The fatal flaw in machine learning AI

Machine learning is a great thing. Pexels-photo-251287The fatal flaw is our expectations of their capabilities exceed the ability of these systems as designed. Especially when they interact with people. Take the news feed on LinkedIn or Facebook for instance. Their design goal is "show people what they like to see". People will stay longer on the platform and consume more advertising. It works very well. The problem is it collapses everyone's worldview and they can no longer absorb new ideas and viewpoints. Just the opposite of being well read. The social good diminishes.

My point being is that whatever the AI designer's target goal is, that will become extremely efficient. But like all computer systems, they are extremely inflexible in looking at macro impacts, yet we deploy them without understanding the unintended consequences. I like collusion avoidance in a car because it enhances my driving skills. I don't like autonomous driving systems because it makes me a stupid, disconnected driver. 


My 7" 40GB $62 Traveling Tablet: Amazon Fire

On a trip to Indianapolis, I left my iPad on the plane. Tracking it did no good. I was not happy. For Christmas this year I got the cheapo Fire with Firea wimpy 8GB of storage. For $12 I added a 32GB card. Total cost = $62.

It's a great traveling companion. The screen's not the absolute best but works just fine for watching streaming movies (supports all the streaming services I use) and with 40GB downloaded movies. And yes you can read books, play games and store a lot of music. Since it has Bluetooth I use my Bluetooth stereo buds. To protect the screen I slipped it into a standard shoe shining mitt you get at the hotel.

The problem with using your phone on a trip is the battery consumption. Sometimes on a trip to NYC I've depleted the battery so much, I'm afraid I won't have enough to run the Lyft app. Now I turn my phone completely off and just use the Fire. If I happen to leave it on a plane again, I don't think I'll have as much angst.

Business Case for FHIR and Argonaut: Patient Directed Post Acute Care

FHIR's purpose is an ambitious effort that defined healthcare standards and API's to accelerate ArgonautProject_logodevelopment of useful web-like applications from Electronic Health Records (EHR).  Project Argonaut is a private sector based initiative to build useful applications to drive adoption.

Applications for whom?  The current applications are heavily internal EHR focused among multiple institutions and EHR vendors.  

Although the industry talks about "patient centric" medicine, in reality that has as much relevance as "natural" has to foods you buy at the supermarket. Outside of a beauty contest called "patient satisfaction" there is not much in the field focusing on the patient's point of view.

And therein lies the killer app that can drive widespread adoption. Build an app that patients demand.

And it's not personal health records (remember Google Health?)  Instead it needs to solve an immediate problem for a large number of people. I suggest an app  that over 1 million new patients every year have a need for 3-6 months. That happens to be the number and recovery period for total hip and knee replacements (not partial, or repairs) in the United States.  

There are evidence based protocols and procedures for optimal outcomes from surgery to full recovery over a multitude of care settings.

But who is coordinating this effort over the entire lifespan?  Is it:

  • The insurance company (or Medicare)?
  • Primary care physician?
  • Rehab center?
  • Physical therapist?

All of those answers are incorrect.

In  healthcare the presiding belief is that care manager is on the provider side of the house. In reality it is the patient or in many cases the patient's advocate. The patient advocate role could be the patient, significant other or a family member. Things get very complicated when the advocate is caring for Mom who lives three states away.

This is reality.

300px-ArgonautDeviceWe need to build a post acute care platform with a FHIR component.  But it goes beyond health records.  This is the place that the patient advocate can coordinate, collect, manage and distribute information needed for the patient's full recovery (the use cases are numerous blog posts in themselves).  Everything from medication reconciliation, getting DICOM images to the therapist, through hiring Task Rabbit to take Mom to the grocery store or find a plumber to fix the leaking faucet.

By moving the focus to the patient, several authentication and state issues can be avoided. Firstly, patients have access to their records regardless of the state involved.  As for authentication, the patient has a key ring of OAuth tokens where their identity is confirmed by each medical provider or other entity.  (it's signing into a new service using Google in reverse, where the patient is Google).

Create an application like this and FHIR will take off - Because the patient advocate will demand it.


Protecting yourself from hacked credit card readers: Google Wallet & Apple Pay

First TJX with 90 million accounts stolen, then Target with 40 million accounts stolen and now Home Depot with 56 million accounts stolen.  I found it interesting that Target was hacked through a flaw in Microsoft Active Directoy.  No news yet on the details of Home Depot.

What's a person to do?   Buy a phone with NFC payment option.  The newer Android phones have Google Wallet and it looks like Apple Pay is coming soon.  When you activate Google Wallet, you link it to a credit card.  I prefer American Express because they have great anti-fraud detection and allow you to dispute un-authorized charges from their website.

When you are shopping, look for the wireless payment option on the card reader.  Most grocery stores have them and big chains like Best Buy.  Walmart is too cheap and does not (use cash if you must shop there).   WirelessWhen you touch your phone to the card reader, Google prompts you for a pin.  And that's it.  What's interesting is that to the card reader, it looks like a single use Master Card regardless of your actual credit card.  And you need a data connection, because Google sends out the authorization code real time.  Pretty cool.  On your credit card statement it will say GOOGNFC*merchant name

If your phone gets misplaced, you can deactivate your wallet from the Google website.  Of course you have 2-factor authentication for your Google account, right?  And you have a lock code on your phone. And a good PIN for Google Wallet.

