Measuring prevalence, reliability and variation in high-risk prescribing in general practice using multilevel modelling of observational data in a population database

摘要: Measuring prevalence, reliability and variation in high-risk prescribing general practice using multilevel modelling of observational data a population database Bruce Guthrie,1* Ning Yu,2,3 Douglas Murphy,1 Peter T Donnan1 Tobias Dreischulte2 1Quality, Safety Informatics Research Group, Population Health Sciences Division, Medical Institute, University Dundee, UK 2Tayside Medicine Unit, NHS Tayside, 3Institute Epidemiology Health, Faculty Sciences, College London, *Corresponding author b.guthrie@dundee.ac.uk Background: High-risk primary care is common known to vary considerably between practices, but the extent which varies among individual practitioners (GPs) not known. Objectives: To create safety indicators usable existing electronic clinical examine (1) patients, GPs practices including measurement (2) changes over time prevalence practices. Design: Descriptive analysis logistic regression routine data. Setting: medical record Participants: For reliability, 398 26,539 patients 38 Scottish change prescribing, ≈ 300,000 particularly vulnerable adverse drug effects registered with 190 Main outcome measures: GPs, five non-steroidal anti-inflammatory (NSAID) prescribing. 19 previously validated indicators. Results: Measurement at GP level was feasible only for newly initiated drugs similar NSAIDs are usually by GPs. There moderate total NSAID [intraclass correlation coefficient (ICC) 0.034], this indicator highly reliable (> 0.8 all practices) distinguishing because large number being measured. initiation (ICC 0.055) larger 0.166), did reliably distinguish had > 0.7 half study. Between quarter (Q)2 2004 Q1 2009, percentage exposed measured 17 that could be examined whole period fell from 8.5% 5.2%, largely driven reductions antiplatelet use. Variation increased decreased five, no relationship rate DOI: 10.3310/hsdr03420 HEALTH SERVICES AND DELIVERY RESEARCH 2015 VOL. 3 NO. 42 © Queen’s Printer Controller HMSO 2015. This work produced Guthrie et al. under terms commissioning contract issued Secretary State Health. issue may freely reproduced purposes private research study extracts (or indeed, full report) included professional journals provided suitable acknowledgement made reproduction associated any form advertising. Applications commercial should addressed to: NIHR Journals Library, National Institute Research, Evaluation, Trials Studies Coordinating Centre, Alpha House, Southampton Science Park, SO16 7NS, UK. v Conclusions: moderately cannot easily routinely difficulties accurately identifying actually prescribed often shared responsibility specialists. initiation, there approximately three times greater than Most above average worked were themselves average. The observed 2009 falls These results consistent improvement interventions more appropriate targeted on or higher need understand why design evaluate reduce it. Funding: Funding Services Delivery programme Research. ABSTRACT Library www.journalslibrary.nihr.ac.uk vi