Implementation of policies (decisions) in the health sector is sometimes defeated by the system's response to the policy itself. This can lead to counter-intuitive, unanticipated, or more modest effects than expected by those who designed the policy. The health sector fits the characteristics of complex adaptive systems (CAS) and complexity is at the heart of this phenomenon. Anticipating both positive and negative effects of policy decisions, understanding the interests, power and interaction between multiple actors; and planning for the delayed and distal impact of policy decisions are essential for effective decision making in CAS. Failure to appreciate these elements often leads to a series of reductionist approach interventions or 'fixes'. This in turn can initiate a series of negative feedback loops that further complicates the situation over time. In this paper we use a case study of the Additional Duty Hours Allowance (ADHA) policy in Ghana to illustrate these points. Using causal loop diagrams, we unpack the intended and unintended effects of the policy and how these effects evolved over time. The overall goal is to advance our understanding of decision making in complex adaptive systems; and through this process identify some essential elements in formulating, updating and implementing health policy that can help to improve attainment of desired outcomes and minimize negative unintended effects.

Background: Change of first-line treatment of uncomplicated malaria to artemisinin-combination therapy (ACT) is widespread in Africa. To expand knowledge of safety profiles of ACT, pharmacovigilance activities are included in the implementation process of therapy changes. Ghana implemented first-line therapy of artesunate-amodiaquine in 2005. Drug utilization data is an important component of determining drug safety, and this paper describes how anti-malarials were prescribed within a prospective pharmacovigilance study in Ghana following anti-malarial treatment policy change.

The CEM methodology is a useful tool for monitoring the safety of widely available and utilized medicines, particularly in an urban environment where spontaneous reporting yields poor results and where the availability of various regimens and high levels of medicine usage can give valuable 'real-life' safety data. The types and frequencies of events reported reflected the types of events expected in patients prescribed antimalarials and nearly all events reported are listed in the summary of product characteristics of the medicines involved.