Dianna Smith

Expertise in small-area estimation of health outcomes and inequalities, teaching experience in MSc-level public health and undergraduate, MSc in geography.

My masters, doctoral and some postdoc research focused on the spatial analysis of food access and health outcomes, including the identification of ‘food deserts’: areas with poor access to healthy and affordable food. These previous studies have contributed to my primary research goal of facilitating a more widespread understanding of the measurable extent of social inequalities, with a focus on the basic needs of individuals. An additional area of research and collaboration is food poverty risk and provision of emergency food through food banks.

I collaborated with colleagues at QMUL to explore different methods of geographic data visualisation, which has led to three papers published and in preparation. We have focussed on the different mapping styles available for health data (choropleth, interpolated [smoothed], point data) and applied this to diabetes risk in East London. Non-academic outputs included Diabetes Risk Report which led to discussions with a national charity about using the maps as a way of identifying high risk people in the area for an educational intervention. Some maps were used by the local Director of Public Health as part of highlighting health needs to the Local Authority. The local acute trust incorporated some of the maps in their new public health strategy, and some of the maps were presented to the Board of the acute trust, which helped galvanise support for tackling wider determinants of health in the local area.

My doctoral thesis necessitated the creation of small-area (output area (OA)) population estimates of diabetes and adult obesity prevalence. One of the main outcomes of this research was a unique methodology that creates area-specific models to best reflect the heterogeneous OAs in the study area. Typically, prevalence estimates are based on a ‘global’ model configuration. SimHealth instead uses different model configurations to more accurately represent the diverse areas, and incorporates a unique clustering step to improve model estimates.

Throughout my MRC-funded postdoctoral fellowship I have extended this work with a holistic approach. Building on the 2009 methods paper which showed the advantages of the adapted methodology, I have applied the method to estimate mental health and alcohol consumption across all of England at the Lower and Middle Super Output Area (see Riva & Smith, 2012). Most importantly, I successfully validated the modelling methodology using smoking data in New Zealand (Smith et al 2011) and intend to tested this method against a different estimation approach with colleagues at the University of Leeds (Harland et al 2012).