Rugged landscapes: complexity and implementation science

Abstract

Background: Mis-implementation-defined as failure to successfully implement and continue evidence-based programs-is widespread in public health practice. Yet the causes of this phenomenon are poorly understood.

Methods: We develop an agent-based computational model to explore how complexity hinders effective implementation. The model is adapted from the evolutionary biology literature and incorporates three distinct complexities faced in public health practice: dimensionality, ruggedness, and context-specificity. Agents in the model attempt to solve problems using one of three approaches-Plan-Do-Study-Act (PDSA), evidence-based interventions (EBIs), and evidence-based decision-making (EBDM).

Results: The model demonstrates that the most effective approach to implementation and quality improvement depends on the underlying nature of the problem. Rugged problems are best approached with a combination of PDSA and EBI. Context-specific problems are best approached with EBDM.

Conclusions: The model's results emphasize the importance of adapting one's approach to the characteristics of the problem at hand. Evidence-based decision-making (EBDM), which combines evidence from multiple independent sources with on-the-ground local knowledge, is a particularly potent strategy for implementation and quality improvement.

Keywords: Agent-based modeling; Complexity; Evidence-based decision-making; Mis-implementation.

Fig. 1

Illustrative landscapes. The number of local peaks in the fitness landscape increases with K (N=4 above)

Fig. 2

Universal problems. When K is low, PDSA alone performs well. As K increases, a mixture of PDSA and EBI performs best. (Parameters used to generate figure: N=10 and S=0.)

Fig. 3

Context-specific problems. When S>0, a combination of EBDM and PDSA outperforms all other implementation procedures. (Parameters used to generate figure: N=10 and K=4.)