If there was a way to accurately and consistently predict which patients were likely to use the emergency room in the upcoming year or incur a preventable chronic disease in the next 10 years, how would such a capability impact patient care? In 2008, the United States will spend approximately 12.5% of its GDP ($1.79 trillion) on the treatment of chronic diseases that, in many cases, could have been avoided if the patient was targeted early for preventive therapy. These diseases will account for more than 70% of all U.S. deaths this year. Ironically, the analytical techniques required to make prediction and prevention-based care a reality already exist. However, these advanced methods are not routinely and consistently applied within the healthcare industry, a circumstance that may be partially rooted in confusion over what analytics are and how best to strategically and sequentially leverage more advanced methods to achieve a competitive advantage in the healthcare marketplace.

"Analytics" is most generally defined as an investigation based on the property of numbers. Functionally defined, analytics is a hierarchy of increasingly sophisticated and insightful methods that include graphical study, descriptive statistics, inferential statistics, predictive modeling, and optimization to derive the most meaning from every data point available for investigation.

Once relevant health data, which may include hospitalization, laboratory, claims and pharmacy information, is accessed and collected, one fundamental assessment technique is to graph and/or plot data to ascertain easily recognizable relationships and areas requiring further investigation that may not have been visually apparent from the raw data. For example, plotting a patient's systolic blood pressure over time may reveal a systemic increase or decrease that, as a result of data volume or random noise, may not have been observable by examining the raw data. Such graphical analysis is often accompanied by calculation of various descriptive measures, which may include sums, averages or frequencies

Within the field of descriptive statistics, Online Analytical Processing (OLAP) is a popular method for rapidly assessing multi-dimensional data. In addition to individual patient information like time to discharge being desired, aggregates of this data, such as average time to discharge for all a physician's, department's or facility's patients may need to be examined. Perhaps, average time to discharge for just the month of October is required. These aggregate summaries involve performing many calculations and are often time-consuming to produce on-demand within large data sets. With OLAP, many of these aggregates are pre-calculated such that they are available almost instantaneously, thereby making descriptive analysis more efficient.

Inferential statistics embody a class of powerful techniques whereby conclusions are drawn about a larger population and its members based on what is found in a random subset from the population. For example, if medication noncompliance is assessed for a small and random sample of a population being considered for enrollment into a managed care plan, it may be possible to use this information to assess noncompliance of the entire population without the need to measure every person within that group. Alternatively, it might be possible to determine with confidence that a particular member or set of members in the group is at risk for noncompliance based on analysis of the population to which the member or set of members belongs. Examples of inferential methods include Analysis of Variance (ANOVA), regression, t-tests, and factor analysis.

Predictive modeling and optimization are generally regarded as data mining techniques. Data mining methods often relax many of the assumptions of inferential statistics, enabling analysis of data that may be noisy, error-prone, or non-conforming to other requirements frequently encountered when using classic inferential technique.