creating an algorithm of whether a patient should have surgery based on those patients who actually did, who are probably more affluent than those who didn’t. Or in cases of a system that can infer information that the patient declined to provide, such as smoking or HIV status, which may cross ethical boundaries.
  • Avoid “label leakage” in model testing, such as randomly assigning X-rays between training and testing sets without recognizing that patients have multiple images, which would then overweight the model’s accuracy.
  • Break out the model’s testing results into the specific areas where it either excels or fails. Potential users need to know what a particular model works well in adults but not pediatrics, for example.
  • Use clinically relevant evaluation metrics that look at the positive predictive value and sensitivity. A model whose high false-positive rate predicts a situation that requires high-cost, limited-value drug therapy isn’t going to be useful.
  • Publish results with restraint, sharing code, data sets, and documentation so that other researchers can make their own assessments of usefulness.
  • Test the system on real-life patients in silent mode only, where clinicians review the predictions without acting on them. Then move on to randomized controlled trials while recognizing that randomization at the patient or physician level is difficult and could endanger patients.

Reader Comments

From Mo Exposure CEO: “Re: links. Thanks for linking to our company’s news item. The response from HIStalk readers was amazing.” Thanks for deciding to sponsor the site as a result, especially since I don’t run fluff news pieces, meaning your announcement had to earn its way into my news post. Items I mention sometimes get a lot of clicks. Even webinar announcements sometimes get a couple of thousand clicks, and announcing a new sponsor always draws several hundred. Sponsor support comes from having loyal, influential readers, so my only job and outcomes measure is to make it worth their while to return.

From She Lives on Love Street: “Re: [RCM business line omitted.] Word on the street is that it’s been sold to [acquirer name omitted].” I’ve emailed the rumored acquirer’s PR contract but haven’t heard back. I’m running the redacted version to remind myself to follow up.


From Core Cutter: “Re: Apple. This CNBC article suggests internal trouble with its health offerings.” Reporter Chrissy Farr sometimes writes good health IT-related stories, but this isn’t one of them. I suspect her editors are pushing her too hard into drumming up questionably researched, speculation-based stories that hold minimal news value, as in her never-ending quest to take guesses at “what Amazon is doing in healthcare” because naming those big names draws eyeballs, but leave their owners with little useful information. This one tries to extrapolate Apple’s health-related departures into “differing visions for the future” that aren’t backed up by the stated facts. My take:

  • She interviewed “eight people familiar with the situation,” none of whom are the people whose departures she noted, and those people she spoke to are simply speculating on why those people left.
  • The five folks listed as having departed held wildly unrelated Apple “health” jobs, ranging from marketing to wellness clinic executives. It’s not like a mass exodus, either in numbers or in area of focus.
  • The denominator of health-related jobs at Apple isn’t given, so we only know that it’s five positions out of hundreds.
  • The story reports from the unnamed sources a difference of opinion among health-related employees about Apple’s direction, but those weren’t tied directly to the departures and those former employees didn’t say that’s why they were moving on.
  • The clickbaity, present-tense headline implies a sudden uptick in internal tension, but does little to back that up with facts.
  • We don’t have anything to suggest that Apple is disappointed in its health-related results or that it would like to change direction.
  • Health and health IT have always had high turnover, some of it based on unreasonable expectations or finding out that big companies just want to make money instead of making people healthier, but in Apple’s case there’s also the possibility of parlaying an Apple credential into an even better job.
  • Even if the story is right in claiming internal tension, so what? You’ll know if Apple makes major product or organizational changes. Speculating beforehand may be entertaining doesn’t really add value, except for the sites trying to sound insightful.

