��������������������������
plicated systems can usually be reduced to independent component parts and the impact of changes in one component can be examined ( Miller & Page , 2007 ). Much of our experimental and analytic armamentarium is devoted to the analysis of such systems , and the search for independent causes ( Pearce , 2006 ; Rothman , & Greenland , 2005 ; Rubin , 2005 ). Complex systems , on the other hand , consist of multiple interacting causes with dense , and sometimes nonlinear , feedback . Altering or dissecting out one component of a complex system dramatically changes the behavior of the system ( Miller & Page , 2007 ). This is a critical distinction and is a critical element motivating building a bridge between the study of complex systems and SDOH .
Complex Systems
There is no agreed-upon definition of
what constitutes a complex system , but there are general characteristics that appear in most definitions . Reference is usually made to systems with multiple parts that interact in relatively simple , but structured , ways while generating system output that is not reducible to the sum of the individual components . First , the elements of a complex system ( people , molecules , nation states , etc .) are heterogeneous with regard to their properties and behavior . Second , current states influence future states and often the dynamics are nonlinear . Third , networks connecting elements of the system can have dramatic effects on the behavior of the system . Fourth , feedback loops allow for learning , adaptation , and modification of behavior . Fifth , what has been called emergennce is often found ( Axelrod , 1997a ; 1997b ). In such cases , small changes can have large and unpredictable consequences , with the whole being greater than the sum of its parts . Finally , stochasticity is the rule .
Importantly , because of the structured interactions , nonlinear dynamics , emergent system behaviors , and other properties characterizing complex systems , the piece-meal dissection of such systems is not likely to be particularly informative , and may even misleading . This is a critical insight for the consideration of complex systems approaches in our SDOH scholarship — in many instances , failure to consider SDOH using a complex systems approach is not only noncontributory , but may potentially be wrong . Informed by this realization , complex systems approaches have been applied to many areas of inquiry , including economic systems , social systems , biological systems , ecological systems , environmental systems , information systems , organizations , geography , psychology , and political strife , and within the last two decades they have been quite successfully applied to infectious disease transmission and more recently are beginning to be applied to topics in social epidemiology and public health ( Auchincloss et al ., 2011 ; Axelrod , 1997 , Axelrod & Epstien , 1997 ; El-Sayed et al ,. 2012 ; Epstein , 2006 ; Galea , Hall , & Kaplan , 2009 ; Galea , Riddle , & Kaplan , 2010 ; Hammond & Ornstein , 2014 ; Lempert , 2002 ; Luke & Stamatakis , 2012 ; Marshall et al ., 2012 ; Ness , Koopman , & Roberts , 2007 ; Yang et al ., 2011 ).
�����������������������������
While there are many approaches to understanding complex systems , agent-based models ( ABMs ) represent a potentially fruitful way of exploring dynamic systems and their behavior where there is interest in both individuals and populations , multi-level and multi-scale determinants , nonrandom patterns of connection between individu-
93