Bottom line, if that card reader was hacked, the bad guys only get a fictitious credit card number that can't be used.  Not bad.


Saving Healthcare in the US: Focus on Efficiency, Efficacy and Motivation

I won't bore you with the statistics on how the US spends more and gets less than any other industrial nation in the world and consumed about 17.9 percent of GDP last year.  Instead I want to focus on the goal of colleagues of mine who are serious about shaving 1% of GDP in healthcare.

How to do this?

First off, you need to measure precisely how much you actually spend on each patient.  Then you need to examine how much utilization of resources you actually are using for patient care (not how much you bill).  Focus on maximazing work flow and resource utilization and you now improve efficiency and save money.

But that does not mean you are doing the right thing by the patient.  Next you need to measure efficacy to see if you are getting the best results for what you did.  By doing so we found in occuptational health that patient outcomes improved while physician visits decreased by 40%.

While this is all well and good, it doesn't matter if no one uses it.  So you need to have the proper motivation.  Many times this means a cultural change in the organization.  For example we had a case where an organization could save $8M a year by employing these methods.  The medical director killed the project becauses he did not want his patient outcomes measured.

Efficiency, efficacy and motivation, when implemented, will change the landscape of healthcare.

HealthIT 2.0: Time for the Hospitalist?

Imagine taking care of 15 patients a day.  And you've never met them before.  And coordinating care among three shifts of nurses, labs and specialists.  That is the plight of the hospitalist. 

HealthIT 1.0 has failed them.  Patient histories from multiple sites of care?  Disparate PAC systems, care coordination?  Medication reconciliation?  There are bits and pieces but no system does it all.

Instead the 1.0 vendors try to bolt on new functionality to very old legacy systems.  Epic is based on MUMPS that was developed in 1967.  And they are the leaders.

The next generation 2.0 vendors will disrupt the establishment by focusing exclusively on the physician / care providers and the patient.  And we're seeing examples of this from outfits like Doximity, Practice Fusion, Hello Health and Image32

It will get better.


HIMSS12 - Aventura - location based personas for hospitals

With over 1,000 vendors at the 37,000 attendee HIMSS12, I admit getting a wee bit glassy eyed.  Then a friend showed me Aventura.  It's a virtual desktop for use in hospitals.  Yet this is not just another EMR, it is a framework in which all your applications run.

AaWhat's cool about it, is that it is location aware (through a tap of a RFID card).  Not only does it hook you up to the nearest working printer, but it changes your applications.  For instance it might be great to IM and email in your office, but not in a patient's room.  There you want to see that patient record.  With an extensive rules based engine, it makes sure you have the tools you need, when you need it.

And for those who carry around an iPad, as soon as you swipe your card, your iPad displays just what you need.

But that's only half of the story.  The person who gave the demo was an actual living and breathing developer. 

Why is this important?  Because when your developers actually work with real customers, great things happen to your applications.

iPad Physical Design Flaw

The iPad is a nice looking device, unless you want to hold it, or carry it.   It is not a handheld device, it is a prop-able device.   Aa  With a focal length designed for Steve Jobs' eye sight propped on his folded leg.  You cannot use it flat because of the glare (and it skids) and you cannot hold it without fear of it slipping from your grasp. Steve should have dimpled the shell.

I predict that every iPad user will buy a case to compensate for this flaw which will ensure that the aesthetics will only be seen the first time you take it out of the box.  And kudos to the first person who tells me of a case with a hand strap on the back so I can hold it in landscape mode.

Netbook + Mac Mini: A perfect combination

I wrote an article for PC Week in 1990 about the new Intel 386 SL chip.  This was the first chip made specifically for laptop computers.   And everything changed.  I suspect the Intel Atom chip will change everything again.

Aa For 19 years I have never believed a laptop would be a "desktop replacement".   For me the ergonomics and price factors have always favored a desktop for prolong work.  Which is why people plug in full size screens and keyboards into their notebooks.   But they pay a hefty premium.

When I travel, small and light is good.  But I still need to see what I'm doing and I would like to run "real" applications.  The new crop of 10" netbooks with nine hour battery life does the trick.  But for real work, I like to use a desktop.   Ergonomically and budget wise, it just makes sense.    So rather than dropping $2K on a powerful enough notebook (plus screen and keyboard) you can spend $1K and get a great netbook and a Mac Mini.   I like the mini because 1) it's very quiet and small and 2) it's OSX.   The netbook is XP, but since I run the same application on both platforms, that's o.k.

The key to making this work is Dropbox.  All my current project files are stored in Dropbox, and automatically synched on both machines (plus stored on Amazon S3 for safe keeping).   

And since I have a netbook, I can use a very small flip phone that fits in the change pocket of my jeans.   Since all my email is forwarded to Gmail, I can read and write from any email address.   I admit I'm not as good using T9 as some people, but it works for me.

And as a bonus, the netbook fits in a standard file folder.   Very nice going into meetings.  And here's a secret, you can upgrade those netbooks to 2gb memory for $30 or so.   

So it is possible to have it all and not go broke.   (and yes, that Compaq was my original portable computer)