HIStalk Announcements and Requests


I Google-discovered Zenni in helping a friend get eyeglasses and it’s pretty cool. You key in the vision numbers from your eye doctor’s prescription (they don’t want the actual prescription), use a ruler to measure the distance between your pupils, and then head off through a long list of frames to choose your glasses. It takes a couple of weeks to received them in the mail from China and then you’re set unless you need to bend them a bit like your optician does for a perfect fit (not necessary in his case). The biggest draw beyond convenience is price – a pair of snazzy progressive glasses cost him less than $50 (they look exactly like his $300+ pair from Costco), single-vision sports glasses with polarized lenses were $60, and no-nonsense single-vision sunglasses were $15 (!!). At these prices, you could stash a pair of prescription sunglasses in every car, get some glasses set for computer monitor distance, and get backup normal glasses for next to nothing.  What you end up with is pretty much exactly what the optician would sell you for five times the price after two trips to the store. You still need an eye exam every year or two, but what happens afterward is Zenni’s strong suit. You could do a life-changing but inexpensive good deed by treating someone who can’t afford glasses to a pair of Zennis.

Color me skeptical: a new Frost & Sullivan white paper (which you can download only if you provide work details) predicts that clinical decision support systems “are poised to become the user interface of choice for clinical interaction with health IT,” replacing the EHR. My take is exactly opposite – clinical decision support systems will feed their information and recommendations through the EHR, disappearing in the background but providing no less of a service in recognizing that clinicians rightfully want everything placed into their EHR workflow and design. Nobody in their right mind would suggest that CDSS systems contain everything a clinician needs to see, or to visualize how those systems would interact with the user when several are in use (one for radiology image appropriateness, one for antibiotic stewardship, etc.) I think F&S is way off base here, and had I cared enough to download the report, I bet I would find some CDSS vendor involvement. The HIMSS rag gave it a dramatic headline, a pointless stock art photo, and a non-critical acceptance of what the report’s author said, assuming they paraphrased it accurately. This is one of those reports that predicts huge growth in some market segment, knowing that a more realistic report wouldn’t exactly fly off the shelf.


September 5 (Thursday) 2:00 ET. “Driving 90% Patient Adoption Across Your Network: How US Dermatology Partners is Showing Us The Way.” Sponsor: Relatient. Presenters: Michele Perry, CEO, Relatient; Sara Nguyen, VP of applications and integrations, US Dermatology Partners. US Dermatology Partners is helping its physicians reclaim time they can spend with patients and is turning patient engagement strategies into business results across its 90 locations in eight states. Attendees will learn how US Dermatology Partners defined its patient engagement objectives and physician-optimized strategies. They presenters will provide advice on starting or accelerating  patient engagement goals.

Previous webinars are on our YouTube channel. Contact Lorre to present your own.

Acquisitions, Funding, Business, and Stock


The newly hired, cost-cutting CEO of India’s second-largest hospital chain will slash its expenses by 20% in trying to recover from the misappropriation of company funds by its previous owners. Publicly traded Fortis Healthcare will reduce doctor pay, replace people with software, close underperforming hospitals, sell non-essential assets, and ensure that nurses perform only those tasks that lower-paid employees can’t do.

EHealth Exchange announces go-live of a national, single-connection, InterSystems-powered gateway service whose charter members include the VA, AdventHealth, InterCommunity CCO, and OCHIN.


  • Vanderbilt Health chooses Sectra for PACS and VNA in diagnostic radiology and cardiology.
  • Medical records retrieval vendor Womba chooses Allscripts Veradigm EChart Courier to aggregate provider patient records to its attorney clients.



Spok hires Matt Mesnik, MD (Vigilant Diagnostics) as chief medical officer.


Josh Hoders, MBA (DrFirst) joins Forward Health Group as sales VP.


FDA hires Vid Desai (Vyaire Medical) as CTO.

Announcements and Implementations


A KLAS review of drug diversion monitoring technology finds that Omnicell Analytics is the most widely adopted solution, but many customers fail to achieve their expected outcomes because they decline to pay the extra cost of EHR integration.Medacist RxAuditor has a lot of customers, but outdated technology and workflows cause most of them to use only its simplest dispensing reports in then chasing down problems via manual workflows. Kit Check’s Bluesight for Controlled Substances holds promise based on early adopter reports.


