Predictive analytics uses statistical analysis of big data to predict outcomes.  The data typically includes historical, as well as current observation, with thoughtful consideration of the data fields to be reviewed.

The analyses of this type are often described as "statistical learning" because the prediction model increases in precision with additional new data. The predictive modeling techniques use artificial intelligence to create a prediction profile or algorithm based on past patterns within the data analyzed.  Clinical prediction models have been prevalent, but using predictive modeling to forecast revenue is a relativity new concept.

There are many methods for revenue forecasting, from the simple formula of charges, multiplied by the gross collection rate, then factoring in the payment lag or AR Days.  A more complex way is to build an expected collection rate, based on the contracted fee schedule, reduction for multiple procedure payment, bad debt and free care reduction and specific carrier payment policies.  These are just a few examples, which will produce an expected revenue amount.  However, that happens when there is a delay in payments, sometimes expected, such as January with high deductible plans and not expected. Predictive modeling offers the lens to see those downturns in revenue and plan accordingly.
Getting the prediction profile correctly defined is the first step. Brian Montambault, a data scientist at Physicians Professional Services in Medford, MA explains it this way, "It is imperative to build the different components impacting revenue.  We complied the historical payment data for our universe, going back to the beginning of time." The profile looks at the deviations and the local relationships, such as December's charges in relation to January's payments, as well as growth over time.  The model was tested against historical actual revenue totals, examining the monthly deviations and local similarities with the model improving with each additional revenue totals added.

The predictive model was tested for four months before putting the tool into operations.  The revenue predictions out performed standard forecasting tools, with predictions matching the actuals with under a one percent variance.  The prediction model learns the monthly fluctuations in the revenue cycle, which is important for anticipating changes in the monthly cash flow. Montambault asserts that "Using predictive analytics enables more informed decision-making by taking into account the payment trends identified using a powerful statistical tool."

Diana Allen is the CEO of PPS, the financial services company for Harvard Medical Faculty Physicians at Beth Israel Deaconess Medical Center, Boston, MA.