In this evolving world of electronic records and clinical information systems, more and more attention is being given to predictive modeling (PM) as a valuable asset to improving our nation’s healthcare. Perhaps once considered, a method used only by actuaries to identify populations at risk to a health problem, PM is evolving as a means to answer questions as quickly as the data it depends on becomes available.
PM is a process of statistical analysis using historical data to predict future trends and behavior patterns. These techniques are in evidence throughout our daily lives performing calculations to identify patterns and answer questions about performance or guide decisions. Fraud detection, underwriting, marketing, investment management, and clinical decision support rely on PM to identify and guide future actions. These models of course are only as good as the analysis, assumptions and data accuracy used. And in healthcare, historical error rates and data validity present an enormous challenge to accurate prediction of the future.
Last week, CMS announced its program launch of PM to fight Medicare fraud. Beginning July 1, 2011, the Affordable Care Act of 2010 has made available funds to the President’s Campaign to Cut Waste enabling CMS to identify fraudulent claims before they are paid in much the same manner credit card companies detect fraud.
There should be little doubt to the value and knowledge gained from the data of historical practices. All that is required for PM to exist is a fast computer with lots of number crunching capability, a program to query, sorting and analyzing of the data, and individuals with a need to know and ability to answer question of relevant nature about the business in which they are engaged.
For example, consider the time and resources spent in rehabilitation setting FIM™ goals and working to achieve them over a length of inpatient days. Currently, we rely on our experience and training to establish some future target expectation and proceed to plan and deliver therapies to achieve that expectation. Some do it better and more accurately than others, some just guess and hope for the best, while others don’t bother and just treat the patient. PM applied to this process could look back at all patients classified identical to the current patient and calculate the average FIM rating achieved upon discharge for the historical group. Using the historical average of outcome achieved as a basis for comparison, a therapist would immediately consider past performance and outcome for all similar patients and fine tune expectations for the current patient. The planning and resources applied in treatment could be measured against these evidence based expectations rather than guessing.
Now for the 50% of clinicians who set FIM goals accurately and achieve them, PM will provide little value other than to substantiate their good judgments. However, improving the accuracy of the 50% of clinicians who guessed wrong … priceless