Using flawed, uncertain, proximate and sparse (FUPS) data in the context of complexity: learning from the case of child mental health

Miranda Wolpert, Harry Rutter

The use of routinely collected data that are flawed and limited to inform service development in healthcare systems needs to be considered, both theoretically and practically, given the reality in many areas of healthcare that only poor-quality data are available for use in complex adaptive systems. Data may be compromised in a range of ways. They may be flawed, due to missing or erroneously recorded entries; uncertain, due to differences in how data items are rated or conceptualised; proximate, in that data items are a proxy for key issues of concern; and sparse, in that a low volume of cases within key subgroups may limit the possibility of statistical inference. The term 'FUPS' is proposed to describe these flawed, uncertain, proximate and sparse datasets. Many of the systems that seek to use FUPS data may be characterised as dynamic and complex, involving a wide range of agents whose actions impact on each other in reverberating ways, leading to feedback and adaptation. The literature on the use of routinely collected data in healthcare is often implicitly premised on the availability of high-quality data to be used in complicated but not necessarily complex systems. This paper presents an example of the use of a FUPS dataset in the complex system of child mental healthcare. The dataset comprised routinely collected data from services that were part of a national service transformation initiative in child mental health from 2011 to 2015. The paper explores the use of this FUPS dataset to support meaningful dialogue between key stakeholders, including service providers, funders and users, in relation to outcomes of services. There is a particular focus on the potential for service improvement and learning. The issues raised and principles for practice suggested have relevance for other health communities that similarly face the dilemma of how to address the gap between the ideal of comprehensive clear data used in complicated, but not complex, contexts, and the reality of FUPS data in the context of complexity.