We know that general practices and individual general practitioners (GPs) differ in the frequency with which they prescribe drugs and request blood tests. Previous studies suggest practices tend to be consistently high or low prescribers across of drugs for quite different conditions. For example antibiotics and sleeping tablets. This is partly because general practices care for different types of patients. But GPs and practices also seem to have different preferences for prescribing and requesting blood tests. It is important to understand this because some drugs should be prescribed more and others less. It is also possible to have too many or too few blood tests.
In this research we aim to understand differences in practice preferences for prescribing and requesting blood tests. We will calculate the number of prescriptions of different types of drugs and the number of blood tests undertaken in different general practices. We will apply statistical and computing analysis methods to investigate variation in patterns of prescribing and blood tests between different GPs and general practices. We will determine how much of the variation is due to differences in the patients they care for and how much is due to differences in practice preferences. We will investigate if general practices’ patterns of prescribing and blood test are consistent over time. We will try to find consistent patterns in practices’ prescribing and blood tests.
Although prescribing and requests for diagnostic tests increase yearly, there is variation between general practices preferences for prescribing and diagnostic tests. The aim of this research is to understand general practices’ and GPs’ preferences for prescribing and requesting diagnostic tests.
We will analyse of rates of prescribing and diagnostic test ordering across the entire CPRD dataset. Analysis will take place at two levels, general practice and individual prescriber (GP). We will group prescriptions into drug categories using British National Formulary (BNF) chapter and calculate age, sex, deprivation and morbidity standardised prescribing rates for each BNF chapter. We will similarly group diagnostic tests, into broad categories (e.g. liver function tests, full blood count etc) and calculate, age, sex, deprivation and morbidity standardised rates of test ordering.
To investigate variation in prescribing between individual GPs, we need to attribute the prescribing or diagnostic test decision to the GP consulted. We will identify consultations for common conditions where a first prescription or a diagnostic test may occur. Prescription and consultation dyads will be identified for each of the main BNF chapters, e.g. proton pump inhibitors for dyspepsia, antidepressants for depression, statins for high cardiovascular risk etc. Equivalent consultations and diagnostic test dyads will be identified for each of the main category of diagnostic tests (e.g. full blood count and tiredness; spine X-ray and back pain).
At the general practice and individual GP level we will investigate stability of prescribing rates and rates of diagnostic test ordering over time. We will investigate cross-correlations between rates of prescribing and diagnostic test ordering using dimensionality reduction methods (principal component analysis). We will investigate relationships between rates of prescribing and diagnostic test ordering and other practice characteristics including practice size, deprivation banding and measured continuity of care.
Number of prescriptions, grouped by BNF chapter, at the general practice level and the individual GP level.
Number of diagnostic tests requested, grouped by categories of test, at the general practice level and the individual GP level.
Tom Marshall - Chief Investigator - University of Birmingham
Doaa Radwan - Corresponding Applicant - University of Birmingham
Brian Willis - Collaborator - University of Birmingham
Practice Level Index of Multiple Deprivation