Over the last few years, the triple aim has taken center stage in health care. Through more effective identification of individuals at higher risk, health care systems can become more strategic about resource allocation in order to achieve the triple aim. A tool called predictive analytics has created opportunities for customized prediction and relative risk scores to achieve this very goal. Most initiatives that utilize predictive analytics, thus far, have only concentrated on one side of the bell curve, fixating on over-utilizers of health care services. Although this is an important field of study that further helps identify risk factors and obstacles to care, it may not be a complete picture of opportunities for resourceful quality improvement within a population. Other great opportunities lie on the other side of the bell curve through the "Positive Deviance" approach to quality improvement. Positive deviants are the high-risk individuals whose uncommon behaviors and methods enable them to find better solutions to problems than their high health care-utilizing peers, while having access to the same resources and facing similar or worse challenges. Positive deviance enables the community to discover these successful behaviors and methods and develop a plan of action to promote their adoption by the high risk population and health care system at large.  Improvement may be better achieved by looking at both sides of the curve.
Predictive analytics (PA) is a term that describes a plethora of statistical and analytical techniques that investigate current and historical data to make predictions about the future.  PA utilizes these techniques to identify patterns in the data associated with a specific endpoint. Once the data is analyzed, a weighted formula is created from the recognized data patterns. These formulas can be used for the creation and application of more effective predictive risk scores that would enable health care providers to identify patients in need more accurately. Without this tool, many patients that are at an increased risk may be overlooked, and opportunities may be lost to apply preventive measures. This tool creates timely, actionable opportunities to make real change in daily patient interactions. By harnessing probabilistic prediction power from a diverse set of data sources, including home grown, community specific data, it is plausible to change the landscape of how we practice patient care through the lens of population health.
Predictive analytics has also established its utility in governmental public health. The Chicago Department of Public Health is leading several predictive analytics initiatives in partnership with the City of Chicago Department of Innovation and Technology (DoIT) and civic-minded organizations. At DoIT, their work in this area has predicted the likelihood rodent complaints and aids crews at the Department of Streets and Sanitation in preventive baiting. The pilot program thus far has improved preventive rodent baiting in the city by eighteen percent.  Foodborne CHI, managed by Chicago Department of Public Health and the Smart Chicago Collaborative, is a platform that is using Twitter, machine learning, and big data analytics to improve inspections, predict outbreaks, and improve the health of residents.
In New York City's Office of Policy and Strategic Planning, predictive analytics has been used in multiple areas to improve outcomes. They have created a "five-fold return" on building inspectors' time looking for illegal apartments that have an increased risk of catastrophic events.  Predictive analytics can be a game changer for local health departments.
Health System Framework
For the individual patient being admitted to the hospital, predictive analytics has already started to impact on care. By identifying individual patients who are at higher risk for readmission in a timely manner, there are opportunities for intervention. This invaluable insight enables care team members to create a more productive impact through the prioritization of those in greatest need. This notion has already played out at Parkland Hospital in Dallas, Texas, by Dr. Rubin Amarasingham and colleagues. They created a predictive analytics system called the Parkland Intelligent E-Coordination and Evaluation System, or PIECES, that has helped target individual patients to get the care they need. PIECES identifies high risk patients by analyzing information available through the hospital's electronic health record. Once these higher-risk patients were identified, they were assigned as high priority for the hospital's case managing staff. Since PIECES has been running, Parkland hospital has cut its 30-day readmissions for Medicare patients with heart failure by 26 percent with no increase in staffing. 
Positive Deviant Paradigm Shift
The lessons learned thus far show that even a small improvement in prediction can have a significant impact in how to allocate resources. What we have yet to explore is how that small improvement could be used to enhance the positive deviance approach for patient care. The use of positive deviance has been shown to have great potential for improving quality of care in the health care setting.  The synergy between predictive analytics and positive deviance creates a new approach to health care delivery that is currently untapped. This approach could help open the door to explore how positive deviants traverse through the social determinants of health in order to sustain their well being. When studying social determinants of health, researchers have shown that health care accounts for only 10-25 percent of the variance in health over time.  The two fields of predictive analytics and positive deviance may work hand in hand by providing an effective approach for studying and spreading local behaviors that have been proven to overcome social determinants of health.
As we move towards a value-based health care system, the identification of agreed upon quality indices is paramount. Through this approach there may be opportunities to develop new quality indices based on these identified positive deviant behaviors. The behaviors identified could help customize quality metrics based on measurable outcomes that are contextually specific to each care system. As we move towards partnering with patients to create quality metrics, this could serve as a more evidence based approach to achieve this goal.
Conflict of interest disclosures: Neither author has any conflicts of interests to disclose.
1) "What Is Positive Deviance?" Positive Deviance Initiative. N.p., n.d. Web. 09 Dec. 2013.
2) Nyce, Charles (2007), Predictive Analytics White Paper, American Institute for Chartered Property Casualty Underwriters/Insurance Institute of America, p. 1
3) United States. City of Chicago. Office of the Mayer. City of Chicago Launches Preventive Rodent Baiting Pilot. By Molly Poppe. Chicago: City of Chicago Office of the Mayer, 2013. Print.
4) Howard, Alex. "Predictive Data Analytics Is Saving Lives and Taxpayer Dollars in New York City." Strata. O'Reilly: Strata Making Data Work, 26 June 2012. Web. 28 Sept. 2013.
5) Amarasingham R, Patel P, Toto K, et al. Allocating scarce resources in real-time to reduce heart failure readmissions: a prospective, controlled study. BMJ Qual Safdoi:10.1136/bmjqs-2013-001901
6) Bradley, E., Curry, L., Ramanadhan, S., Rowe, L., Nembhard, I. and Krumholz, H. 2009. Research in action: using positive deviance to improve quality of health care. Implementation Science, 4 (1), p. 25.
7) Institute for Alternative Futures. Community Health Centers Leveraging the Social Determinants of Health. Alexandria, VA. March 2012. Available from http://www.altfutures.org/pubs/leveragingSDH/IAF-CHCsLeveragingSDH.pdf.
For more information please visit: http://www.positivedeviance.org/