Investigation of antibiotic prescribing and antibiotic resistance in humans. Katie Stewart Supervised by: Louise Matthews, Marian Scott, Dirk Husmeier and Colin McCowan k.stewart.3@research.gla.ac.uk 18th May 2017 Katie Stewart Influences on antimicrobial resistance 18th May 2017 1 / 1
Introduction Overall PhD project Aims: to develop statistical methodologies to investigate geographic, demographic and environmental risk factors for antibiotic resistance. Existing work and relation to my focus There has been existing work in AMR in humans. My focus: addressing statistical problems and modelling on sparse data, associations between environmental factors and indicators of AMR. Katie Stewart Influences on antimicrobial resistance 18th May 2017 2 / 1
Context of Associations Figure: Woolhouse, M., M. Ward, B. van Bunnik, J. Farrar (2015). Antimicrobial resistance in humans, livestock and the wider environment. Philosophical Transactions of the Royal Society of London B: Biological Sciences 370(1670). Katie Stewart Influences on antimicrobial resistance 18th May 2017 3 / 1
Currently Used Open Access Data GP prescription data for October 2015 - November 2016 for the 963 GP practices in Scotland as of January 2017. Focus on antibiotics which treat UTIs, namely Nitrofurantoin, Trimethoprim, Cefalexin, Aminopenicillin, Ciprofloxacin, Co-amoxiclav. Also have population estimates for Scotland, Scottish Index of Multiple Deprivation, GP Practice population and workforce information. Katie Stewart Influences on antimicrobial resistance 18th May 2017 4 / 1
GP locations in Scotland Figure: GP practice locations within Scotland. Katie Stewart Influences on antimicrobial resistance 18th May 2017 5 / 1
Considerations for analysis Areal unit to be used: Scottish data can be split by health board, intermediate geography or data zone Demographics of an area - eg. population breakdown, deprivation levels, access to services. Smoothing data over all smaller units - prescription data is assigned to the GP practice but the patients are not necessarily living in the same area Katie Stewart Influences on antimicrobial resistance 18th May 2017 6 / 1
Scottish areas Data Zone Intermediate Health Geography Board Number 6976 1279 14 Total containing a GP practice 706 581 14 Average size 500 to 1000 household residents 2500 to 6000 household residents - Katie Stewart Influences on antimicrobial resistance 18th May 2017 7 / 1
Temporal Patterns of Antibiotic Prescriptions in Scotland Figure: Count of antibiotic prescriptions per 100,000 population by Health Board, October 2015 - November 2016. Katie Stewart Influences on antimicrobial resistance 18th May 2017 8 / 1
Trimethoprim:Nitrofurantoin Prescriptions in Scotland Figure: Ratio of Trimethoprim to Nitrofurantoin prescriptions by Health Board, October 2015 - November 2016. Katie Stewart Influences on antimicrobial resistance 18th May 2017 9 / 1
Spatial Maps by Health Board, December 2015 and July 2016 Figure: Count of antibiotic prescriptions per 100,000 population by health board, December 2015 and July 2016. Katie Stewart Influences on antimicrobial resistance 18th May 2017 10 / 1
Next Stages Challenges Data availability and accessibility. Scotland prescription data - choosing an appropriate spatial scale, assigning prescribing data to areas where patients live. Using environmental/agricultural data with human data Future work Investigate other data sources (environmental/agricultural) and how to use these with the existing data. Statistical modelling of AMR in humans. Katie Stewart Influences on antimicrobial resistance 18th May 2017 11 / 1
Statistical modelling Aims: to develop statistical methodologies to investigate geographic, demographic and environmental risk factors for antibiotic resistance. Statistical modelling in order to determine the risks for antimicrobial resistance in humans through data from humans, environment and agriculture. Modelling strategies will need to take account of the spatial structure of the data - eg Conditional Autoregressive Models (CAR models). CAR model random effects: φ N(0, τ 2 [diag(w 1) W ] 1 ) CAR model full conditional distribution: ( n ) φ k φ k, W, τ 2 j=1 N w kj φ j τ n j=1 w, 2 n kj j=1 w kj Katie Stewart Influences on antimicrobial resistance 18th May 2017 12 / 1
Thank you for listening. Katie Stewart Influences on antimicrobial resistance 18th May 2017 13 / 1