Journal on Policy & Complex Systems Volume 1, Number 2, Fall 2014 | Page 101

��������������������������
measurement styles of various disciplines . The complexity of simulation models of SDOH means that these problems cannot be solved easily by sensitivity testing , as the number of parameters and combinations of parameters becomes unwieldy , although there are some new techniques for sensitivity testing and pattern analysis that may hold promise ( Grignard et al ., 2013 ; Grimm et al ., 2005 ).
There is also a natural tension between the K . I . S . S . (“ keep it simple , stupid ”) ( Axelrod , 1997 ; Axelrod , & Epstein , 1997 ) approach and the desire to build larger , more all-encompassing models that extend to the range of factors covered by SDOH perspectives . This is a tension that is increasingly found in other areas in which agent-based models are being used as well ( Grignard et al ., 2013 ). While simple models , sometimes called “ toy ” models , have advantages as it is easier to see how they work and they can be informative in their simplicity , they lack the appeal of being able to address the tangled web of causation that characterizes SDOH ( Galea et al ., 2009 ). There are issues of model size — number of agents — as well . Basic phenomenon of emergence and nonlinearity can be demonstrated with models with a small number of agents , but to adequately and realistically address agent and environment heterogeneity may require larger population of agents . In principal this is not a computing challenge , as ABMs involving millions of agents are possible .
Good model design , a solid basis for estimating parameters , and careful sensitivity testing are all critical elements in applying the complex systems toolbox to the SDOH . But , there are critical training needs as well . Because of the breadth of factors being considered , understanding the SDOH at its best is an interdisciplinary effort , and while there are now many such interdisciplinary groups of researchers , these groups seldom include those with background in the modeling of complex systems . Likewise , very few of those trained in such techniques have had any training in the understanding of SDOH . While attempts have been made to bridge such divides ( Institute for Systems Science and Health and the Network on Inequality , Complexity & Health both sponsored by NIH / OBSSR , Agent-Based Modeling Bootcamp and Incubator for Health Researchers at the University of Saskatchewan ; MPH curriculum redesign at Columbia University ) ( Agent-Based Modeling Bootcamp for Health Researchers , n . d .; Begg , Galea , Bayer , Walker , & Fried , 2014 ; Institute for Systems Science and Health , n . d .; Network on Inequality , Complexity & Health , n . d .), much work remains to be done in creating pathways for success in integrating these two approaches .
Creating well-designed , well-documented , and informative simulations of the complex nature of SDOH is a big challenge , but there are additional challenges to the funding of such efforts and the acceptance , publication , and use of the results from such simulation models . While there has been considerable progress made in the acceptance of the importance of SDOH in the last few decades , the use of ABMs and other complex systems techniques to study SDOH is in its infancy . Acceptance of new perspectives and techniques in a scholarly enterprise is never easy ( Gebbie , Rosenstock , & Hernandez , 2003 ; Sorenson & Bialek , 1993 ), but there are some steps that can be taken . First , we need to acknowledge that there is considerable interest in training in this area . At a recent NIH conference on Complex Systems , Health Disparities and Population Health organized by the Network on Inequality , Complexity & Health ( http :// conferences . thehillgroup . com / UMich / complexity-disparities-popu-
99