The URL of Sonoma Valley Hospital (CA) is “maliciously acquired,” forcing the hospital to change its prized three-letter domain name of “svh.com” to “sonomavalleyhospital.org.” The hospital’s URL registration was good through late 2021, but someone updated it using credentials from an unknown source to take control, which an expert contacted by the local paper says is nearly impossible to reverse. I checked the Whois for the URL and it’s now running on China-based servers with “registrar lock” turned on. I thought it was straightforward to contact the web registrar or ICANN with proof of ownership to get the transfer reversed, but regardless, hospitals should:

  • Use complex passwords for their domain service’s website.
  • Change the registration address if it points to the same domain since otherwise you’ll lose the ability to be contacted if someone grabs that URL.
  • Turn on the “registrar lock” option of your domain service so it can’t be transferred.


A fired VA hospital chief pathologist is charged with three counts of involuntary manslaughter after a review of his needle biopsy cases showed a misdiagnosis rate of 10%, 10 times the expected rate. The VA says he was responsible for at least 15 deaths and an unknown number of incorrect diagnoses. Colleagues had complained of his erratic behavior for years, but the VA let him continue working while he underwent drug and alcohol rehab, finally firing him in 2018 after a DUI arrest.


Kaiser Health News covers the plight of several “no-stoplight” rural towns that believed Miami entrepreneur Jorge Perez, who promised to save their tiny local hospitals but instead used them in a massive insurance fraud scheme that took advantage of higher lab billing rates for rural hospitals. A 14-bed hospital cranked out bills for $120 million in just six months, of which $80 million went to the hospital’s new owner and little to the hospital, as employees reported running out food, cleaning supplies, and IV fluids, with patients in one of them displaced because their hospital beds were repossessed while they were still occupying them. When insurers eventually stopped paying, 12 of the hospitals filed bankruptcy and eight closed. Perez paid $3.5 million to settle the the DoJ’s false claims charges and says he’ll now focus on his software businesses.

Sponsor Updates

  • Hackensack Meridian Health Jersey Shore University Medical Center reduces stroke-related readmissions by 50% after implementing Vocera Care Inform to provide personalized audio discharge instructions and educational materials.
  • Healthfinch announces several new customers of its Epic-integrated Charlie Practice Automation Platform and its exhibit at Epic’s UGM next week.
  • Aprima will host its 2019 User Conference August 23-25 in Grapevine, TX.
  • Artifact Health publishes a new case study describing how its mobile physician query tool helps Western Maryland Health System accurately code episodes of care in a quality-based reimbursement program.
  • Burwood Group is raising money for the Boys & Girls Club of Greater San Diego.
  • Wolters Kluwer Health releases six new Audio Digest Topical Collections for CME.
  • CoverMyMeds and Culbert Healthcare Solutions will exhibit at Epic’s UGM August 26-29 in Verona, WI.

Blog Posts



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


HIStalk Featured Sponsors


Currently there are "6 comments" on this Article:

  1. My family’s been happily using Zenni for years. Never a problem with the quality of their products. My last order was for progressive bifocals, but I never read through my glasses and I found the near-vision portion at the bottom to be annoying when I walked on trails (couldn’t see my feet!), so I re-ordered for single-vision. Easy to order, and so inexpensive that the cost of a 2nd pair wasn’t a problem.

    • Dittoing on how awesome Zenni is, I’ve been using them for a couple of years. Most people don’t realize how much of an absolute scam optical shops are. A few pieces of CNC’d polycarbonate in a cheap plastic frame is worth $300? Doubtful. That being said, your optometrist is going to fight you on this and might even refuse to give you your full prescription (leaving off PD measurement is common, but not hard to measure yourself, as you mentioned).

