We all use predictive analytics and when looking a medical claims to find patterns of abuse and fraud, this is good thing and why HHS has incorporated this into the processing of Medicare claims. Those patterns are there to find but all keep in mind depending on how much the parameters of these queries are utilized, imageyou can also get some false positives and the company through many years has had the time and expertise to look at those.  Data is data an depending on how the queries run and what they look for, that determines the results you see.  You will never have 100% in this area by any means.  If anyone could completely get rid of false positives in this area, everyone would be knocking on their door but this is analytics with humans, so it will never be, but can be improved.

Health Insurance Underwriting Practices With Prescription Data – How Does This Work


The company has had success with predicting for many years and profits from it very well.  The old Ingenix MedPoint Prescription Profiling has been bringing in money for years and is used for underwriting policies, again to assess risk.  I’ll repeat again such numbers and information should always be used in context.  When you go non linear with some of the models that are out there you create “relations” in data that really do not exist and thus the decision to go this route or stay with linear data comparisons is always a question.  The old data base that was used for 15 years to short pay doctors and hospitals was also in a way predictive analytics as when United and all the other major insurers used it, they could see and predict what their expenses would be in this area for percentages of out of network charges based on “known” pricing factors that were already determined.  Of course later this didn’t work as New York found those analytics were faulty and resulted in short payments and the AMA finished with that class action suit not too long ago.  Aetna recently announced their settlements in this area as they also took advantage of the predictive analytics built into the out of network charges. 

United is not the only company touting analytics on steroids today and again there’s some value in some of what they say, but the savings numbers are all out of whack as there’s economists around today to predict this stuff either.  It kind of looks like economists are slowly becoming a dying breed as they just relay on the numbers the Quants model these days.  Paul Wilmott makes that pretty clear in the video documenting how Quants work today and you can view it in the left hand side of this blog, an eye opener on how the progressions of Quants and their technologies has evolved.  It’s much more difficult to be one today as everyone’s in on the game.  In addition we have this other agency out there, the MIB  that collects and markets all kinds of insurance information.  You will find almost any type of insurance company subscribing to their data and were basically the beginning of insurance data collecting for underwriting. 

The MIB – Health Insurance Bureau Business Intelligence Mining May Go Beyond Just Healthcare Information

“MIB, Inc. is the premiere provider of fraud detection information for individually underwritten life, disability income, long term care and critical illness insurance.  MIB member companies rely on its Checking Service for the fast, secure aggregation and  exchange of data to combat fraud, improve underwriting effectiveness and increase product line profitability while ensuring fair pricing for all applicants.”

If you look a little further here with the MIB you can see where they are now marketing data that gives a risk assessment on how long they anticipate you will live.  So again we are not short on analytics out there today and we will continue to see reports and studies promising billions in savings but it’s just not there in those amounts as companies tout.  Sure there are savings to be had but everyone wanting a sale is exploiting numbers beyond belief so don’t get “Algo Duped” with some of this.  Think of the stock market if you will with algos and think of the Knight Capital case too and see how algorithms can quickly make or break a company and we do have rogue algorithms running around out there today and one of those can shoot a good portion of analytics right down the tubes, so again these savings numbers are only as good at the day they are put out there as complexities and the uncertainties of how algorithms work will keep it that way. 

MIB Solutions and Hooper Holmes Working Together to Assess Mortality Risk – Analytics and Consumer Files Used for Underwriting And To Estimate How Long You May Live And What Your Body Will Cost Over Time

Again I am not saying that there’s not room for improvement as there is, but nobody and I mean nobody can accurately predict some of this as the money you save today might be spent on a new drug, device or crisis tomorrow and that is the unknown…better know as life. The MIB has their write up on diabetes too with predictive algorithms and risk assessment so they can take your numbers and mxi them with other data they have around and supposedly are able to predict about when you will die or how many years you have left, but the human body has been known to fool many, so again not 100%. 

On the readmissions, UCSF is already doing what is outlined in this report, so what they are doing is using medical records to work with patients and using humans to make sure the patient gets the treatments and care they need, and an EMR did this, which is where clinical care is, and by such use they can pretty much develop the best care methodologies with clinical data, unlike the outrageously quoted cost savings like in this report from United.  Medical records have been using predictive analytics since they have been around but it has been called this:)  In other words it gives the clinicians the information they need to evaluate patient care.  Let’s model their example and give all hospitals a grant to do the same. 

UCSF Medical Center Reduced Readmissions With Heart Failure And It Was Not That Magical Flipping Algorithm Everyone Is Looking for That Did the Trick


So with all this being said about models and predictions I think it’s time I hussle over and get a copy of Emanuel Derman’s book “Models Behaving Badly” , which was reviewed here by the Wall Street Journal, to see what else I might learn about modeling.  Here’s a few paragraphs of what was said in the review…

”He then migrated to the center of the financial world in the 1980s, using a mix of mathematics and statistics to value securities for the trading desk at Goldman Sachs in New York. He had hoped to use the methods of physics to build a grand, unified theory of security pricing. After 20 years on Wall Street, even before the meltdown, he became a disbeliever.

The basic problem, according to Mr. Derman, is that "in physics you're playing against God, and He doesn't change His laws very often. In finance, you're playing against God's creatures." And God's creatures use "their ephemeral opinions" to value assets. Moreover, most financial models "fail to reflect the complex reality of the world around them."

It is hard to argue with this basic thesis. Nevertheless, Mr. Derman is perhaps a bit too harsh when he describes EMM—the so-called Efficient Market Model. EMM does not, as he claims, imply that prices are always correct and that price always equals value. Prices are always wrong. What EMM says is that we can never be sure if prices are too high or too low.”

Enough said for me here with these 3 paragraphs and he’s the expert on models and humans as “God’s creatures”…as it’s true and especially in healthcare so we need to keep that human element alive and well in healthcare.   We had a small Twitter chat where he stated that only medical doctors should carry the “MD” credentials after their names and PHD folks should not be in that arena:)  I tend to agree with that too.  Let’s face it we are loosing a bit of this as why else would the AMA come out with this statement…

AMA Reaches Out to Doctors To Remind Them Patient Welfare Must Come First As Rising Pressures From Insurers and Hospitals Can Surmount At Times

I also wonder if their analytics are so right on target, how this happen with the AAFP confronting them on the fact that in many areas of the US through complex contracts that they are paying doctors less than Medicare?  Is that legal?  I don’t know but United said they were surprised this was happening?
 


So again when reading reports and looking at models in healthcare, I think Mr. Derman has some good advice to offer and keep this in mind when you see what all are touting today with analytics and savings that are just a projection and again in agreement with him the models do not reflect the complex reality of the world around them…again take a hard look at what you projected in healthcare.  Here’s the latest below on their latest report and you can read the article and link below on what United claims with yet another model for Medicare and Medicaid and it kind of sounds like they are really clamoring for more business from CMS and Medicare?  BD 


“UnitedHealth touts predictive modeling as solution to healthcare fraud and preventable hospitalizations”


http://www.unitedhealthgroup.com/hrm/UNH_WorkingPaper9.pdf

http://www.healthcarefinancenews.com/news/predictive-modeling-solution-healthcare-fraud-and-preventable-hospitalizations?topic=04,14,19

0 comments :

Post a Comment

 
Top
Google Analytics Alternative