A methodology to access and analyse routine clinical data in Scotland (SCI-Diabetes) to investigate the effect of organizational arrangements on the care of people with type 2 diabetes and diabetic foot ulcers

Introduction. Diabetic foot ulcers (DFU) are a devastating complication in people with type 2 diabetes (T2D). Lower extremity amputations (LEA) have been used as a primary outcome to evaluate health care provided to people affected by DFU. Rates of LEA can highlight trends in relation to socio economic differences, geographic variation and different organizational settings.A meta-analysis conducted by our group showed that the impact of LEA rates can bepartially explained by organizational arrangements which reduce their incidence e.g. multidisciplinary teams, dedicated teams, pathways of care or combined complexinterventions. It is not clear whether such differences can be captured by large clinicaldatabases collected from everyday practice. In the framework of the EU HealthPros project, we conducted an empiric study usingnational data from Scotland to adapt principles extracted from the literature for predictive modelling. We aim to identify a data model for applying the definition of organizational arrangementsassociated with the reduction of LEA among people with T2D observed in the literature, tothe data stored in the Scottish diabetes register (SCI-Diabetes). To build a predictive model investigating the association between structures, processes of care, type of services provided and LEA rates, while adjusting for personal case-mix characteristics. To compareand validate the model on different databases for national & international comparisons.Methods. We developed a data mapping algorithm to translate a set of criteria derived from the literature into source code applied to an extract of SCI-Diabetes. We compared the local data dictionary to the contents of a national database of general practitioners in England, todefine a set of data elements that could be reused in different contexts. We used the standard set from the International Consortium for Health Outcomes Measurement for further generalisation. The algorithm was compared with the views of technical and clinical experts at the coordinating centre. A basic set of characteristics at cluster (e.g. rural vsurban, geographical, deprivation index) and individual levels (e.g. age, gender) were identified for categorisation and inclusion in the statistical predictive model. Statistical analysis involves multivariate GEE logistic, multilevel and Cox models. Results. The algorithm allows identifying a cohort of patients whose pathways of care can be followed from diagnosis to any LEA event and/or last visit. Through the extracted database, the experience of the cohort is followed from diagnosis of DFU to the LEA event, or the lastobservation without event. The mapping tool and model estimation are still in progress and will be further refined by the final presentation at the Conference. The results will be compared to the output of a cross sectional analysis of the dataset of people with T2D of the Royal College of General Practitioners Research and Surveillance Centre in England.Conclusion. We developed a methodology to investigate and compare the effect of organizational arrangements on the trajectory of the disease from the occurrence of DFU up to any LEA. Our study documents the method, criteria and algorithm to investigate the variability of outcomes across centres involved with the HealthPros project, starting from England,Denmark and Germany.