| |
In accounting terms, current healthcare analytics is about using the General Ledger, to spot the highest and lowest cost items. The use of quick ratios and cash flow are yet to come. Even words such as "predictive analytics" mean nothing more than using Farmer's Almanac to predict weather. Currently correlations and historic data are used in regression relations to predict the future, even when the underlying mathematics restricts extrapolation of these results to the future.
|
|
| |
| |
Economists found a way out of this quandary by switching to dynamic modeling using sophisticated mathematics of game theory and linear programming. In 2004 we opted to mimic economists and try to model chronic populations as a dynamical system to predict the time evolution of the population. In other words, we hoped to predict the numbers of people in a population that will develop various chronic diseases and cancer. Since over 75% of all costs go to treatment of these diseases, we expected to be able to forecast total costs accurately as well. |
|
| |
| |
Technical barriers |
|
| |
| |
Without going into the details, we drew inspiration from a 225 year-old equation and borrowed the mathematics used in describing a cloudless sky to formulate the time evolution of chronic populations as a matrix differential equation. |
|
| |
| |
In the process we had to answer some basic questions. Is it possible to prove if a Medicaid program is beneficial for the state's enrollees? How does one mathematically describe the health of a population? How about changes in the health of a population? How does one describe cost of healthcare? |
|
| |
| |
It took 5 years to complete the work and another 5 years before a patent was awarded. We validated the results by analyzing Medicaid data in New Mexico for the period 2010-2014. The 12-month forecasts of disease spread had less than 5% error and the costs were in agreement too, except for 2014. |
|
| |
| |
CareMaps - To improve care, lower costs and mitigate fraud and abuse |
|
| |
| |
The CareMaps analysis yields three important results. (i) It shows the strengths and weaknesses of the various prevention programs in the state (ii) It shows the variations in premiums throughout the state in terms of treatment costs of chronic diseases and cancer and (iii) It reveals regions with anomalous costs that are usually precursors to fraud and abuse. |
|
| |