  2. Very interesting take on the Frost & Sullivan CDSS report. Since I wrote it, I thought I might provide some perspectives. As our report notes, there are EHRs and then there are CDS systems. One is written on a base of proprietary code, often in archaic coding languages and built on the notion of closed systems. The other is more modern, largely built on open systems and incorporating human factors to ensure low impact human interaction. EHRs are the former and CDS are the latter. Now, CDS are often developed in response to specific pain points: new regulations, new dosing requirements, etc. As a consequence, CDS are now mostly point solutions, but the CDS vendors are expanding their reach and are discovering something interesting: the more they expand, the more EHR data is required. At some point most of the data in an EHR is then in the CDS. We have not reached that point yet, but when that happens,it makes logical sense to invert the model, where an EHR gates most health It functionality and change it to a model where the EHR is actually subordinate to a more user friendly interface: likely one that looks a lot like a CDS. This has happened before in other verticals. In network management, for example, early network management applications were challenging to use, so vendors arose that essentially coopted the UI and subordinated the network management apps to a single pane of glass. So, while the headline was somewhat (maybe a lot) hyperbolic, the essential notion is solid: EHRs are not user friendly, but CDS generally are. This is an unstable dynamic and we believe that someone will move to address the problem. Also, we did talk to practically every CDS vendor as well as every EHR vendor and healthcare delivery organization. The interesting thing is that the idea that CDS would take over did not come from the CDS community, but came from the healthcare delivery community. No CDS vendor paid us for our conclusions. I am happy to discuss this with anyone that would care to connect. I can be reached at mjude@frost.com

    • Mike,

      I read the original and I read your comments.

      You make some sweeping assertions without any basis in fact. “EHRs are not user friendly, but CDS generally are.” – what is “conclusion” based on?

      You assert “CDS are often developed in response to specific pain points: new regulations, new dosing requirements, etc.” – are there any serious CDS systems certified for Meaningful Use AND in general adoption by doctors at the point of care?

      You assert “One is written on a base of proprietary code, often in archaic coding languages and built on the notion of closed systems. The other is more modern, largely built on open systems and incorporating human factors to ensure low impact human interaction. EHRs are the former and CDS are the latter.” – again, based on what? How do you even judge “modern”? Is Oracle (circa 1978) modern? Is SQL Server (fka SyBase) modern? Any open source references with truly “modern” languages you can link us to? If these new wonder kids of healthcare are not proprietary, then you should be able to provide a good selection of open source references.

      And, on what fact base do you assert that EHR vendors don’t “incorporate human factors to ensure low impact human interactions”?

      This article seems to lack real study and instead looks like a low grade garbage report most commonly found with newbie Wall Street analysts telling us why Practice Fusion is the future of healthcare IT.

      Been doing this work for a life time and have watched Arden Syntax, many companies come and go, and dealt with many of the tough issues of providing dependable and beneficial decision support to doctors at the point of care. Your article and commentary don’t seem to hold water.

    • Mike,
      I agree w Mr. H. Your example of network management is just one application not a full business operational system. An EHR is a enterprise wide mission critical transaction system. Furthermore, a CDSS is akin to a Mangement Decision Support System in commercial industry, and I know of no situation where a MDSS has totally replaced a SAP or Oracle ‘transaction’ system.

  3. Oh dear – please update with the name of the RCM vendor when you can! We’re about to outsource with a Brand Name RCM.

Founding Sponsors


Platinum Sponsors




















































Gold Sponsors












Reader Comments

  • Bob Smith: Sault Ste. Marie...
  • DrChiefHustler: Dr Jayne.....I have always loved your blog and I read it everyday.....but I will fight you over Dr Mostashari! :). He's...
  • IANAL: That ideas thing isn't just a Cerner problem. The last five years on Epic's ideas site there has been this simple thing ...
  • John: Nice to finally see a well thought out COVID federal response plan. But sadly makes me wonder: where might we be today i...
  • Brian Too: Just don't get one of those "synthetic leather" chairs. I got one, against my better judgement I'll admit. It was su...

Sponsor Quick Links