Type 2 diabetes medications lower blood sugar to stop patients from developing health problems, but they only work if taken regularly. We have previously shown that patients who do not take enough of their medication, as measured by the number of prescriptions issued by their doctor, have worse control of their blood sugar than patients who get regular prescriptions. Therefore, it is really important for the doctor to determine which patients are more likely to be at risk for not picking up their prescriptions, to help address this issue.
We predict that factors such as whether previous prescriptions are collected on time, patient clinical features (e.g. age, gender), other tablets, and other conditions, may help predict how regularly patients pick up prescriptions. We aim to develop a statistical model that will take information available from a patientÂ’s clinical record, to predict how likely they are to regularly pick up prescriptions for their diabetes medications. The aim is to inform future work to enable patients at risk of having problems to be flagged up on GP computer systems. This will enable the doctor to introduce systems to help e.g. by sending reminder messages to mobile phones and similar digital devices.
Adherence to diabetes medications is a clear predictor of glycaemic response: patients with <80% prescription coverage have a worse response 6 months after starting therapy. It is therefore important for the clinician to determine patients at risk of non-adherence to enable delivery of interventions aimed at improving agreed medicines use.
We aim to investigate factors that predict adherence to medicines, assessed by prescriptions issued, including clinical features (age, BMI, gender, ethnicity), socioeconomic status, number of diabetes or other routine medications, and comorbidities. We will also study patterns of past prescription coverage on first line metformin treatment to assess whether this is a predictor of future adherence on second and third line therapies. Initial analysis will focus on predicting adherence as measured by medical possession ratio over the first year of treatment, but additional analysis will investigate if predictors change when studying other measures such as persistence of prescription pick-up. We will restrict our analysis based on patients starting a new diabetes therapy from 2004 onwards, allowing for higher quality data (when QOF returns came into practice) and reflecting modern prescribing practices.
We will develop statistical models of predictors of adherence using regression techniques (linear when examining predictors of adherence on a continuous scale, logistic when examining predictors of <80% v >80% adherence). The long-term aim will be to implement such models in clinical practice to flag-up patients requiring additional support.
Primary: 12 month Medical possession ratio on second line diabetes therapy
Secondary: 12 month persistence
Andrew Farmer - Chief Investigator - University of Oxford
Beverley Shields - Corresponding Applicant - University of Exeter
Andrew Hattersley - Collaborator - University of Exeter
HES Admitted Patient Care;Patient Level Index of Multiple Deprivation