Message Font: Serif | Sans-Serif

No. of Recommendations: 2
JLC,

Flash back even further to my college statistics class where we were learning methods to access the accuracy of weather predictions. When the weatherman says 30% chance of rain, how accurate is he if it rains 1/4 days vs 1/3?

</snip>

Weather forecasts have improved immensely since you were in college (e.g., more satellites to collect data, more supercomputers to process it.)

Big Data will find a lot of low-hanging fruit in health care if Google has an economic way to read and analyze paper records.

intercst
No. of Recommendations: 3
An new type of algorithm...

Way back in residency training my ICU attending was working on the APACHE Scoring System, a mortality/morbidity predictor using current lab data and physical assessments. The purpose was to help evaluate the usefulness of continued care.

Flash back even further to my college statistics class where we were learning methods to access the accuracy of weather predictions. When the weatherman says 30% chance of rain, how accurate is he if it rains 1/4 days vs 1/3?

So my question, you get an estimate of 19% death, how accurate is it really?

JLC, who learned long ago that it is hard to put numbers on death whether it be in time or chances.
No. of Recommendations: 2
JLC,

Flash back even further to my college statistics class where we were learning methods to access the accuracy of weather predictions. When the weatherman says 30% chance of rain, how accurate is he if it rains 1/4 days vs 1/3?

</snip>

Weather forecasts have improved immensely since you were in college (e.g., more satellites to collect data, more supercomputers to process it.)

Big Data will find a lot of low-hanging fruit in health care if Google has an economic way to read and analyze paper records.

intercst
No. of Recommendations: 4
JLC, who learned long ago that it is hard to put numbers on death whether it be in time or chances.

My prediction: 100% chance.
No. of Recommendations: 1
Google had created a tool that could forecast a host of patient outcomes, including how long people may stay in hospitals, their odds of re-admission and chances they will soon die.

That reminds me of something we covered in a Robust Engineering Green Belt course. That was similar to 6-Sigma Green Belt courses leading to a Black Belt certification, but more focused on Taguchi's work for robust design.

One case study was the analysis of the best way to assess if a patient that arrived at the emergency room with heart attack symptoms was really having a heart attack, or it it was something else. Hospitals ran tons of tests and used many patient factors in the determination. It turned out that you could get the most accurate assessment by using two tests and the interaction between those tests (that is, the results of the two tests, plus a calculation based on the two tests combined) than all the other stuff. It was odd that patient weight didn't matter. (I suppose obesity may increase your chance of getting a heart attack some day, but was not significant in determining if you are having one right now.)

So a computer program can use factors that have mattered in the past, and not use things that don't matter but people still consider because they seem like they *should* matter.